Slashdot Mirror
SGI behind Linux: it's official
Cornel Ciocîrlan writes "SGI announced yesterday that it will support Linux as its fourth operating system. The press release talks about contributions to the open-source community in the area of high-performance file systems, OpenGL, high-bandwidth I/O, compilers and other scalability features. "
What the hell is UNICOS®?
"HP-UX and AIX really only scale up 16 processors, Digital UNIX even less so"
This doesn't surprise me. I know a kernel hacker who worked for Digital who was always finding the same bugs in the Digital Unix kernel as those he had come across in the Linux kernel. (Cue Twilight Zone music.)
Coincidence?
"sieve-like security"? You're being kind. It is almost as bad as NT. I have worked with large systems since, well, that was what they were called, I made the jump to UNIX pretty early, and I am still amazed at what ships out of SGI. SGI is famous for this. Than and 400MB tarball binary patches. More than a few people pointed out that the fact that Netscape was raiding the two most bug-tolerant places around (Apple and SGI -- both barely better than M$) for staff was a Bad Thing. I would rather keep IRIX somewhere else -- not in Linux, thank you very much.
...
And not to gloat, but I haven't heard ant FreeBSD weenies saying "yippee -- maybe we can have some of SGI's elephantine bugware released under a BSD licence!" Perhasps we should take a hint from the crabby old men for a change
Sorry about that -- one handed with coffee in the other and an eye on ADSM ...
Yes it is, but it only scales to 14 CPUs. With Parallel Sysplex, you can get 34 machines, but that is a cluster. Even with the Uber-OS, it is still not MMP.
I recall the release of GLX as OSS, but what is this OpenVault mentioned in the article? Some kind of version control? And why did the article lack links of any kind? Seems like it might have been nice if the words GLX and OpenVault were links to the respective stories...
I work with SGI's and they are marvelously engineered. I would love to see them push Linux too, but I'm suspicious. They announced months ago they were going to support Linux but nothing has happened. In this article, they say they don't have any current plans and that makes me think nothing is gonna happen. They, along with the other corporations need to do more than just talk hype.
Load level means really nothing until you investigate further. Use sar & top and find your bottleneck, a high load level means that you have lots of processes waiting to use the CPU. You could be I/O bound or memory starved and the CPU idly twirling it's fingers waiting for your disks or waiting for a block of memory.
I've had a load over 40 before when I had over 200+ sendmail processes running at once and over 70+ popper processes running at once. There is nothing bad about this mail was comming in fine, humanly you couldn't notice any slowdown at all with the machine. That is the mark of a good machine, throw everything at it and it keeps chugging along, try that with NT.
The information on the ping times can't even be guessed at what causes it. Traceroute is much more usefull than a ping, ping just checks accessibility. When you say "other way" are you pinging from a connected server? Pinging from a server on the internet has way better performance than pinging through a modem (of course 500ms is still high even for that). You have something else wrong than your Irix OS, hell CNN, ATT, etc. are running it and they are not low bandwidth sites in any shape or form.
College students all over the world are thinking about "Enterprise Linux Distribution". Eventually they will even know the definition of the word 'Enterprise' if they keep at it.
Seriously, though, who is going to write this "Open Source" Enterprise Linux? Shouldn't they maybe be fired for not working on what their employer is paying them to do? (in almost all cases certainly not Linux)
Intel has already picked AIX with some SCO utilities as the reference Unix for Merced.
Obviously you haven't used Irix 6.5, which is a much improved release of one of the finest operating systems in the world.
Last year, I seem to remember SGI (Then Silicon Graphics Inc...*cough*) issued a press release saying that they are planning to port Irix to IA64 (i.e. Merced / McKinley). I haven't seen anything recent, though on it. You might be able to dig up that press release on their site..
Here's the SGI press release..
I agree, the graphics software for Linux can't even come close to what's on IRIX. They have a few cool graphics software for NT now on those new SGIs. When they bring it to Linux I'm goign to buy one.
Are you suggesting that some Linux hacker stole Digital Unix code? Wouldn't surprise me. Explains some things, actually.
If the linux file servers aren't compatible, why are you using them? People need to quit this Linux Jihad crap and use what works.
The SGI press release speaks much of interoperability. The question is
what mechanisms they intend to employ to those ends. The only thing
specifically mentioned is their official support for Samba. While this
is a Good Thing, it is important to remember that Samba is, at its base,
a mechanism for providing interoperability on one specific vendor's
terms: Microsoft's.
If SGI is truly interested in pushing interoperability, what they ought
to be pushing, IMO, is open-standards, cross-platform interoperability
mechanisms such as CORBA, Java and LDAP.
Ok lets face it, obviously most of you reading this are younger and NOT working in LARGE hardware manufacturers shops. Good software that scales well and is robust takes a hell of a lot of time.
SGI is probably hard at work porting stuff.
I see from a few web sites that the 320 VW runs
linux 2.2.5 pretty good. ^_^
Lets face it, Linux is NOT ready for serious
HARD CORE production work at the same level as the
Solaris's and Irix's of this world. A few examples you say? Async IO. I wouldn't bother
with databases until this happens in the kernel.
Sgi(should this be capatilized now?) or IBM could
bring this to the table. No journaled LARGE file
support. ext2fs and xiafs,et al are all well and good, but something like SGI's new cxfs would
be absolutely phenomenal for beowulfs!! clustered
file systems are the future. Especially with
larger sites moving to SAN and failsafe solutions.
"But XXX!" you say. "There are plenty of places
using Linux in production work!" Yes, but where is the money for an IBM or an SGI? SGI's big install base was the Indigo2. A kickass desktop 10 years ago. Now they do scalable FAST servers and cling to the remnants of super-computing, while trying to revive the low end with NT and Linux. No margin, what value-add with a linux
box? Has sgi EVER been easily affordable?
Sgi has a lot of cool products I'd LOVE to see
go OpenSource. (performer, cxfs,failsafe,nqe,
4dwm,etc... But I bet we'll only see older
versions like xfs instead of cxfs. That'll still
rock! Sgi backed out of selling the secrets of
Cellular Irix to Mr. Gates & co. perhaps
they'll bring decent SMP and excellent scalability. (ala' ASCII Blue Mountain anyone?)
They are planning 256p machine sales now. Why
not make their future clusters on Intel with Linux? Granted a lot of kernel work will have to
go on again.
A buddy of mine sez that the first Linux only
boxes from SGI are ALREADY out in the field!!
the 1400L or somthing. IBM is still trying to
get their first linux only up and going...
And where is SUN and HP?
And yes I do know WAY more about this then I'm saying, but I am covered by a DEC, an AT&T, a
SUN and SGI non-disclosure agreements. This stuff
is all public or conjecture.
A unix guy..
Is anyone thinking about an "Enterprise Linux Distribution?
I wouldn't hold your breath. Linux is largely developed by volunteers, many of them college students (and most of the rest under 30). I seriously doubt any of them can afford to buy a Cray to play around with in their spare time.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Posted by OGL:
Question: what's stopping SGI (or anyone else for that matter) from enhancing Linux? I can't think of any reason why a cxfs patch wouldn't be integrated into the kernel proper.
-W.W.
Well, yeah, but then the FreeBSD guys believe that all the SysV folks will See The Light and switch to a simpler (but nonstandard) initialization sequence.
Again, the world keeps spinning. The great thing is that with Linux, we can pick up after SGI, fix the bugs, and still have a better system than when it started. I don't really think that the BSD folks would mind that, either, if SGI wanted to use FreeBSD or NetBSD as the in-house OS of choice.
There's a lot that I hate(d) about Irix, and when I saw the CERT advisory about "4dgifts" and other unpassworded default accounts I decided that security wasn't top priority on Irix.
Still, their I/O has always been top notch. If we can have better, fully reentrant I/O in Linux and ditch the crufty old bits of Irix, why not?
Incidentally, the crabby old men are usually right, as I'm sure you were implying. But sometimes they don't see the big picture.
Remember that what's inside of you doesn't matter because nobody can see it.
>>...if SGI decides to hold close it's dead IRIX >>technology the community will lose a very good
;-))
>>scalable operating system.
With sieve-like security and a tiny user base. I used to use Irix and loved it, but come on... the world keeps on spinning. Cellular Irix will probably show up on the ultra-high-end; go get yerself a O2K, maaan. As for the midrange, who cares? Linux scales as well as Irix on an O2...
I'm overjoyed that SGI is bringing in the heavy I/O artillery for Linux. Unless you really despise all us unwashed Linux users, you should be too. AOL will probably be enough (by themselves) to drive Irix *support*, but maybe not *development*, especially on the low end. (AOL runs AOLserver on O2Ks with Sybase as the main backend; they're keeping all 3 of these in business by my estimation
Incidentally, Irix goes to 128-way on the big CC-NUMA systems. It effortlessly did 20-way on our (straight SMP) Onyx when I was at Cornell... I don't disagree that it rules, but try explaining that to a PHB that thinks GUI hooks in a server OS kernel are Modern.
They OSS'ed OpenVault, why wouldn't they do the same (or similar) with XFS? Well, methinks they may take the opportunity to engineer something better. I went and bought the Be book on Filesystem Design when I realized the level of flexibility the VFS gives you. It's pretty cool.
Anyways, SGI == I/O and we should all rejoice. The chances of NT retaining a lead in brute-force I/O (which is a big, big hangup for Linux in the scalability/multithreaded department) should now be slim-to-none. Hah, Hah... and we all thought SGI had sold out. Maybe they just pulled an IBM.
Long Live SGI.
Remember that what's inside of you doesn't matter because nobody can see it.
Oh well, even though they're not announcing it, I guess we should assume that IRIX is officially dead.
If they didn't officially announce it, why should anyone assume this? I mean, given the MIPS processor line has been extended at least through the R14000?
Personally, I prefer IRIX to Linux. Free or not.
According to the press release, they are moving technology from UNICOS to Irix, so I guess Irix is not _quite_ dead yet, but it might be reserved for high-end systems.
Makes sense. Apparently one of the main reasons that they don't just port IRIX to IA32 is that they want to take advantage of all sorts of 64 bit stuff in IRIX, and porting to a 32 bit platform now would disturb that.
I wouldn't discount IRIX yet. Let's see how it does on IA64/Merced/McKinley. Could well beat Linux, since gcc is likely to suck on IA64.
EEK!
:/
World.std.com!
I'm working on moving away from them currently.
try fingering my account sometime. Only time i've ever seen a load balance of 16 on a Unix machine. Before that, the highest I'd seen was a 6.0 on a heavily used mailhub/proxy server/Netscape Calendar server at my old job.
Not to mention that when I dial into the World, I get 500ms ping times to world.std.com. When I access the internet in any other way, I get 100ms or less. The server responds quick unless you're directly connected to it.
I tried traceroute, but it didn't help.
I was dailing into the server, so it showed 1 hop.
Someone suggested it could be whatever is connecting the modems to the server itself.
Not quite. DG/UX scales to 128 processors, and I believe Dynix/ptx and NCR's Unix are similar in that respect. Sure, none of them are major players in the Unix world, but they certainly do scale.
Oh, and isn't OS/390 technically classed as Unix these days? :-)
"The invisible and the non-existent look very much alike." -- Delos B. McKown
Modern filesystems are extremely complex. Once I put together a very simple filesystem -- one less complicated even than DOS's FAT -- and it was a *nightmare* to write. Implementing extensible files, process synchronization and caching dealing directly with inodes and raw sectors makes for some very ugly code.
If a new filesystem is very similar to an old one that is already in the kernel, things are much easier as quite a lot of code can be reused. Howevre, if cxfs is a journaling system (I don't know too much about it), then we are almost starting over.
I imagine that putting cxfs support into the Linux kernel would be a Big Deal. SGI would have to open source a lot of IRIX code that presumably cost them a great deal to develop, or there would have to be an extremely large Linux development effort.
If we are lucky, SGI will decide that Linux really needs a cutting edge, high performance (better than ext2fs), file system and that they might as well just give us the filesystem code out of IRIX.
One way or the other, I'm liking SGI more and more.
--Lenny
R12000 is out
R14000 is on the way
The interesting stuff (Beast) was killed,
but I pray they'll resume as:
1. Merced is delayed
2. EPIC will show how wasteful of resources it is
SGI/MIPS has always shown how it is not just about
MHz. There's one very good BINARY COMPATIBLE opportunity open:
a THREADED processor
Of course, for O2K/Cray style machines, a low-volume, expensive VECTOR processor would
be a better match for most nodes in most applications.
Oh well, even though they're not announcing it, I guess we should assume that IRIX is officially dead. If they're not porting IRIX to Intel, and they're certainly not going to continue developing their MIPS hardware line after spinning MIPS, Inc. off last year, then what's going to happen to that IRIX source tree?
What a shame. IRIX and Solaris are the only serious scalable UNIXen on the market. HP-UX and AIX really only scale up 16 processors, Digital UNIX even less so, while IRIX and Solaris can handle a good 64 processors with a reasonable backplane. This means that if SGI decides to hold close it's dead IRIX technology the community will lose a very good scalable operating system.
Betcha the admins over at world.std.com are frowning over this... A good consolation prize might be freeing the source for XFS. Who thinks SGI might be willing to take such a drastic course of action?
Why does it make me nervous that all the financially troubled companies (ie. those who get beat up by MS) turn to Linux? Ok, so there's IBM and Oracle too.
Hmmm: "The more you tighten your grip, the more they'll slip through your fingers".
Perhaps if we all boycotted Cavedog... nah.
"The biggest problem with communication is the illusion that it has taken place."
The highest I've ever seen was on a RS/6000 model
;)
370 doing Sendmail 8.6.x duty.. The box got up
to a load of 160 before crashing.... Granted, it
was because of a nasty tight script loop (which wasn't mine, honest!!) and didn't reflect a realistic situation, but lord knows if I wasn't on a 3164 I would have gotten a screenshot (had to go terminal because everything else went to shit)..
My Linux laptop is regularly loaded > 2, but that's because it's usually doing a kernel rebuild at the same time as a KDE rebuild (or a GTK rebuild, or a wild attempt at a Mozilla build, or....
they're presenting a paper called:
sounds like they might be open sourcing XFS , cool!
"Yeah well
The fonze doesn't do conga lines. But i'll tell you one thing, if sgi had any sauce, they'd start releaseing xservers with insane 3d support and not pull a mozilla and say, "here's the code guys, make it work on linux and we'll make more money from it."
this is really good news. it was refreshing to read a press release that mentioned more than just 'cutting edge' this and 'industry leader' that -- SGI seem to have a healthy attitude towards Open Source and hopefully we'll begin to see Linux become a realistic environment for running up a 3D workstation or Non-linear digital video suite. Some of us more graphic types would like a little more to play with than the Gimp and Xview.
And I KNOW Bowie J. Poag gonna make some sweet Propaganda themes with a copy of Maya for Linux!
--
Rare Window - free your photos
Golly, I guess that means the Compaq Wildfire project is in big trouble. Well, I suppose that's possible, but I would be surprised if scalability of the OS is the problem. For those not obsessive Q-watchers, Wildfire is an alleged large scale SMP Alpha box, yet to be introduced, with plans up to 72 processors.
I concede the point that IRIX and Solaris have the most scalability street cred. And Wildfire is still in the labs, rather than the stores.
ObLinux: Is there a well defined set of goals for scalability other than SMP support for large numbers of processors and not crashing (sorry NT :-)? Is anyone thinking about an "Enterprise Linux Distribution?"