Slashdot Mirror
Experiences of Running Linux on a Mainframe
xneilj writes, "Linuxplanet has an interesting article where a guy decided to install the native Linux on the company mainframe in their lunch hour. Interesting article if you're wondering why anybody would pay seven figures for a box when you can get a high-end pc machine for a fraction of that. Author Scott Courtney reckons if you put Linux on it 'the mainframe of today may in fact be the best damned Web server you ever saw.' "
Great job! I think this qualifies as "Toll of the Day". Keep up the good work, dmg.
Yes the minis were getting so powerful, and now the PC's are getting so powerful. What the Pee Cee Wee Nees don't seem to get, is that Moore's law applies to mainframes just as much as to Pee Cees.
Lakehead University, in Thunder Bay Ontario (Canada) purchased a brand new CRAY system (in conjunction with 4 other Canadian Universities) Instead of paying several hundred thousand for Cray's OS. They did some research into alternatives. After about three months of study, they decided to install Debian Linux on it instead.
So does the IRS
And that UDMA/66 HDD beats the hell outah all the DASD connected to the mainframe.
Looking at the replies above this I can't help but laugh. How could anyone have taken this seriously? Only a person who considers himself 1000x smarter than everyone else could believe that anything so ridiculous was in earnest. This fantasy of genius coupled with the reality of dull-wittedness is what allows trolls to succeed. Fortunately, Slashdot is full of such high-ego/low-wattage folks.
Even if he called you a jerk, you cannot go around threatening to kill people. Not only is it illegal, it makes you look bad.
Note to Rob: allow readers to weed out comments by poster name.
Oh, great, a psuedohacker.
Indicators of a psuedohacker:
* Mentions Beowulf in a blatantly incorrect manner.
* Has a quote from 1984 in his sig.
Just think about the letters "dmg" for a while.
Thank you.
Bzzt wrong. . .but thank you for playing.
IBM had a system 32, 34, 36, and 38 (38 was the precursor to the AS/400).
For its time, the System 36 wasn't a bad little mini. . .
Linux: Our os is based on 30 year old technology and has a few apps that are a bit flaky, and you need to rebuild your kernel every 5 minutes.
yeah cool "advocacy" dude. Best wait until Linux is ready for the desktop before you start hyping.
it is UNIX
Why don't you come up to the front, Johnny, and explain to the class just what exactly Unix is? I'm sure we'd all love to know. Is Solaris Unix? How about IRIX? HP-UX? Is OSF/1 Unix? Or is Unix just a name that gets sold to the loser of the moment or used by bigots to describe their os of choice? These are operating systems. They look and act somewhat like Unix did. But there hasn't been anything I'd call The One True Unix since version 7. They're all different in little ways; either all are Unix or none are. Take your BSD-is-God attitude elsewhere. I actually like BSD, and you make me ill.
IBM's AS/400 development teams have been working
on getting Linux running on their PowerPC based boxes for some time with a lot of success.
Actually, a friend of mine, who is working at IBM recently booted Windows 98 on the S/390. OK, he was using Bochs for Linux, but nevertheless it was Windows on the Dinosaur.
PS: He claimed to get 386 speed on a ten year old machine running not much more than bochs.
As an ex-IBM employee I can attest to this. IBM's hardware still accounts for roughly 50% of its income. Of that hardware, the majority is mainframe sales. But again, right on the money, the mainframe market is not a growing market. IBM, though, wisely moved away from the mainframe/terminal model to the client/server model. Under this model everybody has their own PC which is networked to Mainframes as the servers. And as a developer who writes client/server apps I can also attest that your average mainframe still blows away even top of the line PCs when it comes to volume of data transactions.
haahahhaha....what a faggot you are! I knew you were a fucking moron and now I know you're a punk bitch. Go back to sucking dick, beeyatch.
It would surprise me very much if one of the design or verification groups in Rochester didn't already have Linux running on the PowerPC model AS/400s. Why would this surprise me? Because they already had a mach kernel ported way back in '96. There is more recent literature about the AS/400 PowerPC architecture available (the author of http://users.snip.net/~gbooker/as400.htm is looking at the old CISC processors... a big difference). The PowerPC architecture lends itself to running multiple OSes... Ever notice how similar some of the AS/400 and RS/6000 boxes look... Hmm... Check out http://www.as400.ibm.com/beyondtech/arch_nstar_per f.htm for a look at 1996-97 technology, kids...
e ntations/itso/ibp27a/notes27a.htm)...
What everyone has been saying about IO, you can see from that document is quite true. See how big the memory and data buses were 4 years ago? (Northstar was a plug-in replacement for Apache - a late '95 - '96 processor - http://www.as400.ibm.com/developer/education/pres
High end PC servers are now starting to use I2O... err.. IOPs in AS/400 land... to offload some of the I/O work to from the processor. Programming these ASICs is really what makes the job of moving Linux to the AS/400 difficult. I can't imagine that IBM would give up the documentation to the I/O or memory control chips... of course there is also the matter of compiling and running 64-bit code... too bad the Mac hasn't moved to a 64-bit chip...
Not an AS/400 expert, but I have an incredible respect for the architecture even though it gets little or no respect...
hahahaha...you are almost as bad as slashdot terminal. You're just a pathetic liar whereas slashdot is a fucking mook. you both deserve each other.
Look I have read books...
A self taught retard, I see. No wonder you sound like a dumb bastard. Mr. Terminal, please go back to sleep.
I was surprised no one mentioned Amdahl's law which provides a limit on the ability of a parallel computer to speed up a given problem. Basically if you have an un-parallelizable part of the problem, then even with infinite numbers of parallel processors you can't get better than the time to solve that un-parallelizable part. Mainframes (and for that matter supercomputers) dominate where the problem can't be broken down into parallel chunks (for example you really want a bank to keep its database of accounts current so that it doesn't allow someone to withdraw the same money twice), but are less relevant if there isn't a serial bottleneck. (Amdahl actually figured this out when IBM was debating whether to bundle a whole mess of slowish newly transitorized 650's into a new "mainframe" or to build a faster machine. The faster machine one).
The latencies involved in starting a new process will make most any HTTP server that forks off a new process every time someone connects suck, regardless of OS. Which is why most decent servers today either maintain a pool of processes or take a multi-threaded approach, and why things like mod_perl (which runs inside the server process and thus doesn't require starting up a new process for the interpreter) are much faster and more popular among serious web developers than CGI scripts. Presumably, any HTTP server on a mainframe could be made to use these same strategies, rendering process startup latency a near-nonissue.
As for "loading new data" being slow on a mainframe: What in the world are you talking about?
Funny that this entire article that describes all the advantages of having a virtual machine approach doesn't mention that this feature is available for Intel machines as well: VMware.
We use it for exactly the purposes that he is describing as advantages of VM.
It is a pity that Intel does not give us better hardware support for it, that would boost performance.
A backup/restore machine would be a part of this scenario, which would backup up the other virtual Linux machines at internal throughput speeds.
Since most mainframe shops (hopefully) already have a backup site which closely simulates the home site, a disaster, such as flood or fire, means just running one big restore process to bring back your whole server farm. Sounds better than trying to configure and restore dozens of individual servers, especially under adverse conditions.
Ah, the Burroughs. CANDE. ALGOL. Good memories. Used a B6900 in engineering school for a short time (then they got PCs) but I really liked the environment. Those line-oriented terminal paradigm (as opposed to char-oriented for Unix and screen-oriented for IBMs) were weird but strangely effective. STX-Xmit. I wonder if Linux could ever be ported to such beasts.
Hmm silly little ol' me thought that IBM's mainframe division lost millions because of those PCs but I guess I was wrong.
Yes, you are wrong. Just sneak into any server room in any decent bank. And keep wondering why they have something called 'Tandem' over there, and keep wondering what are those brand-new cool-looking very-big red-black boxes with IBM logo...
Ah, let me guess - you are Linux freak and you were thinking that everything is running Linux these days...
You know, not everybody gives a damn about kernel versions, release dates, useless features for kids & gamers... Some people still need RELIABILITY! (so, don't wonder why so many banks still use Dynix, DG/UX, etc... educate yourself because posting stupid comments).
You Earn £700 a day? Bullshit. No one on gods earth is worth £182k a year for VB. Like most VB coders tho, youre talking out of your arse. Keep up the good work. Best of luck. Ill see you in Britain's richest list in a few years.
I hereby nominate slashdot-terminal for the very first "Slashdot Kook of the Month" award. Will anyone second this nomination?
Funny, the US Patent office runs most of its business on Hitachi mainframes.
if you search ibm's web site for QVC you will find what they use...
> My god...my brain...it's....it's melting...You have made me stupider.
My god! This is the funniest flame I have read in a _long_ time. Thanks FatSean. I haven't laughed this hard for quite a while...
Well, back to my programming project...
Anybody remember Amdahl UTS, the first Unix to run on mainframes, back in the early 80's? Unfortunately, they treated it as a proprietary product used only to leverage hardware sales instead of selling it to anyone with S/370. A few years later, when IBM released AIX, it faded away.
Maybe I will get some use out of the JCL I learned in college after all :). Don't want to even think about the COBOL classes - although CICS was kinda fun.
I know this is offtopic and probally flamebait. But it's fucking hillarious
Good point. Mainframes are useful for extremely large datasets, like geographical or weather data. But the huge latencies involved in starting new processes or loading new data would make them suck as web servers.
Yeah, a horse's mouth that definitely benefits from fear and FUD. I think Edwards has a lot to gain by up talking up the dangers and not much to lose - so I'd want a second more unbiased opinion before I accept everything he says at face value. And while I can see companies should be careful with their smaller thieves have a forklift and bolt cutters your mainframe ain't leaving the data center. Maybe a Sun pizza box... but not a fulled sized 390. Also, I find his claim absolutely ludicrous that thieves that get 50% of the value of the equipment when the sell. Maybe if you got a truck full of hard to get RAM chips and desperate ask no questions buyer. But a mainframe??? Please... I'm laughing. I can just see the CIO at the board meeting... "Yeah and we just found this mainframe at the pawn shop... half price." Man, every mainframe shop I've every been in will refuse to accept shipment if even the internal plastic bad is thorn - much less missing a single piece of paper work.
According to the article in Intelligent Enterprise, they get about 65 million transactions a day. Their database backend is all custom built sitting on mainframes. Even the big Sun boxes were not upto the task. Very impressive architecture.
The one thing that glaringly stands out after reading the comments on this article is that the vast majority of /. Have become closed minded bigots. 90% don't even read the article in its entirety and they drown out the people who do read and/or know what is even being discussed. I could care less about moderation or ways to sort these people out, but after looking at the moderator points on this article would make me want to throw up my arms in disgust. (I had more wrote here but it was cutting and untrue so I nuked it) BTW... maybe more people should read what they write before they hit the submit button. To the educated people on /. sticking around, BRAVO! But I've had enough and had to speak up.... -Cause I can.
Jesus Christ, that was a good laugh, and I seldom laugh aloud at the keyboard. Man, you rule !
Where are you going to get the data that feeds the multiple high speed printers from? To keep a high speed printer working at its full speed, even to print out bills or cheques, takes a great deal of data and a PC, regardless of how fast, does not cut it here. So do we install a number of PCs to feed a single printer? In the early 1980s someone compared the tasks being done on a VAX 11/780 with simular work being done on a PC, an x386 or x486 I believe, and said that the PC could do the same thing as the VAX. It was pointed out that the VAX did it for about 50 people, and faster, at the same time while the PC could only do it for one person. And that was only what the VAX was set up for in that location. What the VAX could do without even trying is still beyond what any PC can do at its very best today. And the VAX is not even a mainframe computer, and not in that league either, but only a mini computer. Any mainframe can blow any mini computer away without trying and a mini computer still makes a PC very slow. I am sorry but as far as a PC has come is still far behind what is regarded as common in a mainframe enviroment.
- The rifdiculous ease of installing Linux on a VMS machine (The did it in a couple of hours!)
Please, if people want to sprout of shit about this topicAnother post had a good point too: I used to manage a System/34.
You mean... System/32 !?
System/34 doesn't exists...
It appears that LinuxPLanet has been Slashdotted! Oh well, so much for stable Linux web serving. Next time, try BSD for an OS that is not U*IX-like, because, it is UNIX!!!
My daily chargeout rate in the United Kingdom is around £700 (approx $1000), so by reading these postings, the Linux community is getting some very worthwile advice from someone who commands a huge premium for his opinions.
For this reason, I expect you to sit up and take notice when I criticise these mainframe losers for thier attempt to commandeer and hijack Linux for their own ends.
From a marketing perspective this is nothing short of suicide.. Linux has enough of an image problem as it is. The average joe-sixpack sees the Linux community as being a bunch of pony-tailed dressed-in-black Quake-playing open-source zealots who never emerge during daylight hours, and who demonstrate an inability to socialise that would be amusing were it not so pathetic.
The last thing Linux needs if it wants to appeal to the trendy BeOS/MacOS using creative types in the new-media/marketing domain is to become associated with the gray-haired pocket-protector brigade who jealously guarded their machine behind a glass wall, and whose general high and mighty approach was responsible for the triumph of Microsoft and the PC on the corporate desktop.
In short, Linux + Mainframe + Old technology = Marketing death.
Now, I ask the Linux community, perhaps, just perhaps the "open source" bandwagon has rolled just a bit too far with this one. Why should the corporation that can afford to pay $$$$s for a huge mainframe expect to get the OS for free ? Its not as if the community benefits from the port, since most people will never see one of these machines in their lives.
Perhaps it is time for some new exclusion clauses in the GPL. Perhaps GPL only applies to companies who make less that $1bn per year profit.
As usual this "open source" marketing advice is completely free.
Thank You
dmg
"The word you want is RELIABILITY."
Underline, highligh,put up in neon. This is IT.
"Many of the people who are putting together clusters of machines of lower reliability - including those in the management of at least one mainframe company - haven't grokked that concept."
Agreed. In fact I believe that BYTE magazine (before it's print demise) did an article on the *unreliability* of the PC platform compared to other class of computers. The average pc wasn't designed for reliability, for economic and other reasons.
Theres no way in hell gigabit ethernet can compete with ccNUMA or craylink. It just can't move enough data. Gigabit is 120meg/sec on paper, but never in real life. Some real specs for craylink are 600-700 meg/sec, and ccNUMA even higher. If you aren't moving serious amounts of data and enjoy rooms full of cheap pc's then beowulf is right for you. Why have rooms full of pc's when a mainframe will take one rack and a LOT less power and maintenance. Whats the uptime of a pc? maybe a year at best because of the cheap hardware. VAX mainframes are still chugging away. BTW you act like beowulf is something new. VAXen have been clustering since the mid 80's.
That's all nice and good, if the 100M hits per day are evenly distributed throughout the day. But peak demand is going to be considerably higher than average demand - how many hits per second does your site get when it's 6am on the US east coast? How many when it's 2pm? And at 7pm? Unless you don't mind your system becoming unresponsive at peak hours (which probably account for the vast majority of eyeballs seeking your site), you'd better take peak demand into account.
For a second, I thought these were real Slashdot postings! Too Funny! This has to be the best Slashdot post ever!!!
Did you hear the rumor that Packard Bell is buying VA Linux? According to sources, Packard Bell is really into this "Linux Thing".
Please tell me your joking and don't consider this a credible argument.
But to give an equally stupid and flip response: It's a lot easier to plug that one back in again.
Uh, you mean like the old Unix 32V (that's "V" as in "VAX")?
ROFL!
Ahh, my Vic20 could do that in its sleep :-)
...well, maybe if you plugged it directly into the sun.
The interesting thing about AS400 and RS6000 is that the hardware is on a convergence path. Prolly within 4-5 years there will be no "AS/400" or "RS/6000" anymore, although there will be a box to run OS/400, AIX, and maybe even Linux.
The kicker is that they've already developed LPAR (logical partitioning) to enable multiple copies of OS/400 to run on an AS/400. I would be shocked if this weren't expanded to allow multiple *different* OSs running concurrently on the same box.
Dunno why IBM does not push secure a bit more. Apart from RACF, IBM boxen come with some hardware assist that kills bad tasks dead. Partitioning is line #1. Then memory has extra 3 bits for 'keys', plus the concept of trusted vs non trusted user code. Plus everything/resource can be audited. And one can protect memory in multiuser address spaces, and set some very cool hardware assisted traps (SLIP traps) and PER. In short, memory leaks and buffer overflow exploits are impossible, or catchable if they were under root. All system calls and instructions go into a nice hardware assist trace buffer. The resulting dump means you can nail the problem exactly. Which comes to the WORST point about mainframes. All the dodgy, memory leakin code would be dead in jiffy because of program or data exceptions; which may not suit some vendors. Ever since CICS had bounds checking, I have been struggling to understand why memory leaks exist (imagine write protected memory, and the box automatically stops you (accidently) overwriting or reading memory that is not yours). Cheer IBM. If the linux port is true(as in ready for prime time), then the open source community will have a means to finally debug and prove their code, on the greatest debugging box on the planet. Want to audit IPsec. Easy, just turn on tracing. Then the perfect code can go onto PC's that dont need elaborate and expensive hardware protection. ooopps- another negative, better security + fewer security administrators, and quicker exposure of the less capable ones, and accountability.
The player is 4 gigs and currently can only play one movie.. but these are just kinks people, kinks!
Also, you would probably be surprised how much a top class VB developer can command for domain specific knowledge, especially in the fast moving investment banking world, where millions are made based on the output of one Excel macro. Remember these posistions are not widely advertised.
Although i'm sure evolution helps refine a product - the mainframe started off with different base of objectives to the PC. Cost was the main objective for the PC, whereas Mainframes were designed to be GREAT with little regard for cost. They were creating what they thought was a good machine - nothing else. PCs were made after mainframes for christ sake!
<p>So far as I see any particular architecture inherits what designers/technicians know and your "evolution" idea of a particular computer specification doesn't really make sense at all.
<p>Improvements will be made over time, but a specific product doesn't evolve.
<tt><p>====<br>
This is an automatic response to stupidity. Please do not respond.</tt>
- The administration efford would drop to near zero
- Accessing the companies database/intranet would be realtime, since the only limitation would be the X Server connection
- The computing power would be shared ideally between the users and upgrading the server would mean more power for everyone
- You could take your Desktop _anywhere_ in the company
The big question is as always the money. Can anyone estimate what this solution will cost in the means of TCO?keep up
In ways, I think this guy would cause less damage if he were writing code than if he were writing the sort of drivel he has so patronisingly foisted on us. Its quite clear that he has no real concept of the demands of major users in teh marketplace. Firstly, why should a M/F shop run Linux? Well, it *is* cheap (nothing in this life is free). Also, it may well allow users access to new versions of layered s/w earlier in their development cycle (how much longer is the rollout for S/390 versions of packages? - generally at least six months in my experience). There's no point in getting into the usual M/F versus NT battle here. If I were running a small businees with a requirement for small-scale office systems, accounting S/W, and maybe a payroll, I'm always gonna buy the smaller machine. But if I want 24 x 7 availability across a multi-time zone corporation with 1,000's of employees (not to mention customers), I'll have to go M/F. No two ways about it.
Sheez, you people can't recognize a joke to save your life.
That's amazing!!!
If this is the article that I read last week (couldn't get to it, it's slashdotted), it made a nice case for using such a beast for web hosting. Instead of using rack after rack of little server boxes, give the customer a partition on an S/390. The customer can be given root access, since the partition will be invisible to other customers. If the customer slips up and types 'rm -rf /', the image can be quickly restored from a backup.
Want a new DNS, mail, web, FTP, or whatever server? Don't spend $[foo] for a shiny new PC - just fire up a partition for whatever server you want.
Sorry, but 10 PIII 550MHz systems linked in a Beowulf couldn't keep up with a high-end UltraSparc-2 450MHz Sun workstation for most tasks. Sure, if you need a render farm, where the work can be broken down into discrete components of a reasonable size (i.e., "Perform these operations on this 10M of data and get back to me when you are done"), then a Beowulf is great. But what if you need to operate on ten thousand pieces of data at 1K each? Batching the job (you do 100, you do 100...etc) is inefficent, and clogs the network all to hell. But I have both a small beowulf, and an RS/6000 34H (with 64M of RAM). The beowulf may have ten nodes, and be a great system for making MP3s in parallel (send one mp3 to each system), but when it comes to real work (SQL-based databases, etc), the RS/6000 romps the beowulf -- even though it's much older. And that's not even close to a mainframe. The beowulf is all P233/64M systems.
Score 2
The original joke posting was far better than the follow ups. This moderation thing is wierd, used to like it, but now I think I am missing out on good posts
I don't think a bewolf cluster would cut it for working with such things as distributing phone calls over the phone network, or processing huge amounts of active data.. bewolfs are good for processing data that is in storage.. IE> not changing. Huge math tasks and graphics and such are good for bewolf, because the task must be know to be split and distributed evenly. Mainframes have the shear size and power to do this in an active environment.
./ effect strikes again
HA HA HA
You are a goddamn imbecile! Your 486 does not outperform mainframes of the 60s and 70s. Just because you suck at programming and can't write an app that would use massive ammounts of CPU, IO Storage does not mean these
are un-neccessary things. You want to quit CS because it's too hard? DO IT NOW! You want to become a TEACHER? Oh my god...you need to get your head out of your ass first. You're pissed because peoples expectations for fast
computing are driving up prices for an 'average' PC and you can't afford it? My god...my brain...it's....it's melting...You have made me stupider. I will dedicate any and all moderator points I get to banishing your unfounded,
blatantly wrong, ignorant posts to the realm of -1, where no one need read them.
Hey buddy you know what you can go to the lowest pit of hell. If I met you on the street you would be whistling Dixie out of your chest where you would find on closer inspection about 50 bullets fired from an AK-47 assult riffle if I was in a good mood.
Hell I would hazard to guess that my HP48G has more processing power than old mainframes that could barely add a couple of numbers together surrounded by guards of the celestial temple and almost impossible to figure out. Oh yeah an punch cards sucked and swallowed compared to any type of (even crappy interface we have now) alphabetic ordered key input methods. And I don't have to cary a shoebox full of cards either.
Oh by the way you never have moderator access all the time jackass and even when you do it dosn't mean I will post when you have it. Couple this with the fact that you are most likely some snobbish, arrogant, little rich boy from Silicon Valley and you never had to work worth shit in your entire life dosn't mean that I am an idiot. I truely have never had (even in the most complex programs I have even written) to have a mainframe or anything close to that.
Here's a little thing for you. If you ever, ever, ever were to demoderate me to -1 I would just open up another account and be able to post at default of 1 change my wording slightly; the +1 posting thing is a matter of convience. Plus all I have to do to get it back up is say genuinely what you idiot shit for brains people want to hear and presto within say 6 months I have more karma than Rob does. Simple and effective. No fuss no muss I say. I am killing time and that's fine by me. Mmm k?
Thank you...I really, really needed to say that.
No thank you I surely hope that your little mainframe buddies die a screaming horrible death at the hands of reason. You are power mad but the future dosn't fortel the increasing of power more likely the even or more proper distribution of it.
I'm not an expert but I used an AS400 so I've looked at the site http://users.snip.net/~gbooker/as400.htm
and I'm not optimistic : it seems that the way th aS400 handles memory (among other things) will make it impossible or at least very hard to port linux.
But the main problem remain the difficulty to gain informations about AS400 (especially hardware stuff as they consider it as trade secret...)
It's sad because the AS400 is a great platform
despite its age...
Great stuff! What moron moderated this down?
Funniest post EVER
What this brings to the mainframe is a lot of generic canned programs, not to mention the notion of portability of applications. I.e., running Gnumeric spreadsheets on your AS400 rather than some $10K AS400 specific application written and maintained (say $1500 or so per year for the user--the need for maintainance is one of the reasons for the old practice of annual licenses and/or required maintanence contracts). It's a win, win, win situation--IBM and its customers win, and the community benefits (e.g., the ongoing early stage journaling file system efforts for Linux coming out of IBM--they can pay programmers to make Linux a class act rather than screwing around maintaining apps for 1000 users). Most really important apps for mainframes will continue to be pretty much one off deals--I can't see Citibank depending on somebody else to keep their credit card billing operations going (!), but it would be really nice to have a lot of little resources out there (like gnumeric, for instance) that you could apply to things in your databank without 6 weeks of effort from 3 programmers to run a check on some contemplated decision or gather some oddball statistics.
Sane moderators have been taking him down a notch or two. That guy is truely annoying...a Troll if I ever saw one. Or maybe just goddamn ignorant? Hard to tell from here...
Look I have read books and this was more than just one book. Basically if you paraphrase them they say that mainframes were used then. Now we use a client server way of doing things effectively killing most mainframes. Those that are left are the leftovers from the olden days and are falling stars even now. Take that as you will.
They don't understand scalability. They don't understand reliability, they don't understand the development process, they don't have any appreciation of business imperatives, and they do not have a clue about real world systems.
bearing this in mind, do not take what you read on slashdot too seriously. *(unless it is written by JonKatz, and features "tech-savvy geeks").
Are the moderators on CRACK ????
my dual celeron 500 gets 1250 MIPS on that winbench 99 or whatever ziff davis makes.
Not when that one is running through a conduit directly to a breaker box.
No doubt the crazies who anyone think linux on 390 is a good thing will be wanting to run it under Windoze next. Once IBM move in then its down hill for opensource, can it compete with 5,000 people working full time ( some of them don't even wear suits these days!!!) On the other hand maybe this should be adopted as a new benchmark I can just see Bill G. thinking now: "If IBM can do 41,000+ then Compaq and I should be able to do better " the answer is XML
= 1 - (.9) ^ 8 = .57 = 57% of at least one failure
(In other words, take the odds of all 8 NOT failing (90%)^8 and subtract that from 100%)
W0w. If a humorless moron like you has all that karma then I sure as hell don't want any. I'll just keep on posting as AC.
I think you are confusing mainframes with super computers. Super computers are for processing and mainframes are for I/O, data throughput and availability.
If this got moderated up as funny, then I can safely say that moderation is a complete failure.
damn terminal... get a life loser boy
The comment "Interesting article if you're wondering why anybody would pay seven figures for a box when you can get a high-end pc machine for a fraction of that." The above is one obvious example of how the PC guys miss the point. Not only are the costs no longer seven figures (thanks largely in part to PCs) it is not the boxes, but the admin costs, lack of security and slow response times that muck up the PC/Linux/Client Server concepts as saviours of corporate computing. In my company it now takes conservatively 10-20 times as many people to admin the Client Servers running unix as it used to under IBM/s VM operating system, (which BTW is software that has run the unix operating system for almost 20 years as a guest and not hardware as FUDed elsewhere). In addition to that when we accessed the web through VM we did not have to worry about silly things like not being able to boot up our browswer on certain artists' birthdays or some such nonsense. The main point is that certain computing environments have certain strenghts and others other strenghts. Mainframe, PC, Linux, Mac all work well at certain things and the Linux on the mainframe running under IBM's VM operating system would be a great solution for lots of Client Server slow response and admin nightmare problems as well as making the recent DoS attacks impossible as far as effect upon an ISP. Ranting and raving otherwise just makes one sound like a religious bigot or a Dan Quayle.
>As a highly respected practitioner of the science of Marketing, I
frequently post to this forum to attempt to "give back to the
community". I cannot write C++ or Java code to save my life (I'm a VB
expert) so my contribution to this forum takes the form of "open
source" marketing advice.
Sod off. We like the idea of Linux and the BSD's going places Windows never will. Can you imagine anyone wanting Windows 2000 running on a mainframe computer system? Of course not. So take your "advice" and shove it.
I'm wondering if you actually read the article. I know it's long and has words with multiple syllables, but it actually addresses every single question and implied question you asked.
US Citizen living abroad? Register to vote!
This looks like it was a very good article but I was only able to read the first page of it because by the time I was trying to click through to the second page, the damn thing had been slashdotted.
I used to hate multipage articles because it allowed the webmaster to force me to submit to more advertisers-but now I hate multipage articles because of the slashdot effect.
Oh well.
-- Truth goes out the door when rumor comes innuendo. -- Groucho Marx
where the classic supercomputer does not have a cache. The supercomputer streams operand vectors from memory to vector processing units in the CPU and the results are streamed back to memory.
Damn! And I thought all those beowulf clusters on the top 500 supercomputer list were real. Imposters!!!
It's not totally absurd. Just doesn't have anything to do with the example at hand. If your bank, or in my case, Credit Union, is really small, ie has assets less then $3 million, then there isn't enough work to demand a mainframe. A singel server will often do the job. Any real I/O intensive work (printing statements) will be outsourced to 3rd parties who will have mainframes.
Of course, any large national or statewide bank, will use mainframes. To attest otherwise is just plain silly.
-BrentYou are a goddamn imbecile! Your 486 does not outperform mainframes of the 60s and 70s. Just because you suck at programming and can't write an app that would use massive ammounts of CPU, IO Storage does not mean these are un-neccessary things. You want to quit CS because it's too hard? DO IT NOW! You want to become a TEACHER? Oh my god...you need to get your head out of your ass first. You're pissed because peoples expectations for fast computing are driving up prices for an 'average' PC and you can't afford it? My god...my brain...it's....it's melting...You have made me stupider. I will dedicate any and all moderator points I get to banishing your unfounded, blatantly wrong, ignorant posts to the realm of -1, where no one need read them.
Thank you...I really, really needed to say that.
Blar.
Sane moderators have been taking him down a notch or two. That guy is truely annoying...a Troll if I ever saw one. Or maybe just goddamn ignorant? Hard to tell from here...
Blar.
I think you are glossing over years of interesting history and computer architecture. IBM's mainframe division might have lost mililions because of PCs, but not because PCs were superior computing platforms for big data.
I am basing this on what I have read. Most of the time this interesting stuff was going on I wasn't alive so I can't totally say yea or nea on this.
PCs won for the same reason that we have traffic problems on our highways--everybody wants to be a driver and doesn't want to share. This was, it seems, part of the rationale for the federally funded development of the internet (well,
ARPANET): getting scientists to share computing resources bought with federal money (c.f. A History of Modern Computing, Ceruzzi 1998, p 296). This is of course the same phenomenon that drove minicomputers, which were
also replaced by PCs. I'm not saying this is bad, well, not in the case of computing anyway.
From some of the horror stories of various teachers who had to take CS back in the good ol' days (read 60's 70's and early 80's) it really sucked without a PC of your own. My shitty 486 currently ourpreforms all those worthless mainframes from those days and I have complete control over it (well I can't orger it to do something that it can't but some linux apps try).
But PCs still suck at any number of computing tasks, and aren't really improving in areas that can't be mass marketed. That's why my lab bought a very expensive dual Alpha machine instead of spending that money on the 5-to-20
equivalently clocked P-IIIs (these numbers come from real computations). Not to mention a farm of PCs can only handle embarassingly parallel computations at the same speed has the Alpha, and require more programming effort than
the Alpha. The Alpha isn't even close to a mainframe, either.
Buying machines of massive quality just because you can is just silly. I have never ran an app that actually took that much processor power and I haven't actually wrote one myself that could do anything that I would like. Programming is quite difficult (I am currently struggling with a couple of programming assingments myself and they just suck (Just for kicks I am thinking of ditching CS any good schools for say professionals in American History or in history related fields that I could just get a teaching degree in? I really hate programming)) and not something that can easily take control of massive quantities of data. Maybe people should expect things to take a little longer than 10ns or has attention span decreased?
And I haven't even gotten to bandwidth issues that sponsered this thread (well, they're part of the 5-to-20 figure above, in some ways). IBM lost on mainframes because they dealt _only_ with mainframes. DEC lost with
minicomputers because they too were arrogant/ignorant about PCs. And while Intel seems to acknowledge the information appliance ideas, they're x86 tech will only go so far (we can hope, can't we?). But just as information
appliances aren't the best choice for PC-type tasks, PCs aren't aren't the best choice for mainframe-sized loads.
I shudder at the thought of "information appliances" I pointed out once that I think that the only thing that will do for me is increase PC prices at least 4x the current rate because all the mass market decided to get some AOLphone or something like that. It just ends up screwing me in the end.
Oversimplification is a marketing tool. It has no place in intelligent discussion, where flippant remarks are better replaced by _questions_.
I have never done any marketing at all but I still say that at least from all the reading I have done and all the current businesses I have seen in operation that most of them use a client server setup wherein employees have PCs and are connected to a server or servers which run some network aware OS where data and applications reside. The client has more responsibility for data than in the mainframe days and also the server has less work to do but is at least smaller than the mainframes of old in a ratio. This is what I have seen maybe it's different in a few of the richer companies.
Slashdot social engineering at it's finest
A mainframe is NOT a thing of the past !!! What do
you think manages your bank account, possibly your
salary, certainly your IRS account and probably your pension fund ???
A:
It's called a Beowulf Cluster. 10 PIII-550's. Total cost $30,000 as opposed to a 6-figure mainframe. IRS though, well, they're just a wastefull buracracy.
kwsNI
It's quite amusing to see these comments about how mainframes are dead, and that beowulfs are now the 'true way'. I was over at CeBIT in Germany the other day, an saw all sorts of interesting things. Not only IBM's relationship with Linux on more serious hardware (RS/6000 etc.), but also their introduction of Monterey on I-32, I64 and PPC machines makes food for thought... They had one of 6 Itanium machines to be seen on the whole of CeBIT campus (which comprises 7,400 exhibiting companies), and it was running Monterey. What was also very interesting is Unisys' latest data center mainframe - if you can call it that. It had all the crossbar, main storage, I/O and dynamic partitioning stuff that is so typical to mainframes, but its processing nodes off the crossbar (the bit that sits between processing, IO and memory) were each made up of two Pentium III Xeon processors sharing a bus. This was quite amazing - they had their top-of-the-range 32 node box on show, running Windows 2000 DataCenter Edition and SCO Unix simultaneously on two partitions. IBM seem to be going the right way with their server stuff. They're now even producing small single-unit height rack Netfinities for ISPs and Telcos (obviously these support linux). They're being announced sometime next month... But quite a nice move, considering the fact that Sun have those 1U boxen (Netra or whatever they're called). CeBIT is a good place to go for this kind of stuff - it's usually very cutting edge.
No actually it's probably a Stratus box. Stratus These guys make crazy redundant hardware/os stuff. If you have a credit card, get a prescripton from CVS, buy anything on QVC it goes through a stratus box.
I actually did this once, based on originality of the troll, but it got promptly moderated down all the way to -1. It was such a good troll that it actually lived up to the very definition of troll: To fish in; to seek to catch fish from. Here on /., the Usenet sense should prevail, namely: The well-constructed troll is a post that induces lots of newbies and flamers to make themselves look even more clueless than they already do, while subtly conveying to the more savvy and experienced that it is in fact a deliberate troll. If you don't fall for the joke, you get to be in on it. It would seem to me that most /. moderators can't spot the difference between flamebait, offtopic and troll, and sadly Cmdrtaco promotes this by tagging troll as something inherently bad. It's not.
Windows isn't a toy, I've *never* had any fun with it.. it's an embarrassment. :-)
You can always tell the new converts by their inappropriate application of beowulfs to applications where they are highly unsuitable. The reason people still use mainframes is IO throughput. No cluster of PCs can do that.
Someone here on slashdot has a sig that's quite appropriate here: "the average slashdotter seems to think the entire internet could be run of a cluster of beige boxes running linux."
Grow a clue people. This is almost as bad as the idiots who scream beowulf when someone asks about high availability and fail-over.
My dad worked on a system where they had full machine specs (back in the magnetic core memory days).
He has a great story about them not needing floating point stuff at their site but wanting to drive some peripherals without a compatible interface. To cut things short, they ripped out the floating point board and replaced it with a driver for this peripheral.
All was fine until they got an external consultant in to help with debugging part of their system software. Apparently it took him several weeks to get his head around the fact that occassionally the application would appear to do a floating point division and then discard the result - of course this was actually flushing a buffer.
*sigh*
closest I ever got was wiring morse keys into a joystick port - hardware hacking just isn't what it used to be.
At 262 MIPS, new Trinium-II Series 9 processors deliver considerably more power than IBM's G6 'Turbo Opera' processors, which deliver about 201 MIPS of power per engine. The new Triniums not only deliver more single engine power than the G6s, but they are considerably more scalable than IBM's S/390 servers and offer better SMP ratios to boot. A 12-way IBM G6 server has an aggregate of about 1,614 MIPS of power, whereas the 12-way Trinium-II processor offers 2,441 MIPS of aggregate processing power in a single system image. The full 16-way Trinium-II server has 2,969 MIPS of power, 85% more scalability than the biggest 9672 G6. Trinium can crunch 262 MIPS per processor and support up to 16 CPUs in a single enclosure. Performance on a full system tops 3,000 MIPS, compared with a ceiling of 1,600 MIPS on the 12-processor System/390 Generation 6 line that IBM shipped last year. Now, that's a webserver. Well, not really.
The E10K (starfire) isn't a patch on the IBM mainframes that the article talks about. You can only configure domains based around system boards, and you need to use an individual disk to install a separate copy of the OS for each domain. On top of this, I have found that dynamic reconfiguration isn't as reliable as I'd like it to be. OTOH, we are using quite an old starfire now.
You can get large amounts of storage and redundancy on a starfire. But it still isn't a mainframe, although it is still a damn good high end Unix server.
Not being a maniframe type, I enjoyed reading about the s390 and the vm stuff - also, it was a pretty cool thing to have, what did the one guy say, 41 thousand Linux images running at once on one of these? Gives an idea of the muscle these babies have, huh?
"shop smart:shop s-mart" ash
Whoa there! I never said VMWare replaced this, but rather the idea of Linux running under a virtualized system was first (?) demonstrated under VMWare.
I was more thinking that a flooded room of 20 servers stands a better chance of (some) equipment surviving than a similar room with one piece of equipment. After thinking about this, however, I don't think it matters much. :-)
Maybe you mean "kludging together something with baling twine, duct tape, and spit, until I have enough time to rebuild it properly".With a backup and a system to dump to, I can indeed have a system back up and running in under an hour. However to bring a entire system (or systems) back to what I'd consider "safe to release to the wild" I would take a few hours and test everything, so you're right, I do mean to get it back up so (hopefully) nobody saw that the site was down. :-)
But with multiple OSes running on multiple distinct and unconnected partitions, running completely independant, but on the same hardware, that's just ideal.I dunno; I would still rather have them on seperate physical boxes. (I did mean seperate boxes in my original post, not all on one virtual box in the big S/390.) Perhaps it's the cool factor of walking into a room of machines instead of a room of machine. I see where there is very little difference when the big iron is completely redundant and hot-swappable, but something inside me says not to trust a single box.
Financially, I would very much like to see the math to see where the break even line is. I smell an "Ask Slashdot" question, can you?Me too. Rob?
Cute article. But what's it say? It says that the mainframe hardware is capable of fooling the Linux kernel into thinking it's in charge. Hear of this anywhere else? Yup. VMWare.
Now before you fire up your flamethrowers, think about it... The S/390 hardware was designed to hide itself from whatever it was running, which is exactly why it works as well as it does. Hell Microsoft could port NT5/Win2k/whatever to this platform. So could Be. So could OS/2. (Does OS/2 have a port to the S/390 already?) Linux isn't doing anything new or special here, it's just another virtual OS to the hardware.
I'm a huge Linux advocate, but this little excursion into mainframe land by Linux proves very little. What is the point of running a hundred (or 41400 I think the number was) concurrent Linux sessions when Linux is already multuser and multitasking? Clustering? High availability? No, because you can get all that with the existing mainframe architecture. Running Linux apps on Big Iron? Perhaps, but wouldn't it be better to port Apache, etc. to the mainframe OS and run it on a system that perhaps knows a little bit more about the hardware?
I guess it'd be neat to run a bunch of linux sessions on a mainframe to test a new clustering architecture/algorithm but after the test, there's not much use for it, AFAICT. The Linux/390 project doesn't do much but add to the "cool" factor of Linux, which isn't really all that bad. :-)
I work in flight operations for NWA, and we have something like ten mainframes here in the MSP computer center being used to run various applications (some IBM hardware, some IBM clones, some Unisys 2200 hardware).
We still use mainframe hardware mostly because mainframes in general have better reliability/recoverability than large Unix boxes have managed yet to date (and we can't afford to have some of our operational systems go down at all -- downtime translates to flight delays, and an hour outage could strand dozens of aircraft and cost us millions!).
Please don't underestimate the HUGE impact that mainframes have had on large-scale businesses, and don't overestimate the impact client-server had on large mainframe sites. Nothing has changed here, though we have more options for smaller projects now, and mainframe use here is growing if anything.
Almost all of the larger airlines, banks, insurance companies, and government agencies used mainframes for their core processing before c/s became the popular architecture in academia, and as far as I know that's still a true statement today. Client/server sometimes gives you additional options, but for most it isn't a real replacement for a decent mainframe-based system.
--
-Rich (OS/2, Linux, BeOS, Mac, NT, Win95, Solaris, FreeBSD, and OS2200 user in Bloomington MN)
Mainframe/UNIX Bit Twiddler and long time Windows/Linux Hobbyist.
The Theorem Theorem: If If, Then Then.
Wow...I went there...they've changed. When I started there in '89, the 3090 wasn't even on TCP/IP.
Interesting thing about Marist is that IBM decided to take a small college (~5000 students) and give them $10 million in equipment to see what would happen. This amount was upped over the years, and you better believe IBM stood for I Bought Marist, but it was a good place to go.
I'm glad to see that the relationship has continued between the school and IBM...still doing cool things there.
What makes you think that a mainframe has only one network interface? I work with our mainframe every day, and I will freely admit that no quantity of Linux boxes could manage to do what it does. According to IBM, we have the highest transaction rate of any non-military mainframe on Earth. 1,000 Linux machines with RAID 5 and dual gigabit Ethernet cards in PCI66 slots would fail utterly to achieve the numbers we require.
I have to agree, came quite close to messing myself, that was so damn funny. Thanks (I must have missed it getting posted before.)
It was a very interesting article and I think putting Linux on these old beasts is a good thing(tm). I would like to point out a few things:
:-/
Alphas are not "PC" quality or speed! There are only two (new) Alphas that I can think of that are less then $5,000 USD, the DS10 and the UP1000-based systems. The DS20, one very popular system, is over $10,000 USD. Last time I checked a dual processor DS20 was $16,000 USD (I think). The only Alpha I can think of that could be called "PC" quality was the AlphaPC 164SX which has been out of production for some time. A AlphaServer GS140, a 14 processor system, costs about $500,000 USD and are definitely not "PC" quality. The Wildfire (32, 64, and 128-way systems) will be available later this year and are definitely not "PC" quality. The Galaxy systems support running multiple operating systems simultaneous, so this is not a feature unique to IBM "mainframes".
I am sure there are many people out there that would agree with me that OpenVMS is not a "PC" operating system! It runs on Alpha and according to the article Alphas are "PC". So OpenVMS must be a "PC" operating system?
The people who worked on this do not deserve "virtual beer" they deserve the real thing!
The thing that struck me most upon reading this article (last week) is the depth of thought and attention to detail that has gone into mainframe design. A Pentium-II can't even totally virtualize itself without some *serious* trickery.
PCs clearly have a long way to go, as evidenced by that fact that Microsoft's presence in the mainframe market is zero.
Connected to www.linuxplanet.com.
Escape character is '^]'.
SunOS 5.7
login:
Hmm....
I have to return some videotapes...
Actually, many people advocate mainframes as web servers. Check out Schwab's web site. It is served off of a *cluster* of mainframes. Schwab and IBM teamed up to create the networking required to do this. Why, because when Schawb lowered their prices for individuals, they got swamped. In response, they bought a bunch of mainframes, created redundant data centers, and now server their customers.
Specifically, in terms of hot-swaps and reliability comparisons is what I'm really thinking. I understand the whole IO issue (really, I do... I work with MF guys all day...). ;-) My focus is on smaller companies than Chase or the IRS (or any of the other gargantuan examples already suggested). These middle sized companies don't need this kind of throughput, typically, so it doesn't make sense for them to spend the money usually... or does it?
Sujal
politics, food, music, life: FatMixx
Maybe it's just that it is early in the morning (sort of) but your math seems a little off... it would seem that if each box has a 10% chance of going down on a given day, and you have eight of them, your odds of sucess are closer to 20% (100%-(8*10%)) than they are to 99.999999%. Obviously, both of these numbers are wrong... anyone who has taken stats more recently than me is welcome to correct them...
~luge
IAAL,BIANLY
But you don't have an 80% chance of a failure a day. If that were true, 10 machines with a ten percent failure rate would guarantee a failure every day, and that just doesn't make sense. My challenge to someone whose stats are fresh still stands...
~luge
IAAL,BIANLY
Duh! The calculations should be 0.99^50 and 0.99^100. The resulting odds in my post are correct, I just wrote the calculations wrong.
"If you're on the ball as an admin, you shouldn't have many script kiddy problems."
:)
If you're in charge of an S/390, I hope you *are* on the ball
Linux or OS/390, the choice is irrelevant - at the end of the day you have to secure it.
--
Peter
Mainframes, from my understanding are built for:
1) Fault tolerance
2) Massive I/O
They accomplish this, generally, by a lot of virtualization in HARDWARE, while normal PC's do this virtualization in software using stupid tricks like time-sharing (and calling it multitasking)
The PC, though, is about to pull up its pants, so to speak. Crusoe is already virtualizing instruction sets of all the fanciest desktop processors. SMP is going to make a big leap when virtualization fully leaks into the desktop. This will probably happen inside of two and a half years.
Either way, I am projecting that the life of the mainframe is 10 years, maximum. Likewise, the paradigm of desktop computing as we know it is also going to be gone in this same span of time. Both will probably be replaced with some sort of scalable, componetized architecture that eliminates (or greatly reduces) obselescence, improves reliability, and increases performance.
Damnit slashdot.. I was halfway through that article when it got splatted by the /. effect.
Fuck!
Mainframe == data pipes the size of my love rod
SuperComputer == terflops the size of my love rod
So, mainframes are data pushers. (like masssssive I/O to disk / network / processes)
supercomputers are number crunchers. Lots and lots of flops. Flopping.
Both cost a god awefull sum of greenbacks.
SGI wrote one. I don't know if it was included
/??"
in their latest flurry of code releases. Probably
was - lord knows they're never going to make
any money from it.
"I need a pilot's license to get to
Heh.
K.
-
-- Proud descendant of semi-nomadic cattle-herders.
That's all fine and well, but you're leaving out the virtues of high throughput and reliability. Where do those fit into your handy cluster? The 2 do not compare.
-lx
UNIX, a cousin of Linux has been running on mainframes and super computers since the mid-80s.
In fact, a lot of the attraction of UNIX was that
it was tuned for high-power I/O, and the feeble
capabilities in MS operating systems.
Looks like I can dust off the mainframe lines in my resume, maybe change "timesharing" to "Application Server".
The idea of margin-shaving ($x/user) may seem attractive at first but carried to it's logical conclusion leads to the PC revolution all over again. There are still individual PCs with more computing power than the equivalent mainframe process devoted to their needs. Centralized IS still is about as innovative as dry toast, further compounding the issue.
Also remember that unused capacity is a sunk cost. All sorts of hilarious schemes have been developed in the mainframe data center to recover investment in un-used capacity. They all seem to reduce net reliability and flexibility. On the PC unused capacity seems to benefit the user or M$ more or less.
I do kind of miss my 3270-horrible and VM and Assembly and APL. But I don't miss SNA and APPC/VTAM and PDS and ISPDF et al. I don't recall Mainframes as being incredibly fault-tolerant per se. It has more to do with how much you want to spend on reliable systems planning, administration, and applications development.
For those who think Beowulf or other clustering is unsuited to running billing or invoicing or other mundane apps - don't get your hopes up. I imagine a network of pic microcontrollers could do the job. The reason for the big iron of course has to do with the Tons of Slow Overhead (this is mainframe joke, look up TSO), which requires lots of computing power in the first place.
Final point, for those of you with ISPs: The main reason for buying a mainframe is because...you already have a mainframe. But it is funny to look at server racks packed to the gills and compare them to big iron, which is boxes of racks packed to the gills.
Okay, that's so. It's just a small detail :) Anyways, you are right. Beowolfs have no place banks, even though small banks have non-mainframe servers.
-BrentLinus made special exception to binary modules.
So, Linux is closer to LGPL license than GPL.
Je ne parle pas francais.
No... IBM lost billions in it's PC business, but made up for it with mainframes, workstations, software and services. Mainframes aren't nearly as glamorous as PC's, as they ship in such small volumes, but the margins on those are immense compared to those from PC's... Like 40-50% rather than the 5-15% you see in the PC marketplace.
is this lineo's tcp/ip stack for dos free? i'd die to see dos do some tcp/ip. any other options?
also it looks like there are no DOS binaries at the boa site handy. does the code compile cleanly on DOS?
Hmm silly little ol' me thought that IBM's mainframe division lost millions because of those PCs but I guess I was wrong.
Mainframes lost to pc's because pc's are usually better home/office workstations than a terminal connected to some mainframe.
That don't change the fact that there are still big jobs around that a mainframe does better than any pc.
Its a lot harder to yank out two hundred power cords than one.
Come on! Anybody who ever saw a mainframe knows they thought of that one. First, these things are usually installed without a pull-able socket. You don't fit a mainframe in your building, you fit the building & infrastructure around the mainframe's needs.
Second, they don't allow any fool near the machines, that helps uptime too.
How many of you, if you discovered a buffer overflow situation that would let you enter a command string on such a system, would know what to do?
Not many, so the only exploiters will be hostile corporations and governments. And maybe a few discruntled former employees working in places where they can test on compatible hw.
But if someone do crack your os/390 webserver, how many can fix that program? And how many can fix the mainframe after the hackers did their worst to it? Someone who makes such an effort won't be doing it for nothing. And yes - it has happened already.
they are shared, and far too expensive to waste time with a game that often crashes, forcing reboots.
Come on. A mainframe don't crash if used as a gameserver, no matter how buggy the game. Just as linux don't crash when used as gameserver. A good os don't crash - neither from "bugs in games" nor from internal bugs. The only reason you don't get half-life on mainframes the price for the wasted time.
<I>Also, not enough game-related hardware such as GeForce 256 and SBLive! cards are available. </I>
Well, would you want to pay the price of a mainframe-compatible sound system?
I was running Linux on some UltraSparc machines not too long ago and, while I found they were fine as workstations, as servers they seemed to be not quite ready for prime-time. We have a Sun Enterprise 4500 with 8 cpu's that we will probably migrate to Linux eventually but I'm not sure if the hardware is very well supported, particularily some of fibrechannel stuff.
:-), try ultrasparc, powerpc, mips, arm, etc.
I think that if someone is interested in doing some kernel hacking, some of the alternate architectures can use a lot of work. The x86 stuff is very well supported but if you have access to some other hardware and want to make a name for yourself (or just have some fun hacking
In Soviet Russia, hot grits put YOU down THEIR pants.
A smiley would have been a good thing. There are a lot of people out there who don't seem to know history very well (including a lot of people who aren't enough to remember it), so it was pretty easy to mistake your post as something other than humor.
Oh my god, this is so true. Porting Linux to one of the most powerful and reliable commercial platforms today, one that's in wide use through the world and backed by one of the most trusted companies in the world... What were they thinking? Stupid, stupid, stupid... man, we are so screwed now.
You're comparing apples & oranges. If you need 24x7 then 1 S/390 won't do the job (but neither will 1 PIII). With Parallel Sysplex you can easily get 24x7 reliability with 2 mainframes (both of which may run on the same box).
What's more, if you run a little piece of software called Work Load Manager and you run OS/390 you have the best scalability available. With Work Load Manager you can define performance directives for your web server. What's more, different areas of the web server can have different priorities so you can manage your web site more effectively.
While this may not sound impressive (it didn't to me the first time I heard it), someone put it into perspective with this example:
Imagine that you're a large company. You not only need reliability, but you also need scalability. The majority of your data resides on OS/390 and you provide web access to it running IBM's HTTP server on OS/390.
You have 2 customers sign on at about the same time. Customer #1 does a very small amount of business with you. Customer #2 is one of your biggest ones and you need to keep them happy. Customer #2 is also the typical 900 pound gorilla who demands responsiveness or they'll take their business somewhere else. With any other web server both customers would typically have the same priority and you need to build an infrastructure capable of meeting the needs of Customer #2 at all times. With OS/390 you can declare performance directives so that Customer #2 gets higher response time than Customer #1. Both customers are happy and you can manage your system resources more efficiently.
I don't know of any other system which can do this. If you think it's worthless, you'll probably encounter it. I bet it won't be too long before on-line banks start doing this based upon the amount of money in your account. If you have $10.00 in your account then you'll get crappy response from the bank. If you have $10,000 the bank's web site will probably be much more responsive.
Have you read the original posting?
He talked about his bank using beowulfs, not about a non-mainframe server.
Beowulfs in this context is absurd.
The post I replied to asked what my bank and my accounting department is running. They are running beowulfs.
/.-readers the name of the bank which is doing that.
Facts please. Tell your fellow
This is absurd.
My statistics was about five years ago, but I think the way it works is that you have to start looking at multiple chances.
In other words, the chance of one box going down today is 10%. However, the chances of two boxes going down is 10% of that original 10%. The chances of three boxes going down is 10% of the chance of two boxes going down which in turn is 10% the chance of one box going down, and so on.
So that's why the 99.999999% availability is most likely the correct answer. The probability of all eight machines doing down in one day from "natural" causes is statistically unlikely.
On the other hand, if the electrical workers in the street accidentally disconnect power to the building where the web servers are located, you're screwed.
You're scary.
Blar.
Client server is dying...mark my words, son, mark my words....
Blar.
Kudos to everyone for teaching me a lot with these threads. It's been my impression that many people (including myself all to often), don't have much of an idea of what "enterprise computing" actually means.
When you have 8 servers, each with a 10% chance of going down on a given day, then you have an 80% chance of at least one failure a day!!!
Granted, when one of them goes down, you only lose about 12.5% of the total capacity. But, what happens when the failures cascade from one server to another? What about all that extra software needed to make the 8 servers work together properly? The more complexity you add, the greater the chance of failure.
Now, while each server by itself may have a 10% chance of failure, what are the chances of failure after you've added the linking software and hardware? Let's assume that the linking software has a 5% chance of failure per day, and the extra hardware also has a 5% chance of failure per day. What you end up with is a 90% chance of failure per day, and of that 90%, there is a 10% chance of a total system failure bringing down all servers, per day.
-- Error: Cannot find file REALITY.SYS - Universe halted, please reboot!
PLEASE tell me you are kidding to cover up the fact that you are full of shit on this specific issue.... Have you ever actually been *in* a glass house? When we do our yearly UPS maintanence on a holiday weekend we have folks come in in the middle of the night just because it is such a rare thing to see the IBM *off*...
Can't sleep, clowns will eat me....
I did say "classic supercomputer", as in Cray/SGI vector supercomputers.
Because the I/O channels on mainframes can handle a lot more data than PC architectures. It's easier to have one extremely fast DB on a mainframe than to have a distributed DB on slow I/O devices with a lot of update conflict resolution to deal with.
Linux: more like a tricycle to a supertanker.
>Basically if you have an un-parallelizable part of the problem, then even with infinite numbers of parallel processors you can't get better than the time to solve that un-parallelizable part. Mainframes (and for that matter supercomputers) dominate where the problem can't be broken down into parallel chunks (for example you really want a bank to keep its database of accounts current so that it doesn't allow someone to withdraw the same money twice).
Thanks for posting. This is basic and easy to miss among the hype and latest craze d'jour.
Using a mainframe to emulate a mess of PCs is an expensive misuse of mainframe. Trying to emulate a mainframe with a mess of PCs gets very complicated and expensive for something that doesn't quite work.
We need a new moderation category: poster obviously smoking crack.
dave
There are different kinds computers like there are different kinds of transportation. We have fast sporty cars, big trucks, and for some transportation needs we have trains. The average company does not need to buy a locomotive. But if you have big transportation needs, the workload scale of that locomotive overshadows massively parallel.
You wouldn't want to drive one million cars from the factory to the dealer. No, first you load hundreds of cars on a train to a regional facility then truck a dozen cars from there to a dealer.
Large Data shouldn't be any different. Crunch large transaction sets on big systems, then transfer those numbers to smaller systems.
This is a boring sig
come on now! wake up and smell the corporate
budget! The S/390 is a very awesome piece of
hardware that is used and trusted by huge numbers
of companies. a thing of the past, it is NOT.
A year spent in artificial intelligence is enough to make one believe in God.
This just made my day.
It's hard to believe that you can buy a $491,000 computer online. Purchases like that usually require a large number of people just to process the paper work, on both ends. Yet IBM lets you do it with with the click of your mouse. Unbelievable.
Go not unto/. for advice, for you will be told both yea and nay (but have nothing to do with the question)
True, security by obscurity is bad if it's your only defense (and in this case it definitely isn't). But as any general will tell you, trying to fight off a hundred people is a lot easier than fighting off a hundred thousand.
Garg
Garg
Alumnus, Xavier's School for Gifted Youngsters
whats happened to the moderation on this story?
guess informed opinions are a rarity these days
what next porting linux to pcs? I cant wait
But the key will be whether ZDNet will ever admit that Linux is "Enterprise-Ready". If running on Big Iron doesn't make it so, then I wonder what does.
Cheers,
Slak
But this is a high-quality troll. Has Jesus Christ ever checked out segfault.org? You'd fit in over there
It is a nice troll, but do we need it three times? I moderated exactly the same posting down in the Ebay may bid for Sotheby's thread, as "Off-Topic", (its current rating is +3 there), and it first appeared on the Linux 2.3.48 thread yesterday (where its current rating is +4). So perhaps I should have moderated it as "Redundant" instead?
Perhaps he is trying to burn off his Karma by repeating himself...
Jeroen Nijhof
Look at all those racks filled with boxes that serve the today's Internet. What are they? PCs? Huh! I would say a custom-built mainframe. Usually, a bad one. Like a car built by putting together a couple of motorbikes.
Now these Sun E3000 and alike. Workstations? Hmmm, not quite.
I think we are talking computer systems from this vendor or that vendor. IBM has been building this class boxes for 50 years already. My neightbor Joe assembles his third rack-mounted dinosaur, so he must have lot of experience.
Protected memory, virtual memory, SMP, failover, load balancing, virtual machine...We keep inventing these things again and again...
The cost part is easy. Modern rack-mounted systems are seven-figures, too.
The time has come to have a hard look at Big Iron.
Can we get some kind of sanity check on the moderation of his comments?
He has to be a script kitty with some other bogus accounts to moderate himself. Could one of the slashdot team look into this and do something about it?
Thanks,
Bobby
Re: News from the Linux Frontlines
That was a truly hilarious troll... if I had any moderation points to my name, it would definately be moderated up!
Moderation is one of the best things about this place - I don't have to read the "Linux SuX, Windows RoX" troll posts, the mindless babble of people with nothing better to do. BUT, if a truly original troll with genuine humor comes along, it can often be moderated up as funny... that's what should have happened this time, IMHO...
Some trolls deserve better...
For those who have their settings as 1+, definately browse down to -1 and do a quick search for "News from the Linux Frontlines"... great stuff!!!
- strabo
Here in the UK, the Income Support system is ran on a group of around 17 ICL VME machines (at least they were about 4 years aog). These were quad machines and some were being upgraded to be 6 node SX models. They have some serious grunt.
The lasers printer which we spooled the covering leters on moved 6,000 feet of paper per minute. Ok, print quality was not the best you've ever seen, but it was still laser-quality output.
Do not underestimate just how much processing power a single node, 10 year old mainframe has....
When MS does it people laugh. When Linux does it its cool? I don't know about you, but Linux has one strong point. Thats small servers and large workstations. Of course its possible to get Linux to run on small to huge platforms, but is it the optimum thing? Say I'm a company putting Linux on a 256 CPU box to do some mission critical task. Will anyone from the Linux community tell me I've lost it?
A deep unwavering belief is a sure sign you're missing something...
I should have spend my last moderation point yesterday.
The real question about this is why? OpenEdition (one of the newer OS upgrades) has a unix shell and all of the same functionality as you would find on a unix platform. Its not AIX and perhaps might not have as fast of a response time as the Linux box sitting on your desktop (but remember the strength of a mainframe lies in Volumn - processing large amounts of data not small amounts quickly), but it is still
a UNIX shell. And no need to learn COBOL - java works just fine on it : )
One problem is that people do not understand that a mainframe is like a beowulf cluster in that it is composed of multiple computers but in a mainframe the different computers are optimized for specific processes (like I/O).
Considering that computers are a human artifact and that like most things created by man, they get more refine with time and usage, mainframes have evolved and solved isssues that PCs haven't yet even hit because they have been around 40+ years.
AFAIK, you are wrong about IBM's profits on mainframes. This is still a large and profitable business and IBM's monopoly position is of course stronger than ever. What changed is that industry *growth* shifted to PCs and IBM largely missed out on this shift (even after they "invented" the Intel/MS PC platform.) Thus, mainframes are still a great business for IBM, although a shrinking one.
No offense... but you really don't know what a mainframe is, what it does, or why they are used.
TERRABYTES of data pass through in mainframe every couple of hours. there are times when there are over ten thousand concurrent connections to the db2 database all doing inserts and reads, then asking for processing, then getting huge volumes of output back.
a PC cluster using ANY os, just doesn't have the I/O links, either inside the machine or too it's outside world, it just ain't there.
As to your comment about apps and what they use, you have never played with serious apps. Try
using cp to backup your harddrive to anouther one on the same bus, and set the threads priority to real time and then go and do something else.
thats a very simple example, and doesn't even come close to the type of data handling mainframes (even those from the 60's and 70's) routinely handle without a problem.
Yeesh.
-Tilde
Old truckers never die, they just get a new peterbilt
I admit it. I help build Mainframes. I have a few things to add to this discussion:g s.asp". Keep in mind that a Mainframe MIP is a little bigger than a PC MIP. Also, these are the low end.
1) These are my comments alone and DO NOT represent the opinions or policies of Amdahl Corporation. (sorry but I like my job and I've been warned before)
2) Mainframes are dead. Nobody makes Mainframes anymore. They have been replaced by "Global Servers". Honest.
3) The most expensive part of any "Global Server" installation is the Software. The license for operating systems and applications can easily be in the tens of thousands of dollars per month. This is one big area where Linux may come in. The price of hardware has come down so much that I'm told most companies view S/W costs second only to total system capabilities.
4) Most of the worlds data may still be on big iron, but most of the applications developers work on small machines. If you think it's sometimes hard to find that neat little app. for Linux, try finding it for OS/390, VM, or UTS. This is another area where Linux would help.
5) FYI. You can by low end Amdahl "Global Servers" on the web. Check them out at "http://eshopamdahl.com:80/eshop/millennium/confi
Another post had a good point too: I used to manage a System/34. The thing was huge, power hungry, had a rubber belt that came off the disk drive ocassionally, and had WIRE-WRAPPED circuit boards.
It was up for months between reboots, and was a real workhorse. It did the jobs that the BUSINESS needed, and did them reliably. And that's what you want in a business environment.
If Linux gets ported to more proven hardware platforms, that's good for Linux. It may not impress those of us who enjoy the bleeding edge, but it helps CUSTOMERS sleep at night.
aem
-a.e.mossberg
Here it is - the obligatory Beowulf post:
;-)
"Ya know if they did create the ethernet wrapper for the internal IBM network card channel you could probably run 10,000 virtual Beowulf clusters and run Linux distributed apps on them! Probably play Quake Arena well too!"
Enjoy!
The Tick - "Spoon!"
"Bah!" - Dogbert
I think that a mainframe is largely a thing of the past due to the fact that most of the investment in the money and other factors will obselete the small parts that make the mainframe.
Running linux on these things I guess would be a waste because I am sure that some crappy mainframe OS would work a little better. I mean at that level when you shell out several million dollars for the computer the OS is trivial.
Slashdot social engineering at it's finest
The whole point of modern mainframes can be summed up in one word: VOLUME. Regardless of the ground that has been covered by intel and co., the different OS developers, and the various efforts to develop high capacity I/O
interfaces, your standard PC platform just doesn't have the ability to handle the sheer volume of data that your average mainframe deals with. PC's have a long way to go before they can even begin to encroach on the mainframe realm
of computing.
Hmm silly little ol' me thought that IBM's mainframe division lost millions because of those PCs but I guess I was wrong.
Slashdot social engineering at it's finest
I worked for 7 years in a warehouse that delt primarily with MainFrame reconditioning. These bad boys are big, hot running, electricity sucking beasts. I love 'em. Not the fact that they are resource devouring monsters, but the fact that they are indestructable. IBM used to produce a controller/disk array that consisted of 4 frames per string. Frame 1 was the controller(3880), and the Frames 2 thru 4 were the drives(3380). There were several models of the drives(A thru K) and varied from (if I remember right) 5 GB to 25 GB. They were huge. Each frame had a measurement of roughly 3.5 X 2.5 X 6 foot. When a tractor trailor rolled up and dropped 40 of them at our warehouse you knew you'd be running them about 20 feet up the floor at high speed to build up momentum so you could hop on the frame and ride that mutha for the remaining 1000 feet across the 'house. If you crashed into other frames in the process, even better. These things were indestructible. We had a full reburbishing center for reconditioning them(paint guns, laquer paint, body shop for doors, texture paint, welding facility, etc.), so if you busted them up, it was no big deal. The most fun was setting up the 6 foot by 1.75 foot doors out in the woods as deflectors for paintball wars(some had 1 foot by 1 foot openings for control panels that worked well for viewing the "enemy"), or laying the 6 foot panel on the floor, standing at the 2 foot from the end point and bending the rest of the door at a 45 degree angle to create a bitchion toboggan. These doors were ultra smooth and created a comfortable ride, very fast moving, but we never developed a steering mechanism. Sort of like the saucers but rectangular. Back to the point...when trucks would arrive, sometimes the driver would inform us that they took a turn too fast and all of the frames had tipped over in the back of his vehicle. We always expressed concern for insurance purposes, but we'd just put them back on their feet with the forklift after the driver left, fire them up, and 90 percent of the time they'd diag with no problem. Now that I've been working with P.C.'s for who knows how many years, I've found that if you do something as simple as knock over a socket 1 machine, the processor pops out or dislodges.
If anyone has connections with a refurb center(I do but I'm not here to advertise), price an old IBM 3090 system with 3 to 6 strings of 3880/3380 drive arrays(probably get the whole setup for under 3 thou.). The processor(3090) is water cooled, and has some fantastic high test rubber hoses for the whole anti-freeze/cooling setup. If you can get extras, they make great garden hoses by adding male and female ends from your local hardware store. Otherwise, you'll need about a thousand square feet to set up a full working mondo.
Good Luck.
The word you want is RELIABILITY.
That sums it up nicely; at the shop I've been working for the last three years, we cycle our LPARs about once a month as a precaution, but we've only had two crashes on our "production" (read: highest traffic) MVS system in that time frame. No crashes on our secondary production system (where users were switched to transparently during the primary's crashes), no crashes on the VM system and none on the development MVS system. On the other hand, I'm counting myself lucky that I've had my NT (Nice Try) running for nearly a week...
There are still things that PCs can do much better than any mainframe implementation I've yet seen. (Like DW/370, a seriously ugly "word processor" for the mainframe; don't try this at home...) On the other hand, I still get a schadenfreude kind of kick watching our LAN people re-invent the wheel on issues of security, change control, network, inventory, etc. 'cause they think any solutions connected to the mainframe must be inately primitive and ineffective... ;p
(Now if I can just get my VM sysadmin to allocate me some more disk space...)
"I'm a scientist! I don't think, I observe!" - Dr. Clayton Forrester
What happens if you make a Beowulf cluster of mainframes?
IBM calls it Sysplex.
"I'm a scientist! I don't think, I observe!" - Dr. Clayton Forrester
In the article he explains why main frames *arn't* a thing of the past. Main frames still power banks, colleges, and government all over the place because nothing can handle the loads quite like they do. Also VM and the virtual machine capabilities of mainframes make them very flexible for clients. And the point of running linux on a mainframe is that you can section off a small part of the mainframe and run linux in as a webserver, samba file server, fire wall, or if you have employees that need to use a UNIX for their work they could have a linux vm running on a small section of the mainframe for them to work on. Another perk is that if any of the vm's go down on the mainframe they can just be restarted and the rest of the mainframe clients don't even know you are gone, or realy that you were ever there. In short mainframes are not a thing of the past for corporations, or organizations that can afford them.
-----
Can I Play With Madness?
I thought it was funny.. the first time I saw it.
Looks like there are still lots of people who haven't seen it though, as it has so far managed to get 7 positive moderations (versus 8 negative...)
And yes, a plain old pee cee can easily serve those 1100 hits/second and saturate a 100baseT card. My old 200MHz Pentium Pro does it quite easily, with CPU cycles to spare. Of course, it runs thttpd and FreeBSD 3, not Apache and Linux. According to my measurements, running Linux would slow it down by a factor of two, and Apache would slow it down by a factor of five. Running both, with a factor of ten less performance, my old machine would indeed not be able to saturate its network card. No doubt this is where people got the idea that they needed multiple machines to serve mid-sized web sites - when they run inefficient software, they are right!
I expect, then, that you don't run a bank (That's a fairly safe bet, from the rest of the comment :). Your standard university programming problems typically don't need to process millions of transactions a day, have uptimes of over a decade, keep multiple off-site backups (securely), print reams of paper in seconds flat, support thousands of concurrent database accesses etc., etc. Big transaction-processing applications use mainframes because they have to, not because they can.
and I haven't actually wrote one myself that could do anything that I would like. Programming is quite difficult
What do these sentences mean? That you've not written a program that needs a mainframe? Few people do on their own. Or do you mean that you've not written a program that does what it's supposed to? I'm sure we've all written progs that don't behave well...
IBM are in the process of writing a redbook (manual) on this, check out http://www.redbooks.ibm.com/residencies/LS-J003.ht ml
For information about running COBOL under Linux (or whatever), check out "http://www.gnu.org/software/cobol". BTW, I don't know any COBOL and I also don't know anything about GNU COBOL except that it exists. But since you seem interested, you may want to go have a look.
Cheers //Johan
Installed the Bubblemon yet?
Actually the Mainframe Software costs are the killer. Sure the hardware initially costs a lot, but the on going software costs are what REALLY make the box expensive.
Why do you think people are embracing Linux on the BIG IRON.
Just my thoughts as a raised floor dweller.
Technology is only a vehicle. People are the ones that drive it.
any chance of it running on the as400? first post?
Do you see the sig? Do you have it in your sights? Why yes, Miss Moneypenny...
I may agree that people today would not buy a mainframe ( although I never run any big system myself ). But what if a company has already one of them and does not know how to use it anymore?
... ). And to me the so much vented 'network computer' idea resembles a bit the 'obsolete' mainframe concept ( or it does? ).
Besides, this is a wonderful hack, independently by its pratical usage ( and why people run Linux on Palmtops?).
Oh, and concepts in computer science never dies. Look at how SUN re-cycled the IBM Virtual Machine concept with Java ( ok, it's software instead of hardware, BUT
*WHINE ON*
I posted this story last thursday, but got ignored. This is life. It's OK. Just let me whine and I'll feel better.
*WHINE OFF*
Ciao
----
FB
Ok, i know this is from IBM's Marketing department but it does a good job of describing the differences between a unix (or a pc) and a Mainframe system. Its not the processing power, its the IO bandwidth. A PC or Most Unix boxes aren't even close. i'm not saying its the end all be all of computers but for what its used for, Lots and Lots of Transactions and no downtime and thats 99.9999% uptime. http://www.s390.ibm.com/marketing/gf225122.html
anything that can make me smile today prolly should be sitting around 5.. course I like cynical trolls but oh well.
Amen to that. I wish this had happened just a couple of years ago. The outfit I used to work for has a huge IBM mainframe network (running VM as it happens) with thousands of concurrent users on every day. They are reluctantly migrating to an NT-TSE network which is costing them huge amounts in server kit and staff time, largely because their end-users won't put up with the clunky block-mode character UI any more. At least this way they could give all 10,000 users a virtual Linux both with X and a wide variety of existing apps, especially graphical access to the Intranet they're building.
Why would a mainframe have any advantage over a bunch of intel boxes? Processing power? Get enough athlons and you can out-process any mainframe, for easily distributed applications such as webserving. High availability? Sorry, even if one intel box is 90% reliable, 8 are 99.999999% reliable. What exactly are you using to judge "best"?
ok then your [sic] infringing on my copyright! Could you as [sic] me next time before STEALING my comments for your own?
better yet, get FreeBSD 4.0, and use jail().
ok then your [sic] infringing on my copyright! Could you as [sic] me next time before STEALING my comments for your own?
The post I replied to asked what my bank and my accounting department is running. They are running beowulfs. I wasn't saying that beowulfs are god, or that they can replace mainframes, just that in my case, his examples are wrong.
kwsNI
Somehow, I think he was joking....
Its NOT about power, its about IO. A 16 node Xeon box is great for computation, but its IO is still bound by the PC way of doing things.
When MS does it, there are API subsets with enough differences to bite the developer with the huge MS GUI always coming along for the ride. When someone does it with Linux, they are free to port any bit of API that they need and discard the rest.
If you are putting Linux on a 256 CPU box, you may well have some very valuable input to the kernel developers about scaling beyond 16 CPUs.
Is it the Optimum thing? That depends on your criteria. If your primary criteria is speed to market, maybe it is the right thing to do. If your criteria is to squeeze the highest possible performance out of your hardware, the answer might still be a free OS, but you'll spend time and money to optimize the kernel. It might still pay if you have few other options to buy an OS for that platform.
Anomalous: inconsistent with or deviating from what is usual, normal, or expected
Anomalous: deviating from what is usual, normal, or expected
Canard: a false or unfounded repor
Linux is an OS that has been implemented on tiny devices like the Itsy and huge platforms like OS/390. There is no other OS that has such a large range of targets.
Now, just imagine trying to use an Itsy "rock and scroll" driver on a bookcase sized mainframe.
Anomalous: inconsistent with or deviating from what is usual, normal, or expected
Anomalous: deviating from what is usual, normal, or expected
Canard: a false or unfounded repor
Certainly. If IBM distributes the network driver as part of OS/390, and not as part of Linux, then I would think that there would be no problem with the GPL. IBM might have to release the header files with Linux under the GPL, though, but I don't think that that would release any of their trade secrets that are in the driver.
(the largest iron WinNT was ever ported to, BTW, were Sun and Alpha ports--and those two ports are supposedly being discontinued)
WinNT didn't have a Sun port; the non-Intel versions were MIPS, PowerPC, and Alpha, none of which survived to make it to the shipping version of Windows 2000.
In fact, the ONLY Big Iron I know of at ALL that uses anything close to a GUI are a) the Alpha and Sun ports of Windows NT and b) a terminal and configuration program for OS/2 designed to act as a console for booting AS/400 boxen running OS/400 (in other words, the OS/2 program largely replaces the blinkenlights).
OpenVMS also has a GUI, the X11-based DECwindows (one can reasonably argue that VMS is a minicomputer OS on steroids, and not true Big Iron at all, but it's certainly more mainframe than its bastard stepchild NT).
-- When did "innovate" become a synonym for "suppress competition"?
Information wants to be free -- but informants want to be paid.
If my mainframe doesn't have a 3d flight simulator interface to the filesystem, it just ain't worth it. (think Hackers...)
---
I'm not ashamed. It's the computer age, nerds are in.
They're still in, aren't they?
This is a little of topic but what is the definition between a mainframe and a super computer? I'm just curious people are always throwing around the term super computer are there specific guidelines between what's a mainframe and what's a super computer.
I've never noticed it before but my thinking cap does sort of resemble a hockey helmet
There aren't any number of PC's that can be connected and controlled in a cost effective way to compete with what I consider to be a mainframe, a complex of 3090-class processing engines like we have in our datacenter in Florida. We crank out over a million bill images each day and process over 3 billion call detail records (wireless phones) each month. If you can describe a PC-based solution that can do this volume for under $200 million, I'm all ears! :-)
There is much cruelty in the universe, John.
Yeah, we seem to have the tour map.
I am the Lord.
I am the Lord.
God Hates Moderators.
I mean, I fucking DIED for you people! I died on the cross, AND I repost funny comments! When will it be enough? Won't you ever be satisfied? *sob!*
Now you made Jesus cry. You're going to burn in Hell, mister! I'll see to it personally!
*sob...*
I am the Lord.
I am the Lord.
God Hates Moderators.
Posted by CmdrTaco on Sunday February 27, @10:36AM
from the rob-sucks-tarballs dept
Linus Torvalds, creator of Linux, accidentally hit his keyboard with his elbow today. We have yet to receive confirmation that the resulting code will be be included in the next development kernel, but we can never be too sure. Here is the code in full:
This won't compile under GCC, so we can only assume the code is pretty experimental. Look for the tarballs to be released this evening.
Torvalds comments, "What? Oh, yeah, I accidentally hit my keyboard with my elbow when I reached to get my tea. What? Is it part of the new kernel? You're kidding, right?"
We'll update the article as soon as we get more information. The Linux world hasn't been in such frenzied anticipation since the release of kernel 2.3.48.9.2.7.42, which was about ten minutes ago.
Interview: Alan Cox farted
Posted by Hemos on Sunday February 27, @10:34AM
from the whats-that-smell dept
Linux guru and hacker-extrodinaire Alan Cox farted earlier today. What do you think this says about the future of Linux development? Alan's ass will respond to the highest moderated posts later this week.
ESR and JonKatz to participate in "Zealot Deathmatch"
Posted by Roblimo on Sunday February 27, @10:33AM
from the die-bitch-die dept
Open source proponent Eric S. Raymond and Slashdot nutcase JonKatz are reportedly organizing a "Zealot Arena Deathmatch" to raise money for the Apache Software Foundation. The fight is expected to be a tough one, because while Katz is genuinely insane, ESR has the power of girly, elfish looks. A spokesman from Apache says that, "while we don't encourage violence, we'll do anything for money."
VA Linux aquired by Klingons, Rob bows down to new alien masters
Posted by emmett on Sunday February 27, @10:32AM
from the star-shit-enterprise dept
VA Linux Systems, owner of Andover.net, owner of Slashdot.org, owner of Rob's ass, was officially aquired by the Klingon Empire earlier this morning. The Klingons, who have recently taken over Kellogs, GM, and Disney, are looking forward to absorbing more major corporations in the near future. The US Government is discussing investigating the Klingons for holding a monopoly over "every aspect of our lives", to which the Klingons responded, "Puny human scum! I will crush you like a bug and feast upon your steaming entrails." Finally, some competition for Microsoft!
Red Hat and VA stock at all time high!
Posted by CmdrTaco on Sunday February 27, @10:31AM
from the i-am-so-rich dept
Dude, have you heard the market reports today? I am so fucking rich! If this keeps up, I'll be able to stop doing this Slashdot crap! Hell yeah!
I am the Lord.
I am the Lord.
God Hates Moderators.
(I learned Unix on Vaxen and used to run an old PDP-11. I thought for sure the reference to "old DEC boxes" at "universities" would clue people in to the joke. )
Yeah, but those words with multiple syllables are located half-way across the country, and for some reason I can't seem to get them here. It is almost as if the connection between here and there were being jammed with millions of requests and response. It is a very odd occurance that I am currently unable to explain.
- Playing games - sure, they're powerful computers, and would make helluvacool half-life or UT servers, they are shared, and far too expensive to waste time with a game that often crashes, forcing reboots. Also, not enough game-related hardware such as GeForce 256 and SBLive! cards are available.
- 3D graphics - mainframes powerful renderers, not good for viewport stuff in 3DS, for the same reason they aren't good enough for games.
- Other typical PC user stuff - most home computer programs are not demanding enough to demand a mainframe, therefore are just not cost-effective
Sure, all this may be because I use Windows, or maybe I have some management tendencies lurking within me. However, I will probably never need a mainframe for the tasks I perform, and most people never will."Assume the worst about people, and you'll generally be correct"
Even more than that though, a lot of times, people can say more in a -1 post than a +2 post. They don't say it in a straightforward way, but there's this huge spin of sarcasm and irony there. Look at parent post. True, most of the comments are kind of cliched, but he said them in a damn funny way.
Oh, BTW, this doesn't even come close to the "I love monkeys" short story troll. Now *that* was funny. I about messed myself when I read it. Here's
a link, not a link to porn, but a link to the story if you're curious about that I love monkey story.
No sig is worth reading.
I think the only point you got wrong is that of REXX being crappy/unusable/etc. The language (with the notable exception of ObjectREXX) is a really good general-purpouse programming language that happens to have some more power in strings and sending OS requests down the line.
I happen to be somewhat touchy on articles that claim a language to be "bad/ugly". If we forget about APL, it's all just different languages that work fine for different tasks.
And for the clustering vs mainframe discussion:
How do you chop a pound of beef?
A. You hire 1000 chinese to do it B. You run it through the meat chopper
The danger from computers is not that they will eventually get as smart as men, but we will meanwhile agree to meet them
Come to think of it, an AS/400 Linux port would be very sweeet! The Small Iron from IBM (RS/6000 and AS/400) are extreamly reliable, optamised for I/O transactions, and have RISC architecture.
And it has to be better to use top instead of wrksyssts
Slashdot. When you positively, absolutely have to argue about it. Darryl Adams mailto:rampart.pnc.com.au
With that famous OPUS era batch file NERF.BAT????
Slashdot. When you positively, absolutely have to argue about it. Darryl Adams mailto:rampart.pnc.com.au
As much as I understand from reading in various web sites - the Network drivers were written from scratch without modifying any existing code - so IBM doesn't need to release the code for it (and Linus allows binary only module)
Hetz (Heunique)
Look, I understand what you're saying, but this isn't a very good argument. It's based on security through obscurity, which most security professionals agree is useless. If there's a buffer overflow, it should be fixed, not ignored because nobody knows what to do with it. I'm not trying to pick a fight here; I certainly understand your argument. I just don't think it's a good one. Let the merits of Linux/390 vs OS/390 stand on their own. If one is actually more secure than the other, fine. But relative obscurity is not a security feature.
Funny, I've been hearing this for ten years now. "The Mainframe is dead." "Unix is dead." Sure. It's beyond me why people try to predict the future in this industry. I've yet to see someone who can do it without emabarrassing himself horribly.
Running linux on these things I guess would be a waste because I am sure that some crappy mainframe OS would work a little better.
What makes you think so? IBM did the port after all; I would assume they know what they're doing. Besides, if the mainframe OS is crappy (which they aren't) why would it be better???
Try again. IBM did not lose money on mainframes last year. It isn't a high growth market, but it still makes up a large portion of IBMs money stream. Maybe 83-93 IBM lost money in mainframes, but last I check they were shipping mainframes as fast as they could make them. (part of that was Y2K hit mainframes the hardest) Nothing can process the volumn of data a mainframe can. (Okay, a cray can do it, but your talking mainframe class price too)
Sounds like the job for a lot of PCs with load balancing or perhaps some form of a beowulf cluster for the actual processing. Also
you could have multiple high speed printers I really can't believe that you need a mainframe to increase something that is on the
device end like a printer.
Beowulf? For billing applications? Do you know how complex parallel programming is?
And how do you like the idea of maintaining 200 PCs: I'd suggest it's a lot more difficult and expensive than maintaining one mainframe. Especially when most mainframe components are 100% hot-swappable: hell, you can do a microcode update on these things without any downtime.
Of course you'd have multiple printers, although the last mainframe printer I was attached to did 120 ppm, and had its own monitor.
--
Failover, sustainability and load balancing are key when building a web infrastructure. You are better off with fifty rackmount P III's than one mainframe any day of the week. That is why no one uses or advocates using mainframes for web serving.
A S/390 has fault tolerance features that the PC world can only dream of. Failover and sustainability are really not a problem in mainframe world, believe me.
--
Nothing beats my DOS based webserver for pure hittage. I run Caldera OpenDOS {used to be available from www.calderathin.com, but I think now it's lineo.com) with packet drivers for an NE2000 and I run the boa webserver[www.boa.org, it's really cool, a single process webserver] ported to a DOS TCP-IP stack. For reliability, it's number 1. I pity the fool that tries to DOS my DOS box.
nWo for life!
------------
a funny comment: 1 karma
an insightful comment: 1 karma
a good old-fashioned flame: priceless
this sig limit is too small to put anything good h
Sorry, but you are figuring your reliability the wrong way around. If you only need 1 or 2 boxes worth of horsepower, then, yes, 8 boxes provides high availability.
OTOH, if you are attempting to gang multiple Intel boxes together to replace mainframe capacity (ignoring I/O bandwidth in this example), then the more boxes you have, the greater the chance one of them will be down.
Let's give PC hardware a break, and assume 99% reliability instead of your figure of 90%. If you need 50 Intel boxes to replace a mainframe, at 99% reliability each, then overall reliability (all 50 machines working at the same time) is 0.9^50, or about a 60% chance. In other words, at any one time, there is a 40% chance 1 of your boxes is down. If you need at least 50 working boxes, then you better have >50 boxes linked together.
It gets worse with more boxes. Do you need 100 boxes working all the time? The odds of keeping 100 boxes running (each with an individual chance of 99%) is 0.9^100, or about 36.6%.
Meanwhile, that mainframe you were so keen to dump probably had an uptime availibility of at least 99.95%.
There's a new reason to start reading comments at 0 and -1...when there are funny trolling posts like this one that slip through the moderators. People see something that is anti-rob anti-slashdot or anti-"linux party line" and they moderate it down.
But this is a high-quality troll. Has Jesus Christ ever checked out segfault.org? You'd fit in over there.
-- Truth goes out the door when rumor comes innuendo. -- Groucho Marx
Odd huh? At the low end, mainframes are not that much more expensive than large Intel SMP servers. At the high end of course you're getting a boatload of hardware but more importantly what you're getting is a discipline, process and operational model refined over the last 40 years. The point of mainframes is not that they cost fortune, it's that they are built and operated to run and run and run and run.
When I was fulltime in MVS ESA world we planned for WEEKS for an IPL (reboot) because they happened maybe twice a year for planned upgrades and never from unplanned events. In fact it was a ground breaking event when we got a VTAM upgrade that didn't make us regen VTAM following LU/PU changes. And 3745FEP's and NCP? Forget it, I've never seen one crash. Ever. In 20+ years. ESCON Channel I/O is 136Mbps or 17MBs out of the box and there are many approaches to multiply that in the aggregate. Throughput? We put CICS in its own region running as an uninterrupted communications task on a low end 9021-200 and got 500tps almost 8 years ago.
The problem though with putting freenix on a mainframe though is the fs. Native freenix fs structure probably can't be implemented well over the native IO or over the native mainframe file structures, PDS's etc.
I'm currently an admin with a few years' experience on various flavors of unix-alikes on peecee hardware. I'd love to get some experience on big iron, but I'm not sure how to go about that.
If anyone is willing to spend a few hours a week mentoring, I'd love the opportunity to learn from someone in the Boston Metro area with experience. You have scut work to be done? I'll do it, if I get to learn. I'm busy, but I can spare a few hours a week. If you're interested and like to teach, please drop me a line (jerkbob_at_pobox_dot_com).
Apologies again about the OT post...
--
A host is a host from coast to coast...
A host is a host from coast to coast...
Unless it's down, or slow, or fails to POST!
I am offering mainframe web hosting for $5 per month. Yes, $5 per month.
Service includes:
100 Meg of disk space
Free name registration
50 Gigbytes transfer / month.
One time setup fee of US $3,000,000 due at signing.
Haha. Obviously you've never seen a mainframe power cord.
... the motor-generator. The electricity from the power plant is not used to directly power the mainframe. Instead, the power is used to drive a motor, which turns a flywheel, which drives a power generator, which provides power to the mainframe.
... theoretically without killing the mainframe. Theory is very nice ...
It's about the thickness of your wrist, and uses a locking connector. I believe that it requires a wrench to detach the power cord from the motor-generator.
Yes
If there is a power spike on the lines, the power spike is absorbed by the momentum of the physically rotating flywheel, thus protecting the mainframe. Beats the hell out of a "surge suppressor" any day.
This is standard equipment on a 3090.
Oh yes, and we have direct connections to two Commonwealth Edison power substations. If there is a blackout on our primary substation, a HUGE, frightening looking switch is automatically thrown, and the mainframe power is switched over to the other substation
IBM came out with a rushed product called AIX/370. We ordered a copy to demo it, and it sucked. It was a dog.
This Linux port is actually the first credible Unix implementation for the 390. (i.e. something that systems programmers are excited about, as opposed to disgusted by.)
Been there, done it ;-)
Most of them run NetBSD with ease. I have actually used some of them as file servers and despite their pathetic CPU power (around a 286-386) they stuff a 10MB ether to the point of congestion (unfortunately there are no higher speed interfaces for them).
They suck for web servers, DNS or whatever else where latency and execution speeds are crucial but they make damn good fileservers after you replace the hard drives with a recent SCSI. And after such surgery they just work. Boot them once and forget them forever.
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/
It's not as good as that 386 running in my basement...it gets 1000s of hits a day and never misses a beat.
(hits with a baseball bat that is.)
My plan is to pimp before they realize I'm a jackass. Hit 'em hard and fast.
I do not belive that mainframes will die anytime soon. I actually hope and predict that they will be around for a long time. /.-ers, the right thing is not always Linux. (don't get me wrong, I am a registred Linux user Linux Counter)
I have said for 18 years now that we must use the right technology for each system. As much as that may hurt some
Mainframes and their OS is extremely good at processing enormous volumes of data and transactions. They are real bad at interactive processing and should not do that. Linux is one flavour of the very good Unix system that is excellent for servers and power workstations. With the introduction of Linux, Unix will be accessible to the "common user", there is a bit left but it will be there Real Soon Now (tm).
I have seen to many projects becoming disasters because of using the wrong technologies. Let's stop the bickering and try to collectively use computer technology wisely.
Look at Lucent WaveLan IEEE pcmcia drivers. You can get one that is GPL (in source) but lacks support for all features of the card (gold version w/ 128bit encryption running at the full 11Mbps). Or you can get one that is binary with a GPL'd source wrapper acting as a go-between between the proprietary product and the GPL'd OS.
The latter is Not bundled AFAIK with any distribution but must be obtained directly from Lucent's website. But the latter driver does support all the features of the card. It seems that here Lucent decided to cover their arses by writing a wrapper to their proprietary bit that was source GPL talking to the pcmcia_cs and the kernel.
So, which is right? Just including a binary or do you Need to write a wrapper?
Adults are obsolete children. - Dr. Seuss
I found this discussion waaay too late for anybody to see this, but I'll post it anyway for my own satisfaction. :)
Could mainframe linux enable mainframe NT?
Install linux into a mainframe VM,
then install VMWare under the linux host,
then install NT under that.
It's sufficiently recursive to make me smile if nothing else. And we haven't even talked about squeezing Transmeta codemorphing in there somewhere. {chuckle}
UNIX on mainframes isn't new. IBM has had AIX/390 out for a long time, and Amdahl had their UTS (UNIX Time Sharing) for their (IBM plug compatible) mainframes way back in the mid 80's.
As far as old DEC boxes, I was using UNIX (specifically 4.2 and 4.3 BSD) on VAXes back in the mid 80's. DEC hardware was the original host for UNIX (the PDP-7), and was the most popular hardware for UNIX in the 70's (PDP-11 family). I know for sure that NetBSD still has support for at least some of the VAX hardware, and you can get licenses from SCO (the current USL owner) for free to run older versions of UNIX (V6/V7) on PDP-11's. Once you get a V7 license you can get a copy of the 2.11 BSD distributions from I believe an organization called the PDP-11 preservation society.
Read the FAQ. Once a person has built up a certain level of karma, every post they make automatically has a +1 bonus on it unless they specifically check a box against it when they post. Likewise, a person who accumulates a certain level of negative karma automatically gets a -1 penalty on every post (which they can't opt out of). Moderation seems to work. Most of the annoying posts (first posters and trolls) tend to get moderated down quickly, and only occasionally does a particularly good post get moderated above 2. I very rarely see any worthwhile posts moderated to -1, so you can fairly easily read with your threshold at 0 and not miss anything good. Sometimes I see good, but not outstanding posts that are at zero (anonymous cowards do occasionally have something worthwhile to say), so I wouldn't personally read with my threshold set above 0. In general, I am pretty thick skinned, so I usually read with my threshold set at -1.
OK, all you entreprenuers, are you listening? I would LOVE to host in an environment like this,
assuming it could be done for a price that's competitive with colocation/dedicated box services. From the numbers I saw in the article, it seems (at least as far as HW goes) you might be able to be very competitive indeed.
Call me when you want to beta-test. Or even sell.
Tweet, tweet.
1. Losing the opportunity to make millions is not the same thing as losing millions. And even if it were, you're off by about three orders of magnitude - think billions.
2. Last time I checked, IBM's revenue was larger than Dell's, Gateway's, Compaq's, and HP's. Combined. Tell me again about IBM's economic failure?
One can really tell that you didn't read the article. I mean.. you can REALLY tell.... Had you read the article, you would realize what rubbish you are talking. Who said they were running it on the bare metal on the mainframe? Do you know what a mainframe really is? Read the article.. it's VERY good..
Hmm.
Though, a 100Base network can't actually do 100Mbps between hosts. 100Mbps is simply the number of symbls that can be put on the channel.
With ethernet overhead, and other protocol overhead, including handshaking, you'll find the maximum throughput for something using tcp is around 85Mbps. And that's if only two hosts are talking.
But if the 99,900 of them are script kiddies who can only use known exploits, they aren't hard to fight off. The 100 knowledgable and motivated ones are the problem either way.
Script kiddies are a problem, but mainly to the 95% of sites that aren't maintained, the ones that have admins who read bugtraq, etc, are fine.
If you're on the ball as an admin, you shouldn't have many script kiddy problems.
It got a 4, which I guess it deserves, but posting the same thing twice is a bit weak.
The original is here - in the EBay story, BTW.
Now the desktop/small server products are tremendously powerful (200-900 MHz is a ridiculous amount of processor power) and are often underused because they are so cheap (Hello, SETI@Home!). The "minicomputer" devices have taken over many of the former "mainframe" applications, and many actually use parallel microcomputer designs (ie, Sequent, Stratus). The "mainframe" has fairly fast processors and is surrounded with very fast I/O devices and I/O processors. When dealing with the huge amount of data which global companies produce and manipulate, often parallelization of data handling is more complicated than using big iron.
Remember, IBM also sold a lot of IBM PCs. They lost a lot of PC business to competitors, particularly when they tried to require MicroChannel use. Most PCs just don't use so much data that they needed MicroChannel, and the additional licensing expenses just weren't worth the minor benefits. During the same time period there was a gap in their minicomputers with their old-tech S/36 before the AS/400 was developed. The big iron is flashy, but a lot more people needed the smaller machines. And there was a lot more competition once the microcomputer technology provided engineers with fast processor components for competing designs.
Evangelists, please remember this the next time Microsoft gets its panties twisted about "scalability".
Microsoft: "Our OS scales from a hatchback to an SUV"
Linux: "Our OS scales from a motorcycle to a freight train"
--The basis of all love is respect
By the same token, Linux will blow a mainframe also-ran Unix out of the water.
--The basis of all love is respect
Professionally doing the e-commerce thing, I am constantly running into the same problems: bandwidth and performance. It's not enough that the program does X, but that it has to do so much X in so little time on our hardware.
This may be an interesting development. A lot of these new outfits can get capital for the asking but not developers. Thus, mainframes are easy to pick up, but mainframe developers are almost impossible to hire. An S/390 Linux port allows you to use your existing Unix staff, with some mainframe sysadmins and minimal retraining, to use high-bandwidth hardware.
obSlashdot: What happens if you make a Beowulf cluster of mainframes? ;^>
--The basis of all love is respect
Errr, what time frame are you talking about? No sane person I know would tell you that mainframes are the be-all and end-all of computing. No computer is perfectly suited for all tasks. When was the last time you saw a tape robot hanging off a PC that was acting as something other than a controller? When did you hear about PCs churning out billing statements for a million customers twice a month? (And finishing the actual print job in a day?) It's just a matter of tuning your platform to the task, and mainframes are seriously tuned for big volume.
Yes, I got the same impression, that they wrote it from scratch.
I'm not sure this matters though. Section 2b of GPL V2 states:
"You must cause any work that you distribute or publish, that in whole or in part contains or is derived from the Program or any part thereof, to be licensed as a whole at no charge to all third parties under the terms of this License."
Notice the phrase "that in whole or in part contains".
Naturally, the driver itself is probably not derivative of other drivers, except for trivial similarities. However the distribution and more to the point the kernel at runtime is a work that includes GPL'd components
Allowing clean room developed code to be linked (even at runtime) to GPL'd code would be a huge loophole in GPL. It would allow you both to create proprietary dervied works, and to expropriate components of GPL'd works for incorporation in closed products, simply by partitioning the code into sections by licensing.
I believe GPL 2 envisions this:
"If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works."
The question is, can the network driver "be reasonably considered indepependent and separate". (And to a lesser degree, is the network driver distributed as a separate work.)
Any thoughts?
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Failover, sustainability and load balancing are key when building a web infrastructure.
Sure, but do you roll your own, or do you buy it nicely packaged up in a cabinet with complete system wide support from the vendor?
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
I'm surprised that nobody else mentioned this, but the article says that the network drivers for the IBM port are closed source (but apparently free beer).
Isn't redistributing a kernel with a non-free module a violation of section 2b of the GPL (the viral clause)?
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Actually, if I wanted to make an incredibly powerful and secure Web server, I'd run it on regular OS390, not Linux/390. Here's why:
The latest OS390 versions come with their own Web server, which is a variant of Apache (the variant part comes because of a difference in the way mainframes handle processes). So that part is already there.
Now, if you run the real Apache on Linux on your S390, by their nature you have thousands and thousands of people who know the strengths and weaknesses of those packages. Only a relative handful of people in world could hack an OS390 system.
How many of you, if you discovered a buffer overflow situation that would let you enter a command string on such a system, would know what to do?
Garg
Garg
Alumnus, Xavier's School for Gifted Youngsters
Um.... sounds like you don't have a clue what you are talking about, and you probably didn't read the article either. Go check out information about transaction processing, reliability, and mainframe I/O structure... then come back and apologize to all of the nice people you have offended with your total ignorance... Beowulf clusters are great for lots of things, but PCs have *terrible* I/O throughput. It's a Big Deal(TM) for lots of things, and I think you should read some more before you run your mouth off.
"It's tough to be bilingual when you get hit in the head."
Then you're doing something wrong... a 10-node PIII 550 Mhz System can stomp all over any RS/6000 out there. I administer a 16-node PIII-500 Beowulf and we routinely get Flops competitive to an Origin 2000 (for 1/30th of the price).
It all depends on: a) your application, b) your choice of networking hardware and software. Flops-intensive software will fare *very* well on a Beowulf. I/O-intensive stuff (like retrieving huge amounts of data off an RDBS) would most likely be better off on a mainframe.
Check out this site and tune up your cluster...
engineers never lie; we just approximate the truth.
Your comment is a long way over the reality horizon and still accelerating.
Intel hardware has a long way to go before it can touch big iron for sheer I/O throughput, which is what a mainframe customer wants. 16-way is barely the right order of magnitude.
Win2K is so hideously unstable that it has no place in the *real* machine room. MS claim 99.95% reliability. Well, that sucks. Given three minutes for a reboot, that's one crash every two days. If I have 5,000 users logged on, I can't afford that sort of downtime. I need genuine 24*7 availability. That means a system that can cope with 5,000 users, a number of whom are developers, hacking away processing tens of millions of records a day and never crash. Not once. Given an operational life of maybe ten years I won't tolerate a single unplanned outage. Sorry, guys, but the only boxes that do that are mainframes.
Having said that, I'd quite like to run Linux in a VM.
Name me one bank running their systems on a cluster of PC's!
I'm an MVS contractor, makeing a nice living from working on these "neolithic dinosaurs". Last year I worked for HSBC - the Hong Kong Shanghai Banking Corp. They are known as Midland over here in the UK. I know what banks run their stuff on, and it sure as hell isn't a Beowolf Cluster!
No flames intended but you're talking a whole crock of shit here!
... Even if it is now owned by the japanese.
Talk about IBM clones... B-) It's hard to get more accurate than one done in the company started by the guy that used to design 'em for IBM.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
I read the other day that Charles Schwab has installed 6 mainframes in the past year and that they are taking new ones as fast as the vendors can deliver them.
There are still some places where the dirt cheap PC style boxes have not infiltrated. As many people have pointed out, mainframes can handle much higher transaction volumes and have much better uptime figures than any other kind of system.
Also, remember that many of the outfits that use mainframes have huge amounts of legacy code in production that might not port so easily to a beowolf cluster, or anything else.
The costs are huge but you need to put them in perspective. I was talking with one of the local Sun techs in my area and he had recently been out to work on an Starfire that a client kept as a hot spare. This is a $2 million dollar machine and it's just the backup! The tech thought it was overkill but the client, a major brokerage house, does $4 million an hour in transactions. The spare machine would pay for itself ini preventing only 30 minutes of downtime.
For the organizations that really need them, the big iron is still worth every penny.
I can't believe how this post has been moderated up to 4 (at time of writing) when the poster obviously hasn't even read the article. Read it. It's very informative and interesting, especially if you don't know what modern mainframes are capable of.
HH
Yellow tigers crouched in jungles in her dark eyes.
She's just dressing, goodbye windows, tired starlings.
Don't know why you can't read the article (slashdotted maybe?) but youre general impressions are completely wrong. With the S/390 mainframe, multiple operating systems can be run simultaneously. You can even debug unstable, crashing OS's without taking the machine down. Very cool.
And are IBM 'brain dead' for porting Linux to their own mainframes? I don't think so.
HH
Yellow tigers crouched in jungles in her dark eyes.
She's just dressing, goodbye windows, tired starlings.
Errr, what time frame are you talking about? No sane person I know would tell you that mainframes are the be-all and end-all of computing. No computer is perfectly suited for all tasks. When was the last time you saw a tape robot
hanging off a PC that was acting as something other than a controller? When did you hear about PCs churning out billing statements for a million customers twice a month? (And finishing the actual print job in a day?) It's just a
matter of tuning your platform to the task, and mainframes are seriously tuned for big volume.
Sounds like the job for a lot of PCs with load balancing or perhaps some form of a beowulf cluster for the actual processing. Also you could have multiple high speed printers I really can't believe that you need a mainframe to increase something that is on the device end like a printer.
Slashdot social engineering at it's finest
Well, the one I work for does. As a hint, it was "attacked" two weeks ago. The traffic this site gets is what I base my comments on.
Its a lot harder to yank out two hundred power cords than one.
Thats certainly for transactions, not serving html.
Failover, sustainability and load balancing are key when building a web infrastructure. You are better off with fifty rackmount P III's than one mainframe any day of the week. That is why no one uses or advocates using mainframes for web serving.
While a mainframe may be able to handle all of the CPU needs for EBay, no single network card can handle all of the traffic in an affordable fashion. And even if one did, EBay would be silly to put all of their eggs in one basket, regardless of its power.
What happens if you make a Beowulf cluster of mainframes? ;^>
Actually, you can virtualize a whole bunch of Linux systems on one mainframe and Beowulf them.
Talk about ridiculous extremes.
Anomalous: inconsistent with or deviating from what is usual, normal, or expected
Anomalous: deviating from what is usual, normal, or expected
Canard: a false or unfounded repor
I'm currently on a Sun Microsystems E10,000 (e10k) course... studying Sun's answer to the Mainframe. It does basically all that the author likes, as well as being Unix, so you can do shell tools, etc, and be in the Unix environment we all like so much.
Okay, so it costs UKP1,000,000+, but so what?!! Solaris is (basically no cost), Linux is free.
The power's there, plus the flexibility. Also, if you know Linux, you know Solaris.
Author, Shell Scripting : Expert Re
Uh, yeah. I have. But here it is from the horses mouth: Robert W. Edwards, president of Risks Ltd., a risk-management consulting firm in Keedysville, Maryland, calls computer hardware theft "endemic." "There's a constant hemorrhaging from big businesses," he notes. "And it's not just PCs that [thieves are] stealing. They'll steal disk drives, even mainframes." That quote is from: This CFOnet.com article
No sig is worth reading.
Anyone ever hear of a 390 getting jacked? I didn't think so.
Read the whole article, or read it again. On S/390, Linux is a tool that runs in a partition under the native OS. The article isn't advocating ditching VM, he's talking about adding Linux to the S/390's toolset. Need a webserver, DNS, etc? Set up a partition, allocate DASD, copy a new Linux image over, set it up.
As the article states, $400 a seat for mainframe power and reliability is pretty cheap, and you're soon going to see IBM offering end-to-end Linux server solutions, featuring S/390, RS/6000s, PC Servers, and probably AS/400 in the near future (I know there's a third-party port in the works, but IBM will probably beat them to market). IBM shops will be able to come close to Write Once, Run Everywhere, and in the case of S/390, on a machine that almost never falls down and can handle an insane volume of I/O. Don't think that won't appeal to the bean-counters.
slashdot broke my sig
Where the mainframe excels is in its I/O channel. You can, for example, move gigabytes of data from one disk to another (or from disk to tape, or whatever) generating only one CPU interrupt. The channel is intelligent enough to do this kind of thing. That's why people tend to be surprised when they find out that the mainframe that just moved a few hundred million records around without breaking a sweat only has 64 MB of memory and a not-so-hot CPU.
Getting Linux onto the mainframe is a very important step, but it's only part of the picture. Facilities then need to be introduced to take advantage of the advanced I/O facilities that are available.
The Intel-PC world's attempt to imitate a mainframe's channel is Intelligent Input/Output (I2O). It does something similar, with intelligent peripherals designed to take the processing load off the main CPU. Mainframes have had this for decades. The Commodore 64 did, as well (I believe the article touched on this, actually). Now the PC world is finally catching up.
Alan Cox is working on getting I2O support worked into the Linux kernel. If the kernel interfaces for I2O are done with a sufficient level of abstraction, it is entirely possible that IBM could adapt them to use on an S/390 box as well.
--
Tired of FB/Google censorship? Visit UNCENSORED!
KwsNI dun said:
Actually, a Beowulf cluster probably wouldn't be the right tool for the job there (it'd be rather akin to using a butter-knife to tighten a flathead slotted screw--it'd work, but there are better tools for the job).
Beowulfs are very good if you need to do processing that can be done very well in parallel such as some formulas. They do not do so well if you a) have to do a lot of parallel work quickly (for that, a supercomputer like a Cray is probably better suited) or b) especially if you have to move and/or work with a massive amount of data (which is what mainframes are used for mostly and which is a job they do very well).
Besides, the entire purpose of Linux on mainframe boxen isn't to link a mess of S/390s into a Beowulf cluster (which, while it would make a very large box, it would be slower than a comparable Cray or even SGI Challenge series). The major purpose of porting Linux to a virtual machine S/390 is twofold--for "bare metal" runs it's basically an alternative to AIX (which, while reliable, is a horrid bastard child of *nix and the entire IBM REXX mess--for a fair amount of stuff which compiles "out of the box" on most *nixes, you have to add REXX scripts in AIX), while the port for "virtual machines" in S/390 (the one largely concentrated on in the article, by the way) is meant largely as a more user-friendly (at least to us folks used to *nixes and OS's invented in the last 25 years or so) alternative to the traditional IBM VM OS's (most notably MVS, VM/CMS, and VM/ESA--I've had experience with VM/CMS, by the way, and trust me when I say that *nix is far friendlier ;) that can also be used to add capabilities that have not existed for IBM/Amdahl "Big Iron" so far (like X-terms--other than VAXen and a few abortive NT ports, there isn't such an animal as a GUI for mainframes--it's all CLI terminals; also, stuff like PPP accounts can be set up (good for unis that may still have some old 3090 VM/CMS boxen about--IBM isn't officially supporting most OS's for the 3090s anymore, VM/CMS has a real dearth of Internet apps, and the daemons that DO exist do have some serious security flaws [most notably IBM VM SMTP--which can be anonymously relay-raped in its default install--and the default mail client which had serious problems in the early 80s with worms being transmitted]) and whatnot).
FWIW--there are a lot of places still using Big Iron, and not necessarily because they have a ton of legacy Fortran or COBOL (yuck) code that has been around since 1965. :) Insurance companies and banks, for example, commonly still use Big Iron because, well, Big Iron is about the only thing that won't choke on the massive amounts of data it must deal with reliably on a regular basis. (Supercomputers might be able to deal with it, but storage is iffy, supercomputers tend to be more temperamental than most Big Iron, and the amount of supercomputer needed tends to be quite a bit more than the average amount of Big Iron costs. Beowulfs are good for small- to medium-sized applications, but would barf on the amount of info needed and/or the nodes required (there is a limit to how many boxen may be linked in a Beowulf cluster; part of this is due to transmission speeds, but part of it is due to limitations of the Beowulf code itself); also, Beowulf clusters can fall down go boom if one node falls down goes boom.)
To give an example of, say, the average group that might use and NEED Big Iron--how about the US Census Bureau. They have to put in something around 250-270 million records in their databases every ten years; in addition to that, they have to keep databases (for comparison and updates, as well as to tabulate trends across decades) of anywhere from 100 million-250 million records--one for nearly every person in the country.
Yes, this is actually stored by computer--I happen to live nearish one of the four big processing centres for the Census Bureau in the US. They basically input everything on terminals from the census forms, thousands of people do...which are ultimately stored on Big Iron and on media totalling anywhere from terabytes to possibly even exabytes of data.
Damn near everything SHORT of a Really Huge Big Iron system is going to choke, hard, on this amount of info. (Especially so when you consider that a fair amount of older data is probably stored on tape or removable big-platter hard disks--usually using legacy storage systems which may not even be commercially sold anymore--which are being converted to more modern storage media capable of handling terabytes of data at a time.) I like Linux as much as anyone (and freely admit to being a Slackware and SuSE bigot ;) but a Beowulf cluster just isn't going to handle that. Not even if you built it from Alphas. Not even if you built it from bloody Playstation 2s. :)
You mentioned the IRS--well, they've got comparable storage and data handling requirements as the Census Bureau does, only worse. :) They have to maintain upwards of 100 million records which must be entered and updated yearly (just from people who fill out tax returns)...plus records from W2 forms filled out by employers of the folks associated with those 100 records...plus trends must be done on ALL these records, and previously archived records which may stretch back as far as 30-35 years or so (again, often on legacy systems and data--good old removable disk packs and 9-track tapes) in order to flag folks (who might be remiss on paying their taxes) for audits...we're talking literally hundreds of terabytes or MORE of data that the IRS must go through on a yearly basis! It's actually kind of impressive that their systems don't barf more than they do when one thinks of the sheer amounts of data they go through...
-Windigo The Feral (NYAR!)
A mainframe is NOT a thing of the past !!! What do
...
you think manages your bank account, possibly your
salary, certainly your IRS account and probably
your pension fund ??? Mainframes, that's what.
Most of the concepts that we enjoy in Linux: high
reliability, scalability, etc..., have been present on mainframes for **ages**. I am not
flaming, but you have to give each one its due
What really sets these machines apart (in my personal experience since more than 25 years)
are two things: reliability, and high throughput.
These machines were designed from the ground up
to serve **hundreds** (sometimes thousands)
of simultaneous **active** users. You have the
hardware facilities to serve literally thousands
of concurrent I/Os (this explains the price,
of course).
Now about linux: I think that's a good move on
IBM's part (who sponsored the port): as I see it,
their long-range view would be: one OS (and one
set of applications tailored to this OS) on every
possible hardware platform. Then application ports
would truly be re-compilations only, which profits
IBM too (as software developer, and also in the
support department)...
If D. Miller could port Linux to high-end Sun machines with 15 processors, why not on IBM mainframe ?
Although I've read articles in the past that have noted IBM running linux on the 370, I liked this more detailed article.
It should be noted that a Mainframe has huge ammounts of I/O bandwidth. I've seen five year old mainframes put new as/400's to shame. By all rights the 400 has more processing power, but the mainframe has such huge ammounts of bandwidth that it can move the information around quicker. When you think of a web server you realize that the actual program isn't all that complicated. Most of the CPU time is dedicated to feeding I/O out to the TCP stack. A mainframe may not be as powerful as the as/400 or a new Xeon, but it has a lot of smaller processors, and they aren't wasting cycles waiting for I/O to give them something to crunch.
Also you could have multiple high speed printers I really can't believe that you need a mainframe to increase something that is on the device end like a printer.
And here is where you betray your ignorance of things mainframe. Mainframe devices are much faster than similir devices in the PC world. A mainframe printer connects like a network device does and utilizes huge transfer rates. Mainframes are built for IO. IO from tapes, to tapes, to printers. Mainframes exist for IO.
Beowulf clusters are built for parallel processes -- think solving intractable differential equations through simulation like weather models where the state of point A influences its neighbors.
Billing and accounting systems don't have the same kind of dependance between data points. It's just that there is a massive number of them.
And they need to be processed quickly. Some of the files I work with every day have 7 million records with 2K of data per record. That 14GB just for that dataset and we get a new dataset every month. And that's just one file. I use about 10 different files. My company has literally tens of thousands of data files on tape cartridges in the mainframe data center. I can process *any* of them on request in a matter of minutes. I could process those 7 million records of data in a half hour if I had the machine all to myself. That's about 9 MBytes/sec of sustained throughput *with* calculations on every one of 7 million records in that time.
Anomalous: inconsistent with or deviating from what is usual, normal, or expected
Anomalous: deviating from what is usual, normal, or expected
Canard: a false or unfounded repor
Unable to connect to the database. Please email.
Looks like a mainframe should be standard equipment for any site mentioned on slashdot.
Next thing you know, they'll be porting it to those old DEC boxes that all the universities have laying around.
Sure, the dozen PIII's will match the Big Iron in MIPS/FLOPS, but it would take a hundred times as many to match the sheer I/O bandwidth of those monsters. An old, low-end IBM 9x2 will handle 4,000 GB of I/O per second and love it. A high end PC will perhaps handle two, and totally thrash. Assume I'm wrong, and a PC could push 10 GB. You'd still need 400 computational nodes + 40 managerial nodes + 2 controller nodes == 442 PC's to match the performance of ONE old mainframe.
Used 9x2, $20,000.
442 PIII@$1800, $795,600. Which is cost effective?
When you're doing simple processing of huge data sets, like bank account updates or IRS 10-40 return valitation, it isn't adding up numbers that bogs. It's the contunual process of [retreive x][save x][print x]. Beowulf clusters have their place, and not as replacements for mainframes.
.sig: Now legally binding!
A PC vs Big Iron Holy War was definitely not what I was expecting with this article. Having worked with computers in various industries, I can tell you one thing, all of this stuff has its place. I was at a large loan servicing shop (like over a million loans) and I can tell you - you cannot configure any PC hardware to do what their MID-range box did. Over a 2 year period, not ONE server bounce, all the while hundreds of users logged in constantly, running massive batch jobs every night, printing out over ONE MILLION statements at a time etc. etc. etc. I am certainly not knocking the PC, I love 'em. But I do like the positive tone of this article, IMHO Big-Iron stuff is way way misunderstood by folks who haven't any experience with them. Heck, even I scoffed at the things until I got a chance to work with them. (PS I know my website's down - gimme a break, I have a full time job, ya know?)
mas cerveza, por favor politically incorrect stu
I think you are glossing over years of interesting history and computer architecture. IBM's mainframe division might have lost mililions because of PCs, but not because PCs were superior computing platforms for big data.
PCs won for the same reason that we have traffic problems on our highways--everybody wants to be a driver and doesn't want to share. This was, it seems, part of the rationale for the federally funded development of the internet (well, ARPANET): getting scientists to share computing resources bought with federal money (c.f. A History of Modern Computing, Ceruzzi 1998, p 296). This is of course the same phenomenon that drove minicomputers, which were also replaced by PCs. I'm not saying this is bad, well, not in the case of computing anyway.
But PCs still suck at any number of computing tasks, and aren't really improving in areas that can't be mass marketed. That's why my lab bought a very expensive dual Alpha machine instead of spending that money on the 5-to-20 equivalently clocked P-IIIs (these numbers come from real computations). Not to mention a farm of PCs can only handle embarassingly parallel computations at the same speed has the Alpha, and require more programming effort than the Alpha. The Alpha isn't even close to a mainframe, either.
And I haven't even gotten to bandwidth issues that sponsered this thread (well, they're part of the 5-to-20 figure above, in some ways). IBM lost on mainframes because they dealt _only_ with mainframes. DEC lost with minicomputers because they too were arrogant/ignorant about PCs. And while Intel seems to acknowledge the information appliance ideas, they're x86 tech will only go so far (we can hope, can't we?). But just as information appliances aren't the best choice for PC-type tasks, PCs aren't aren't the best choice for mainframe-sized loads.
Oversimplification is a marketing tool. It has no place in intelligent discussion, where flippant remarks are better replaced by _questions_.
Some anonymous coward dun said:
*chuckle* Methinks someone doesn't quite get the point with the mention of scalability...
1) Big Freakin' Deal that you can run Office 97 under Win98. For the applications we're talking about here (mainframe stuff--numbercrunching and storing) "pretty" stuff like Office 97 or GUIs in general are neither helpful nor necessary (in fact, they'd be a detriment to the Job that the Big Iron is doing).
1a) I would far from call any Microsoft product "reliable". Yes, this includes Office, Win95/98/NT/2000/3.1/CE/Me/[insert latest marketing spin from Micro$oft here], and IE. Yes, I know of what I speak here--I've had to do more than one repair job when supposedly well-configured Microsoft apps and OS's suddenly developed severe cases of incontinence. :P Compared with some of the stuff I have to put up with re Microsoft stuff (hint: OS's are not supposed to corrupt their essential files over time, nor are they supposed to lock when running programs [necessitating a hard reboot and scandisk], nor are they supposed to crap themselves after 49 days of uptime because even Microsoft acknowledges that neither Win95 nor WinNT are stable enough to stay up longer than that, thus a 49-day reboot is coded in), even beta builds of Linux are marvels of stability :)
1b) Please call me when a version of Windows is widely available for Really Big Iron, such as is used for databases for insurance companies and the US Census Bureau. ;) (AFAIK, they don't exist--not even WinNT ports (the largest iron WinNT was ever ported to, BTW, were Sun and Alpha ports--and those two ports are supposedly being discontinued). Most of 'em don't use *nixes, either--they use stuff you've probably never heard of like MVS, VM/CMS, VM/ESA, etc.) 2) The point wasn't on "who had more apps" or "who was prettier". It was "Who can run the base OS on more stuff"...which Linux beats Microsoft, hands down. (Itsys are teeny even compared to WinNT boxen, and with the recent ports to run as virtual machines under mainframes (not to mention the Linux/VAX project, the Linux/3090 project, etc.) Linux has probably just surpassed NetBSD as the OS which can run under the maximum number of architectures.) It's rather a different cock-fight than the usual comparisons, mind. 2a) I'm not sure that the virtual-machine versions of Linux are quite ready for prime-time (at least for what mainframes tend to be used for), but at least the option IS available should one want to run Linux as a shell (as opposed to a traditional mainframe virtual-machine OS like VM/ESA or MVS). Compared to the OS's that do tend to be used with mainframes, Linux is a fair sight more user-friendly; more people nowadays are familiar with *nixes in general (if from nothing else but student email accounts or computer science courses) than most mainframe OS's. Also--and this may be a shock to you to hear this--using Linux as a virtual engine actually would make it easier for users to set up stuff like Internet accounts--including PPP services for folks who want to use Windows from home. ;)
(As a minor data point to add to that--the University of Louisville recently retired its old 3090 (which had been formerly used in EMCS and IS courses, then [when email first started becoming widely available and the EMCS and IS departments had largely gone to either PCs, an RS/9000, or a combination of SGI, HP, and DEC Alpha boxen] was used as the primary Internet account server for the Arts and Sciences school) in exchange for a DEC Alpha box. This was done for many reasons, partly because PPP is easier to set up on the Alphas and partly because IBM no longer officially supports VM/CMS on the 3090s [which was a Bad Thing, especially since they also no longer accepted security patches for Internet utils and daemons; at the time, there were two rather serious security bugs for IBM VM SMTP that were being widely abused, and I spent much of a summer giving the two unofficial patches to universities who'd been relay-raped by mailbombers :P]. If a virtual-machine version of Linux had been available for the 3090, it's possible they could have kept it in service a while longer instead of selling the thing off for scrap metal. :P)
3) You talk of things being "ready for the desktop"--most mainframes aren't because they have no real need to be. Realistically, the most useful setup for a Linux VM on a mainframe would be either for Internet-related network services (sorry, but Linux does have better support there anymore--at least sendmail and qmail do have protection against relay-raping and are regularly fixed to close any security holes found) or for a shell alternative for folks who are already used to working on *nixes at a shell prompt (instead of them having to learn the command sets for Yet Another OS). It's fairly obvious that you've never done much work with a mainframe--otherwise, you'd realise that there is no freaking desktop...these are Big Machines, things that fill up entire rooms complete with false floors to hide the miles of cable and Halon extinguisher systems. You aren't going to get right at a terminal, and you probably aren't even going to use an X-term with these beasts (unless the Linux VM running has it set up to do so); if you access these things directly at all instead of sending stuff back and forth across a network with the mainframe being basically a virtual disk, you're going to do it the old-fashioned, CLI, type-in-the-commands-on-a-TN3270 way.
Needless to say, unless and until some kind of Windows port makes it to such Big Iron, whether or not it's "ready for the desktop" is completely and utterly moot! Unless a Linux VM is installed and set up to use X-terms, you aren't going to get a pretty interface--the closest the OS the VM is running and your Windows box are going to get is with your Windows box running a terminal emulator like TeraTerm or VT3270. It's going to be done by text, the way Big Iron has always done it since we got away from programming boxen by switching plugs and relays and valves (vacuum tubes for us Yanks) around and went to punchcards and old Teletype terminals instead, before the nutty folks down at Xerox PARC came up with the idea of GUIs in the first place.
(And before you ask--yes, I know what I speak of here, too. I was at U of L back in the days when the 3090 was actively being used to teach Fortran, and also when it was used as the Arts and Sciences Internet server--U of L actually had set up a mess of old VT100 terminals because, other than through a terminal program, that was the only way the students could read their mail! Graphical interfaces for VM/CMS and most other mainframe OS's that run in virtual machines plain don't exist; Internet apps that most folks take for granted (un-relay-rapable mail servers, such things as even text-based WWW clients, etc.) had to be found or just weren't available, and tended to be years behind their *nix and/or Windoze equivalents. Needless to say, life got easier for the A&S students when they retired the old 3090 and got the nice Alpha server running OSF/1 ;) I wasn't QUITE there in the days of punchcards, but apparently they were still being used as late as the early 80's there--that's how long the 3090 was around--and the things were never really designed for anything much besides big databases and number-crunching and maybe BITNET connections. Most of the OS's for Big Iron actually date back to the days before CRTs became widely available, especially the Big Iron using virtual machine OS's. In fact, the ONLY Big Iron I know of at ALL that uses anything close to a GUI are a) the Alpha and Sun ports of Windows NT and b) a terminal and configuration program for OS/2 designed to act as a console for booting AS/400 boxen running OS/400 (in other words, the OS/2 program largely replaces the blinkenlights). There's no need for being "desktop pretty" if all you're doing with the thing is using it as a big-arse server (which most mainframes are)--most of the time you might not even be doing direct interaction with it anyway, and if you have to a CLI works wonderfully. ;)
-Windigo The Feral (NYAR!)
The whole point of modern mainframes can be summed up in one word: VOLUME. Regardless of the ground that has been covered by intel and co., the different OS developers, and the various efforts to develop high capacity I/O interfaces, your standard PC platform just doesn't have the ability to handle the sheer volume of data that your average mainframe deals with. PC's have a long way to go before they can even begin to encroach on the mainframe realm of computing.
The whole point of modern mainframes can be summed up in one word: VOLUME.
Sorry, but that's the wrong word. Volume is necessary, but you can get that with either big machines or clusters of little ones.
The word you want is RELIABILITY.
And by reliability I don't mean just uptime (although that's a piece of it). I mean the machine does not drop bits. Period. Even though the PIECES of it are dropping bits all over the place. (When you have square feet of silicon intercepting cosmic ray secondaries and rattled by thermal vibration it's unaviodable.)
I know of at least one mainframe multi-CPU unix clone (UTS) which has sites with uptimes measured in years. In fact the last time I heard there were software patches that had been enqueued to be loaded the next time it went down, which have been waiting for years as well.
The CPUS are automatically switched out when they fail and manually switched back in once they're fixed. The show goes on. And the processes that were running on the cpu as it failed still do their computation correctly - because the broken bits were caught and fixed as the CPU/memory/whatever hiccupped.
Many of the people who are putting together clusters of machines of lower reliability - including those in the management of at least one mainframe company - haven't grokked that concept.
The more computations you do, the more likely you are to be hit with an error. If your process is mission critical you can use hardware that catches AND FIXES the error, or you can try to write software that detects and recovers.
The software solution is the MUCH harder problem. The hardware fix - which is the mainframe solution - is expensive. But when you're dealing with millions of bucks per hour of downtime, or perhaps per dropped bit (as phone companies, brokerages, banks, and the like are), you can afford it. Mainframes (less peripherals), redundancy and all, have been under a megabuck a pop for some years.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Well, it just isn't so. Granted that clusters and desktops may even equal a mainframe's raw processing power, but a mainframe's real strength lies in its massive I/O capacity... you can connect huge arrays (even RAID arrays) of high-speed disk drives and have them work near rated transfer speed. No desktop comes even close to that.
Over 25 years ago I worked on a Burroughs B6700 mainframe. We had full source (at zero cost) to the OS, compilers, utilities and so forth, and had a great time mucking around with these, fine-tuning stuff, fixing bugs and even implementing new Algol constructs. For some weird reason Burroughs rarely used our fixes, though :-). BTW we also had full hardware schematics - not that these were of much use...
Wrongo! Mainframe = "Serious Business Machine". Nothing will get an MIS director's attention like saying "See, Linux can run both on your PC and on that million dollar IBM mainframe that runs your core business. Windows can't."
Remember, what caused the PC to catch on was not the "home user", but the business world. The PC caught on because MIS directors saw it as a "Serious Business Machine" in contrast to competitors that were "game machines". To these guys, Microsoft is the "johnny-come-lately" that they are not all that comfortable with. For them, IBM is still king. Saying that Linux will run on their big iron while Windows won't says to that that Linux is a serious operating system while Windows is a toy.