I've contemplated doing something like this for a while. Then a new feature came along to make it more challenging:
"Bookmark this group of tabs"
I love this feature. It's how I lay out my morning reading. One click, and then it's ctrl-W until done. All my comics, all the blogs, and a weather report.
Adding to the last point you mention, the most critical thing you do to any plan is TEST IT.
Stage a disaster.
Either that, or just fake it. Take the data to and try to bring up all the data, bring it up on the internet (under a dr.www.?????.com DNS name for example) and see what's accessable.
In case of failure, tune the plan, and try again in 6 months.
In case of success, tune the plan, and try again in a year.
Ritually disemboweling a computer on the network does certainly seem to keep the rest of the network in line for a while.
{wavy imagination lines}
Yes, I'm a computer therapist.
Thank you for coming doctor. Our computers have been cranky ever since we 'realigned' our sysadmin (he didn't SEEM to be doing anything useful). Downtime is on the rise, Our databases return 'luser' to one querry in three, and our CIO's Office Assistant's computer only prints swear words!
Ok. I think I know what the problem is. Do you have a fire ax?
A FIRE AX!!!
Yes. Ahh. I believe I saw one on the wall outside. Follow me please.
{obtains ax}
Now, could you lead me to your datacenter?
uh... ok...
{finds a development box, and repeatedly evicerates it with said ax.}
WHAT ARE YOU DOING?!?!?!!!
I just bought you a few days grace. Go back and hire your Sysadmin again. The boxes will be happy you did. Until then, I've scared them into submission.
If you're backing up serious data, tape is the only viable solution.
The technology is old. Or rather, mature. Most, if not all of the bugs have been worked out years ago. At work we have some LTO drives. IIRC, they can be written to at 20 mbytes/second, and they hold 60 gbytes uncompressed (and 120 using the hardware compression. Granted, this assumes the data isn't compressed already. Realisticly, you can only get about 80-90 GB per tape.). You can engineer tapes to be increadibly resiliant to almost anything. The only moving part in DLT or LTO tapes is a spindle that can be turned by the drive. The spool is unwound into the drive itself, and rewound back into the tape cartridge. There are also 2 spool designs with the 'read area' in the middle like cassette tapes. There are also tapes in the 100/200gb range, and the speed and size keeps increasing linearly.
What other medium's are out there?
1) CDROM/DVDROM Slow to write, and not much data. Good shelf life (20-30 years) though. Usually CD's are used for 'archival' purposes. AKA "The IRS decrees that this data must be kept for the next 15 years."
2) Flash? it's solid state, and no moving parts, but the write speed SUCKS for real data sizes. Also, the density just isn't there. IIRC flash cards top out at 512 mb now.
3) Hard disk? WAY to fragile. We moved a batch of 20 or so servers with about 2 TB of disk this weekend. We lost 7 disks (of around 150).
At this point, untill someone comes up with a remarkably new idea, tape will be the king of long term, high density data storage for the forseable future.
It's a big troll, sure. However it is also a chance to dispense some good advice:
1) There's a difference between PC's and 'Server Class' hardware. The biggest is testing. It will work, and it can be supported easily. Drivers are nice and available (generally speaking). Usually dual proc, usually RAID enabled. You can use RAID to speed up read access, but almost no one does. They use it for redundancy (in case a disk flakes out on you). How much money you spend depends on how much downtime costs. If it really costs, you need RAID 1+0 or 0+1. Go with hardware based RAID if you can.
2) Sun hardware. There are many more advantages to sun hardware than what's obvious. Never over look what a good support organization can do. You pay for it, but if something fried, I can have a part in my hands 4 hours later. Sun's low end desktop's are nothing to write home about. However, if you've got Ultra60's or SunBlade 1000 or 2000's, that's some really class hardware. You can do some supprising things with it. [1]
3) Dual procs. On Desktops, even if your simulations don't benifit from dual proc, if they take a while, and they eat that 1 CPU, you'll be happy to have a second (web browsing, etc). On your servers, it's effectivly a must.
4) RAM. On the servers, crank it. On the desktops, you should probably crank it.
5) Cost. If your work is anything like mine, you have 'capital' money, and 'O and M' money. When in doubt, over spec the machines, so that your less likely to have to request more money from the 'capital' pool than you initially quoted. "Going back to the well" viewed poorly.
6) NIS. NIS is evil and the plague. If your in a relativly local office with good connectivity, it's alright. If you try to spread it over WAN links, you're going to get hurt at some point.
7) NTP? Why run a seperate server when you don't have too. Leverage what's already in use in the company. This leads to my last point (and what was the best point of the parent)
8) Get yourself a real sysadmin. These are decisions that s/he is experienced in, and paid to do. Your trial by fire that would come from this will probably drive you insane. Good sysadmins are a rareish breed. I know, I am one. There are a fair number of good ones out of work now. Find one.
[1] The reason largely has to do with cache. Sun chips made in the last 2ish years have 8 MEGS of cache on them (that's even mirrored so it's 16 in total, but you can only use 8). We built a GIS app, and field tested it on Sun and Intel hardware. The intel hardware could deal with 1 to 4 users with less resources than the sun box could. However, the the sun box kept growing up to several hundered users, while the intel box started thrashing hard after 10 or so. We compared a dual US3 box to a dual Xeon P4.
The article is up and loaded swiftly for me. That said, I was hoping for some more 'meat and potatos'. Short version:
1) I can't get my special app (iNews) to work. I need it for my work, so I'm kinda screwed.
2) Web browsing sucks (because IE is a hog). Safari is in beta, but getting better. He didn't mention Moz or Chimera (or whatever they call it this week).
3) It's very nice to work with. If you don't NEED a piece of software that is windows only, you'll love it.
I recently did some pricing (each with 1 gig ram).
Dual 1.25ghz power mac: $2400 Build your own dual Athlon MP: $1100 Build your own dual Xeon: $1700 (iirc)
I know it's not fair, but that's only because I can't build my own power book. (buy a dual Xeon, and you're in the $2-3000 range too.)
I'd love to have a (reasonably powerful) apple on my desk. I just can't justify the price difference.
www.pricewatch.com shows 181gb scsi drives. There are also 4 x 181gb drives 1 cube away from me for our EMC.
Maxtor may make a 250 gb drive, but you can't use it in your PC. IIRC, ATA133 can only address up to 120gb.
Also, when you're buying scsi, you're not going for single drive density. You're aiming for throwing 10 drives into a RAID 1+0 config (or similar). And finally, those 250gb are new. They're not going to release it for SCSI until they have some experience with it's failure rates, and what not.
Like other's said, it actually lives in a flash ROM chip. Think of it like the BIOS and GRUB combined. It groks networking, all of your disks (tape, cdrom, harddrive, etc), and lots of other things.
You can diagnose hardware errors with it, you can boot off the network with it, you can specify which device you want to boot from, (on multi proc systems, you can specify which CPU will be the 'boot strap' cpu), and you can tell it if you want it to boot at all (if you want it to come up to just the eeprom after a power cycle.)
It's all sorts of goodness. I can't tell you how many times it's saved the day on our sun boxen. And for the hyper masochistic, there's a full blown forth interpreter.
And get this: It doesn't suck. (though it does generally mean something is sucking wind when you are working at that level).
Sun's boot proms have a fully functional forth interpreter in them. Thankfully it's rarely needed these days, but it's there for when it really hits the fan.
Heck, it's even possible to 'mount' a file system, and use a line editor to fix things (granted, it's easier to boot single user, and go from there, but again, if things are really FUBAR'ed...)
NFA's (and Deterministic FA's. They're mathmatically the same thing) are the math backend that define regular expressions.
They're state machines. They're in a given state, and they know how to go to the adjacent states. So given the string 'abc', if you're currently looking at the 'b' (having already seen the 'a'), you know that you'll have a valid match iff the next character is a 'c'. If it's not, you have no match. if you have 'ab[cd]', and are looking at the 'b', you know you have a match if the next char is a 'c' or a 'd', you've got a match. 'c' and 'd' then are the 2 next valid states.
The nifty thing (and the limiting thing) is that true RE's require no memory. Just the knowledge of what state they're currently in. For this reason, no true RE can be written to see if a given string is a palendrome (you can write a RE to match a specific palendrome, but not an arbitrary one).
The difference between a NFA and DFA is that NFA's allow 'null transitions'. This basically says that there are more than 1 state that you can leap out of when you see the next character, because you can go to these special adjacent states without seeing a character, and then leap out. There's also a proof out there that any NFA can be written as a DFA.
All of that said, Perl's extended RE's are not true DFA's. They have some features that can not fit into the DFA model. This is one of Larry's reasons for wanting to make Perl's RE's into true CFG's (context free gramers).
This model is much more powerful than RE's, but at a greater cost, since you have to have memory too. The mathmatical definition of a CFG is a state machine that drags around a stack of memory. The state machine may at arbitrary times push data onto the stack, and later pop it off. It must be done in order though (to match the math model. If you add a second stack, you have the definition of a 'turing machine' (aka the computers on our desk)).
A CFG can be written to match arbitrary palendrome's for example (just push each letter onto the stack, and when you hit the middle, start poping off, and matching each letter. Yes, this is over simplified. The true algorithm is left as an exercise for the reader)
As for configuring networking gear and rackmount equipment, that, too, will slowly come around and start using new technology for their control interfaces. A time will come, in the not so distant future, where it will become financially unfeasable to implement legacy technology on new devices, because the support will be dissapearing, and the devices won't be as functional in the years to come.
I wish.
Sun makes pretty nice hardware right? They have this spiffy box called an E10k, or starfire. Can have 64 CPU's in 1 domain (hmmm... make -j 64), or 16 different, 4 proc servers, or any other combination you can dream of (multiples of 4 CPU's per domain).
How does the SSP talk to, configure, and control all this power? Through something called a JTAG interface. It's been around since the early 1970's. It works at 300 baud. You get your console, and all the configuration information on the box via 300 baud, SHARED accross the entire platform. [1]
Why the heck did they do this? Because it is well tested, and works. The API's are perfectly understood by everyone, the firmware's been around for ages. They could put it in there, and know that it will work. period.
I wish that you were right. But companies will continue doing things like the above.
[1] To be slightly fair, the box was first designed and introduced in '97 or '98, though when upgraded, it still makes a kicking oracle server with rock solid stability
MQSeries is a pretty useful application. It gives you a unified method of talking from anything to anything.
Backend CIS system on a mainframe, and a java based linux webserver need to talk? No problem.
IIS/MSSQL box need to pass data back to DB2 running on an AS/400? No problem.
Sun/Oracle box talking to a custom homegrown app? No problem.
It's a unified API, and a unified 'paradigm' of communication. It can talk TCP, SNA, and every other network type out there. It runs on every os/hardware platform you're likely to see in production. And it offer's a promise of message delivery. It won't just try and give up. It guarantees (sp) that the message will get to where it's destined to go.
The downside is that it's kinda heavy weight. It's also a little network chatty, and it's also rather expensive. However it does make it easier for your developers. If they all know how to get from their application to a Message Queue, then all of your developers can get data back and forth between ANY of your applications.
All of that said, there is some talk in the jabber community about trying to move into the 'middleware' space. That'd be a really interesting thing to watch. Doing that guaranteed delivery isn't something to be taken lightly, and if they could crack the portability problem (how many of us have as/400's to play with?) it'd be pretty cool.
Out of curiosity, why the heck are you using AIX+Oracle? It may be advertised to work, but IBM and Oracle are huge cometitors in the enterprise DB space (DB2 for IBM), so I'm not really supprised your having stability problems.
Oracle's 'reference platform's right now are Sun and/or Linux. If you must stay with oracle, you might want to switch. If you must stay with AIX, you might want to go db2 (though the first will be far easier... oracle export -> oracle import and you're done)
Outside of the integrated amp weighing a ton, there's another reason. If you listen to the 'audiophiles' out there, one of the things that they rave about is 'build quality'. 99% of the time, what they mean by that is 'weight'. "Feels cheeply made" is a complaint about a lot of 'low-fi' and 'mid-fi' gear (low and mid-fi gear is defined as equipment one or more steps down (usually in price) from the gear you currently own (or the gear you're currently lusting for)).
There do exist good, quality, peices of equipment that produce music wonderfully. Trust your ears over your other senses (unless you're looking at video gear. At which point your eyes should only see 2 things: the picture and the price. People can be conned into wasting a lot of green on video gear.)
You must have backup. If you don't have someone else to do this, and you have a shop buging you extreemly regularly you need to get them to hire a partner.
Phrase it something like this: I am the only person who understands how these systems run. I'm the only person who can fix them when this breaks. If I get hit by a bus, you're in a world of hurt. If I don't start getting more sleep, I'm going to burnout, and you're in a world of hurt. I work 70 hours a week regularly, and the backlog keeps piling up. I need someone to help me.
Then (well, after 2-3 months training the person), you can take weeks off from pager duty. There's a reason doctors have on call rotations. You should too.
One of the things that I've done is set expectations at my place of work. 90% of the time, my pager is available to them when they need me. That other 10% is well communicated in advance, and my boss knows that my pager will be on my bedstand, but I'll be in another state.
(Oh, and reguarding the person who saw 'recurring nights of database server issues' as a sign of stupidity, they might be right, they might be wrong. We've had a sun e4800 go really flakely on us recently. It took WEEKS of long nights (since the box was production, and we couldn't take it down in the day) to get the hardware on that box stable (it would work fine for 3 days or so, then flake out hard). Sun wouldn't give us a new box (with at least an understandable reason), and keep insisting on replacing individual pieces. And it certainly wasn't our doing. It turned out to be a bug deep in the IO chassie's firmware.)
You (and the others following this thread) owe it to yourselves to give postfix a try. It is clean, and elegant. All of it's configs live in 2 files (main.cf and master.cf), and it is powerful.
Postfix has solid anti-spam measures, fantastic performance, readable config files (that are well commented from the start... heck there are about 10-20 comment lines per config line), and you only need to set like 3 lines in the config to get a working mail server.
People have some good points, but one thing that package managers need to understand is rollback. ie: I had gnome 1.4 installed. Worked well. I upgrade to gnome 2.1. 2.1 might just turn out to not do what I need at the moment, or it might be incompatable in an important way. I should be able to 'roll back' the system to 1.4 without performing major surgery.
Yes, this will take more disk space (to hold the old versions of all the files, and the metadata to restore them). But in an enterprise environment, you need it. Sun has been doing this for years with their OS patches, and we should be able to steal it for packages too.
That's one of the most insightful comments I've read on/.
Your point about binary compatability on linux is almost painfully valid at times.[1] It'd get a LOT easier if glibc/gcc would finally decide to stop breaking backwards compatability.
There is a work around (most of the time). You could distribute a statically compiled binary. It's not pretty, and it's not perfect, but it will solve a lot of the problems.
[1] Ever had a debian firewall on a pentium/60 with no gcc, and redhat behind that, and no other available linux boxen to compile things on? Updating the kernel on that firewall was a royal pain.
About 99%. And we seem to have gotten a reputation as being rather 'harsh' as a Grand Jury on our attorneys.
This fact has lead to a wide spread belief that grand juries are a rubber stamp on the process. However, you need to realize that about 95% of the cases we see are truly open and shut: "He has confessed that he did this.", or "He was deported, he's back, he's not allowed back, his finger prints match."
As you might guess, we were quite happy when we saw those cases that fell into the 5%. Some were done by amazingly stupid people, others were wide spread rings of very inteligent people who just got too greedy.
Finally, remember that we have a very light 'burden'. Proving that someone probably did something, when you only hear the evidence of the prosicution (sp) is pretty easy.
How would I improve the Jury system? I'd make it a lot closer to my federal grand[1] jury experience.
They call 45 ish people. They need 23. The list of valid excuses is pretty lean. When we got there, there were 26 of us. They put our names in a bucket, turned it, and pulled three names. The judge said "You three can go." Then swore the rest of us in as grand jurors.
This would have to be modified slightly for a trial (or petite) jury. You'd have to give the lawyers SOME strikes for obvious bias, and start with a larger pool for the OJ Simpson cases of the world. Just make the number of strikes proportional to the pool. Once you have 24 'good men and true' (16 + 8 alternates), start the trial.
[1] Note: Grand jury's decide if a case can go to trial. A trial jury (petite jury) is at the trial, with the judge, etc. At the federal level (at least in Atlanta, Ga), they are set for a year term, with 6 months after that for 'recall' status. There are 4 juries at any given time, one for the first 4 tuesdays in a given month. We show up between 1 and 4 days a month (usually 1 or 2. The last three day was 4 years ago, and no one remembers a 5 day docket), and see around 5-8 cases/investigations a day. Our 'burden' is much lighter than a trial jury. We only have to see if the accused had 'probable cause' in this case (aka: did they probably do the deed). We also only hear the prosecution's side of the case.
We also hear investigations, basically a way that an attourney can question a potentially hostile witness under oath. These can go on for a while (which leads to the 'recall' status). We'd come back for minimal time to hear more evidence, or to see if an inditement is asked for from the AG.
It was a fasinating experience. I saw a LOT of stuff I wouldn't have seen otherwise, but it was quite boring at times, and my boss wasn't exactly amused (nothing they could do about it though... The constitution has a minor trump card in this matter. If they even sniff that you're let go because of it, the feds will come down like a ton of bricks).
As an asside, 'private company' generally means a company like google, owned by a small group of indivduals. Verio is a public company. It is owned by it's shareholders.
Your example from your pre-sales day is a totally different world. Verio had already decided to do business with thing.net, and had written a contract to do so. You and this 'someone' had NOT started doing business. You're totally free to choose who to do business with. However, once you start, it depends on the contract.
I very strongly believe that if you say (like in a contract) that you're going to do something, you need to follow through with it.
As for your 'get an account with another service provider', sure that's possible. However, if BillG was really after you, and if all the service providors followed your logic, then you would be silenced, in the medium of the internet.
As for the reference to the middle ages, it's something called an analogy. A story used to illustrate a point. Exageration? Maybe. However the nobles DID work quite hard to destroy opposition. Why do you think the French Revolution was so keen to have a free press? The nobles silenced many a newspaper. Sure, they could have gone back to writing the things by hand (the equivelent of you're suggestion that they should change medium), but that would have been much less effecient.
Verio [probably] said to themselves, "On the one hand we have a bunch of activists who pay their bills, but who aren't a significant source of revenue for us. On the other hand we have Dow Chemical, a gigantic multinational corporation that could throw us a lot of money if we have a good relationship with them." And they made a business decision.
If that's what happened, I really don't see a reason to get all up-in-arms. Yes, this is an inconvenience for the activists. But, if it happened the way I'm guessing, nobody did anything illegal, or even unethical.
I won't go into the rest of your post, which I thought was well and good. However, while I see the logic of your 'wild assed guess', I disagree with your statement that it's 'not unethical'.
Say you have a mindspring.com account. On your personal web page, you post that 'bill gates of borg' icon that he's reputed to hate so much. He causes a letter to be sent to mindspring, and your account is revoked, using the same logic you just called 'a business decision'.
This is precisely the kind of thing that power leads to, if it's not curbed. Look at the nobles of the middle ages. Not held accountable, they could kill anyone they wished without repurcussions. This is just a modern day version of that.
We must protect fredom of speech (obviously keeping in mind the limits WRT slander and libel). We must we must protect those liberties.
Now, if thing.net violated libel and slander laws, they get what is coming to them. If they didn't, then what happened is unethical, and should be illegal.
I can't echo this enough. It's great multiplayer, but not LAN play. You can have a bunch of people playing at one computer though.
It rocks.
Re:All about the benji's
on
Sun vs. OpenBSD?
·
· Score: 4, Interesting
From the article: University of Alberta's Bob Beck said he is forced to buy out-of-date UltraSparc II-based E450 servers instead of newer UltraSparc III-based V880 machines for the university's SunSITE software exchange.
This seems odd to me: 1) OpenBSD doesn't support SMP yet, right? 2) v880's must have multiples of 2 CPU's (up to 8).
Sunsite might be better off grabing some of those 1U v120's, throwing a dual channal diff scsi card in there, and using an a1000 array (or maybe a t3 array... with only 1 cpu you probably need the hardware raid these offer rather than the d1000's or a5200's). More disk, less rack space, less power.
Now, the v880's rock. Great price point, 8 cpu's, 2 FC-AL planes for a total of 12x73 gig disks, 10 PCI slots (2 x 64bit/66MHz), onboard gigabit fiber... the list goes on. It's a great box (for more details, hit up store.sun.com, select servers, find 'low end servers', and select the v880. And note that that's 'list price'. You can get up to a third off of it from most resellers)
For reference: 4x itanium 800MHz dell 7150: 8x73 gig disks is $61,113.00. 4x usIII 900 MHz sun v880, 6x 73 gig fcal disks is $59,995.00
(That's the closest 'apples to apples' match I could make. I chose itanium vs usIII because they're both true 64 bit chips. Though the expansion of the Dell isn't as nice... the sun can add 4 more proc's and 6 more disks. The dell can add more memory... 32gig tops the sun v880, and 64 gig the dell)
Slashdot Mirror
I've contemplated doing something like this for a while. Then a new feature came along to make it more challenging:
"Bookmark this group of tabs"
I love this feature. It's how I lay out my morning reading. One click, and then it's ctrl-W until done. All my comics, all the blogs, and a weather report.
Adding to the last point you mention, the most critical thing you do to any plan is TEST IT.
Stage a disaster.
Either that, or just fake it. Take the data to and try to bring up all the data, bring it up on the internet (under a dr.www.?????.com DNS name for example) and see what's accessable.
In case of failure, tune the plan, and try again in 6 months.
In case of success, tune the plan, and try again in a year.
Jason
Ritually disemboweling a computer on the network does certainly seem to keep the rest of the network in line for a while.
{wavy imagination lines}
Yes, I'm a computer therapist.
Thank you for coming doctor. Our computers have been cranky ever since we 'realigned' our sysadmin (he didn't SEEM to be doing anything useful). Downtime is on the rise, Our databases return 'luser' to one querry in three, and our CIO's Office Assistant's computer only prints swear words!
Ok. I think I know what the problem is. Do you have a fire ax?
A FIRE AX!!!
Yes. Ahh. I believe I saw one on the wall outside. Follow me please.
{obtains ax}
Now, could you lead me to your datacenter?
uh... ok...
{finds a development box, and repeatedly evicerates it with said ax.}
WHAT ARE YOU DOING?!?!?!!!
I just bought you a few days grace. Go back and hire your Sysadmin again. The boxes will be happy you did. Until then, I've scared them into submission.
If you're backing up serious data, tape is the only viable solution.
The technology is old. Or rather, mature. Most, if not all of the bugs have been worked out years ago. At work we have some LTO drives. IIRC, they can be written to at 20 mbytes/second, and they hold 60 gbytes uncompressed (and 120 using the hardware compression. Granted, this assumes the data isn't compressed already. Realisticly, you can only get about 80-90 GB per tape.). You can engineer tapes to be increadibly resiliant to almost anything. The only moving part in DLT or LTO tapes is a spindle that can be turned by the drive. The spool is unwound into the drive itself, and rewound back into the tape cartridge. There are also 2 spool designs with the 'read area' in the middle like cassette tapes. There are also tapes in the 100/200gb range, and the speed and size keeps increasing linearly.
What other medium's are out there?
1) CDROM/DVDROM Slow to write, and not much data. Good shelf life (20-30 years) though. Usually CD's are used for 'archival' purposes. AKA "The IRS decrees that this data must be kept for the next 15 years."
2) Flash? it's solid state, and no moving parts, but the write speed SUCKS for real data sizes. Also, the density just isn't there. IIRC flash cards top out at 512 mb now.
3) Hard disk? WAY to fragile. We moved a batch of 20 or so servers with about 2 TB of disk this weekend. We lost 7 disks (of around 150).
At this point, untill someone comes up with a remarkably new idea, tape will be the king of long term, high density data storage for the forseable future.
It's a big troll, sure. However it is also a chance to dispense some good advice:
1) There's a difference between PC's and 'Server Class' hardware. The biggest is testing. It will work, and it can be supported easily. Drivers are nice and available (generally speaking). Usually dual proc, usually RAID enabled. You can use RAID to speed up read access, but almost no one does. They use it for redundancy (in case a disk flakes out on you). How much money you spend depends on how much downtime costs. If it really costs, you need RAID 1+0 or 0+1. Go with hardware based RAID if you can.
2) Sun hardware. There are many more advantages to sun hardware than what's obvious. Never over look what a good support organization can do. You pay for it, but if something fried, I can have a part in my hands 4 hours later. Sun's low end desktop's are nothing to write home about. However, if you've got Ultra60's or SunBlade 1000 or 2000's, that's some really class hardware. You can do some supprising things with it. [1]
3) Dual procs. On Desktops, even if your simulations don't benifit from dual proc, if they take a while, and they eat that 1 CPU, you'll be happy to have a second (web browsing, etc). On your servers, it's effectivly a must.
4) RAM. On the servers, crank it. On the desktops, you should probably crank it.
5) Cost. If your work is anything like mine, you have 'capital' money, and 'O and M' money. When in doubt, over spec the machines, so that your less likely to have to request more money from the 'capital' pool than you initially quoted. "Going back to the well" viewed poorly.
6) NIS. NIS is evil and the plague. If your in a relativly local office with good connectivity, it's alright. If you try to spread it over WAN links, you're going to get hurt at some point.
7) NTP? Why run a seperate server when you don't have too. Leverage what's already in use in the company. This leads to my last point (and what was the best point of the parent)
8) Get yourself a real sysadmin. These are decisions that s/he is experienced in, and paid to do. Your trial by fire that would come from this will probably drive you insane. Good sysadmins are a rareish breed. I know, I am one. There are a fair number of good ones out of work now. Find one.
[1] The reason largely has to do with cache. Sun chips made in the last 2ish years have 8 MEGS of cache on them (that's even mirrored so it's 16 in total, but you can only use 8). We built a GIS app, and field tested it on Sun and Intel hardware. The intel hardware could deal with 1 to 4 users with less resources than the sun box could. However, the the sun box kept growing up to several hundered users, while the intel box started thrashing hard after 10 or so. We compared a dual US3 box to a dual Xeon P4.
The article is up and loaded swiftly for me. That said, I was hoping for some more 'meat and potatos'. Short version:
1) I can't get my special app (iNews) to work. I need it for my work, so I'm kinda screwed.
2) Web browsing sucks (because IE is a hog). Safari is in beta, but getting better. He didn't mention Moz or Chimera (or whatever they call it this week).
3) It's very nice to work with. If you don't NEED a piece of software that is windows only, you'll love it.
I recently did some pricing (each with 1 gig ram).
Dual 1.25ghz power mac: $2400
Build your own dual Athlon MP: $1100
Build your own dual Xeon: $1700 (iirc)
I know it's not fair, but that's only because I can't build my own power book. (buy a dual Xeon, and you're in the $2-3000 range too.)
I'd love to have a (reasonably powerful) apple on my desk. I just can't justify the price difference.
You're not looking.
www.pricewatch.com shows 181gb scsi drives. There are also 4 x 181gb drives 1 cube away from me for our EMC.
Maxtor may make a 250 gb drive, but you can't use it in your PC. IIRC, ATA133 can only address up to 120gb.
Also, when you're buying scsi, you're not going for single drive density. You're aiming for throwing 10 drives into a RAID 1+0 config (or similar). And finally, those 250gb are new. They're not going to release it for SCSI until they have some experience with it's failure rates, and what not.
Like other's said, it actually lives in a flash ROM chip. Think of it like the BIOS and GRUB combined. It groks networking, all of your disks (tape, cdrom, harddrive, etc), and lots of other things.
You can diagnose hardware errors with it, you can boot off the network with it, you can specify which device you want to boot from, (on multi proc systems, you can specify which CPU will be the 'boot strap' cpu), and you can tell it if you want it to boot at all (if you want it to come up to just the eeprom after a power cycle.)
It's all sorts of goodness. I can't tell you how many times it's saved the day on our sun boxen. And for the hyper masochistic, there's a full blown forth interpreter.
And get this: It doesn't suck. (though it does generally mean something is sucking wind when you are working at that level).
Sun's boot proms have a fully functional forth interpreter in them. Thankfully it's rarely needed these days, but it's there for when it really hits the fan.
Heck, it's even possible to 'mount' a file system, and use a line editor to fix things (granted, it's easier to boot single user, and go from there, but again, if things are really FUBAR'ed...)
NFA's (and Deterministic FA's. They're mathmatically the same thing) are the math backend that define regular expressions.
They're state machines. They're in a given state, and they know how to go to the adjacent states. So given the string 'abc', if you're currently looking at the 'b' (having already seen the 'a'), you know that you'll have a valid match iff the next character is a 'c'. If it's not, you have no match. if you have 'ab[cd]', and are looking at the 'b', you know you have a match if the next char is a 'c' or a 'd', you've got a match. 'c' and 'd' then are the 2 next valid states.
The nifty thing (and the limiting thing) is that true RE's require no memory. Just the knowledge of what state they're currently in. For this reason, no true RE can be written to see if a given string is a palendrome (you can write a RE to match a specific palendrome, but not an arbitrary one).
The difference between a NFA and DFA is that NFA's allow 'null transitions'. This basically says that there are more than 1 state that you can leap out of when you see the next character, because you can go to these special adjacent states without seeing a character, and then leap out. There's also a proof out there that any NFA can be written as a DFA.
All of that said, Perl's extended RE's are not true DFA's. They have some features that can not fit into the DFA model. This is one of Larry's reasons for wanting to make Perl's RE's into true CFG's (context free gramers).
This model is much more powerful than RE's, but at a greater cost, since you have to have memory too. The mathmatical definition of a CFG is a state machine that drags around a stack of memory. The state machine may at arbitrary times push data onto the stack, and later pop it off. It must be done in order though (to match the math model. If you add a second stack, you have the definition of a 'turing machine' (aka the computers on our desk)).
A CFG can be written to match arbitrary palendrome's for example (just push each letter onto the stack, and when you hit the middle, start poping off, and matching each letter. Yes, this is over simplified. The true algorithm is left as an exercise for the reader)
As for configuring networking gear and rackmount equipment, that, too, will slowly come around and start using new technology for their control interfaces. A time will come, in the not so distant future, where it will become financially unfeasable to implement legacy technology on new devices, because the support will be dissapearing, and the devices won't be as functional in the years to come.
I wish.
Sun makes pretty nice hardware right? They have this spiffy box called an E10k, or starfire. Can have 64 CPU's in 1 domain (hmmm... make -j 64), or 16 different, 4 proc servers, or any other combination you can dream of (multiples of 4 CPU's per domain).
How does the SSP talk to, configure, and control all this power? Through something called a JTAG interface. It's been around since the early 1970's. It works at 300 baud. You get your console, and all the configuration information on the box via 300 baud, SHARED accross the entire platform. [1]
Why the heck did they do this? Because it is well tested, and works. The API's are perfectly understood by everyone, the firmware's been around for ages. They could put it in there, and know that it will work. period.
I wish that you were right. But companies will continue doing things like the above.
[1] To be slightly fair, the box was first designed and introduced in '97 or '98, though when upgraded, it still makes a kicking oracle server with rock solid stability
How many people still have a serial port palm pilot?
:-) )
How many people need to configure a piece of network gear?
How many people have needed to get to the serial consone on their unix box?
How many people have rack mount gear that the only console is serial?
Serial won't die any time soon.
(if dell tries to take away serial ports, the admins who run dell.com will probably hold the site hostage until they change their mind.
MQSeries is a pretty useful application. It gives you a unified method of talking from anything to anything.
Backend CIS system on a mainframe, and a java based linux webserver need to talk? No problem.
IIS/MSSQL box need to pass data back to DB2 running on an AS/400? No problem.
Sun/Oracle box talking to a custom homegrown app? No problem.
It's a unified API, and a unified 'paradigm' of communication. It can talk TCP, SNA, and every other network type out there. It runs on every os/hardware platform you're likely to see in production. And it offer's a promise of message delivery. It won't just try and give up. It guarantees (sp) that the message will get to where it's destined to go.
The downside is that it's kinda heavy weight. It's also a little network chatty, and it's also rather expensive. However it does make it easier for your developers. If they all know how to get from their application to a Message Queue, then all of your developers can get data back and forth between ANY of your applications.
All of that said, there is some talk in the jabber community about trying to move into the 'middleware' space. That'd be a really interesting thing to watch. Doing that guaranteed delivery isn't something to be taken lightly, and if they could crack the portability problem (how many of us have as/400's to play with?) it'd be pretty cool.
Out of curiosity, why the heck are you using AIX+Oracle? It may be advertised to work, but IBM and Oracle are huge cometitors in the enterprise DB space (DB2 for IBM), so I'm not really supprised your having stability problems.
Oracle's 'reference platform's right now are Sun and/or Linux. If you must stay with oracle, you might want to switch. If you must stay with AIX, you might want to go db2 (though the first will be far easier... oracle export -> oracle import and you're done)
Outside of the integrated amp weighing a ton, there's another reason. If you listen to the 'audiophiles' out there, one of the things that they rave about is 'build quality'. 99% of the time, what they mean by that is 'weight'. "Feels cheeply made" is a complaint about a lot of 'low-fi' and 'mid-fi' gear (low and mid-fi gear is defined as equipment one or more steps down (usually in price) from the gear you currently own (or the gear you're currently lusting for)).
There do exist good, quality, peices of equipment that produce music wonderfully. Trust your ears over your other senses (unless you're looking at video gear. At which point your eyes should only see 2 things: the picture and the price. People can be conned into wasting a lot of green on video gear.)
You must have backup. If you don't have someone else to do this, and you have a shop buging you extreemly regularly you need to get them to hire a partner.
Phrase it something like this: I am the only person who understands how these systems run. I'm the only person who can fix them when this breaks. If I get hit by a bus, you're in a world of hurt. If I don't start getting more sleep, I'm going to burnout, and you're in a world of hurt. I work 70 hours a week regularly, and the backlog keeps piling up. I need someone to help me.
Then (well, after 2-3 months training the person), you can take weeks off from pager duty. There's a reason doctors have on call rotations. You should too.
One of the things that I've done is set expectations at my place of work. 90% of the time, my pager is available to them when they need me. That other 10% is well communicated in advance, and my boss knows that my pager will be on my bedstand, but I'll be in another state.
(Oh, and reguarding the person who saw 'recurring nights of database server issues' as a sign of stupidity, they might be right, they might be wrong. We've had a sun e4800 go really flakely on us recently. It took WEEKS of long nights (since the box was production, and we couldn't take it down in the day) to get the hardware on that box stable (it would work fine for 3 days or so, then flake out hard). Sun wouldn't give us a new box (with at least an understandable reason), and keep insisting on replacing individual pieces. And it certainly wasn't our doing. It turned out to be a bug deep in the IO chassie's firmware.)
You (and the others following this thread) owe it to yourselves to give postfix a try. It is clean, and elegant. All of it's configs live in 2 files (main.cf and master.cf), and it is powerful.
Postfix has solid anti-spam measures, fantastic performance, readable config files (that are well commented from the start... heck there are about 10-20 comment lines per config line), and you only need to set like 3 lines in the config to get a working mail server.
People have some good points, but one thing that package managers need to understand is rollback. ie: I had gnome 1.4 installed. Worked well. I upgrade to gnome 2.1. 2.1 might just turn out to not do what I need at the moment, or it might be incompatable in an important way. I should be able to 'roll back' the system to 1.4 without performing major surgery.
Yes, this will take more disk space (to hold the old versions of all the files, and the metadata to restore them). But in an enterprise environment, you need it. Sun has been doing this for years with their OS patches, and we should be able to steal it for packages too.
That's one of the most insightful comments I've read on /.
Your point about binary compatability on linux is almost painfully valid at times.[1] It'd get a LOT easier if glibc/gcc would finally decide to stop breaking backwards compatability.
There is a work around (most of the time). You could distribute a statically compiled binary. It's not pretty, and it's not perfect, but it will solve a lot of the problems.
[1] Ever had a debian firewall on a pentium/60 with no gcc, and redhat behind that, and no other available linux boxen to compile things on? Updating the kernel on that firewall was a royal pain.
About 99%. And we seem to have gotten a reputation as being rather 'harsh' as a Grand Jury on our attorneys.
This fact has lead to a wide spread belief that grand juries are a rubber stamp on the process. However, you need to realize that about 95% of the cases we see are truly open and shut: "He has confessed that he did this.", or "He was deported, he's back, he's not allowed back, his finger prints match."
As you might guess, we were quite happy when we saw those cases that fell into the 5%. Some were done by amazingly stupid people, others were wide spread rings of very inteligent people who just got too greedy.
Finally, remember that we have a very light 'burden'. Proving that someone probably did something, when you only hear the evidence of the prosicution (sp) is pretty easy.
How would I improve the Jury system? I'd make it a lot closer to my federal grand[1] jury experience.
They call 45 ish people. They need 23. The list of valid excuses is pretty lean. When we got there, there were 26 of us. They put our names in a bucket, turned it, and pulled three names. The judge said "You three can go." Then swore the rest of us in as grand jurors.
This would have to be modified slightly for a trial (or petite) jury. You'd have to give the lawyers SOME strikes for obvious bias, and start with a larger pool for the OJ Simpson cases of the world. Just make the number of strikes proportional to the pool. Once you have 24 'good men and true' (16 + 8 alternates), start the trial.
[1] Note: Grand jury's decide if a case can go to trial. A trial jury (petite jury) is at the trial, with the judge, etc. At the federal level (at least in Atlanta, Ga), they are set for a year term, with 6 months after that for 'recall' status. There are 4 juries at any given time, one for the first 4 tuesdays in a given month. We show up between 1 and 4 days a month (usually 1 or 2. The last three day was 4 years ago, and no one remembers a 5 day docket), and see around 5-8 cases/investigations a day. Our 'burden' is much lighter than a trial jury. We only have to see if the accused had 'probable cause' in this case (aka: did they probably do the deed). We also only hear the prosecution's side of the case.
We also hear investigations, basically a way that an attourney can question a potentially hostile witness under oath. These can go on for a while (which leads to the 'recall' status). We'd come back for minimal time to hear more evidence, or to see if an inditement is asked for from the AG.
It was a fasinating experience. I saw a LOT of stuff I wouldn't have seen otherwise, but it was quite boring at times, and my boss wasn't exactly amused (nothing they could do about it though... The constitution has a minor trump card in this matter. If they even sniff that you're let go because of it, the feds will come down like a ton of bricks).
As an asside, 'private company' generally means a company like google, owned by a small group of indivduals. Verio is a public company. It is owned by it's shareholders.
Your example from your pre-sales day is a totally different world. Verio had already decided to do business with thing.net, and had written a contract to do so. You and this 'someone' had NOT started doing business. You're totally free to choose who to do business with. However, once you start, it depends on the contract.
I very strongly believe that if you say (like in a contract) that you're going to do something, you need to follow through with it.
As for your 'get an account with another service provider', sure that's possible. However, if BillG was really after you, and if all the service providors followed your logic, then you would be silenced, in the medium of the internet.
As for the reference to the middle ages, it's something called an analogy. A story used to illustrate a point. Exageration? Maybe. However the nobles DID work quite hard to destroy opposition. Why do you think the French Revolution was so keen to have a free press? The nobles silenced many a newspaper. Sure, they could have gone back to writing the things by hand (the equivelent of you're suggestion that they should change medium), but that would have been much less effecient.
Verio [probably] said to themselves, "On the one hand we have a bunch of activists who pay their bills, but who aren't a significant source of revenue for us. On the other hand we have Dow Chemical, a gigantic multinational corporation that could throw us a lot of money if we have a good relationship with them." And they made a business decision.
If that's what happened, I really don't see a reason to get all up-in-arms. Yes, this is an inconvenience for the activists. But, if it happened the way I'm guessing, nobody did anything illegal, or even unethical.
I won't go into the rest of your post, which I thought was well and good. However, while I see the logic of your 'wild assed guess', I disagree with your statement that it's 'not unethical'.
Say you have a mindspring.com account. On your personal web page, you post that 'bill gates of borg' icon that he's reputed to hate so much. He causes a letter to be sent to mindspring, and your account is revoked, using the same logic you just called 'a business decision'.
This is precisely the kind of thing that power leads to, if it's not curbed. Look at the nobles of the middle ages. Not held accountable, they could kill anyone they wished without repurcussions. This is just a modern day version of that.
We must protect fredom of speech (obviously keeping in mind the limits WRT slander and libel). We must we must protect those liberties.
Now, if thing.net violated libel and slander laws, they get what is coming to them. If they didn't, then what happened is unethical, and should be illegal.
I can't echo this enough. It's great multiplayer, but not LAN play. You can have a bunch of people playing at one computer though.
It rocks.
From the article:
University of Alberta's Bob Beck said he is forced to buy out-of-date UltraSparc II-based E450 servers instead of newer UltraSparc III-based V880 machines for the university's SunSITE software exchange.
This seems odd to me: 1) OpenBSD doesn't support SMP yet, right? 2) v880's must have multiples of 2 CPU's (up to 8).
Sunsite might be better off grabing some of those 1U v120's, throwing a dual channal diff scsi card in there, and using an a1000 array (or maybe a t3 array... with only 1 cpu you probably need the hardware raid these offer rather than the d1000's or a5200's). More disk, less rack space, less power.
Now, the v880's rock. Great price point, 8 cpu's, 2 FC-AL planes for a total of 12x73 gig disks, 10 PCI slots (2 x 64bit/66MHz), onboard gigabit fiber... the list goes on. It's a great box (for more details, hit up store.sun.com, select servers, find 'low end servers', and select the v880. And note that that's 'list price'. You can get up to a third off of it from most resellers)
For reference:
4x itanium 800MHz dell 7150: 8x73 gig disks is $61,113.00.
4x usIII 900 MHz sun v880, 6x 73 gig fcal disks is $59,995.00
(That's the closest 'apples to apples' match I could make. I chose itanium vs usIII because they're both true 64 bit chips. Though the expansion of the Dell isn't as nice... the sun can add 4 more proc's and 6 more disks. The dell can add more memory... 32gig tops the sun v880, and 64 gig the dell)