Slashdot Mirror
SGI NUMAflex Linux System On Display @ SC2002
jarrod.smith writes "
According to SGI will unveil
its Intel® Itanium® 2
NUMAflex shared-memory supercomputer architecture (which runs Linux as its OS) at Supercomputing 2002 which runs this week in Baltimore, MD.
The link at SGI says the system will be on display at the show. The exhibit floor opens this evening. Unfortunately I did not go this year. Can those lucky enough to be at the meeting scope it out and post comments?"
Wow!
NumaFLEX... And to think... All that AMD could come up with was Athlon 64.
You'da thunk that they'd at least stuck a period or an 'e' on there somewhere...
eAthlon.64?
Life is the leading cause of death in America.
Additionally, it offers unparalleled scalability in the line of Linux supercomputing. This is a system built to grow with a business, although your business better be pretty much grown already to back the check you'd need to fill out to buy it.
My conclusion: it's an excellent largish solution for academia seeking a more stable environment than can be achieved with Beowulf clustering and excellent pricewise solution for businesses seeking to expand without sinking a lot of money into unnecessary costs.
According to SGI will unveil its
According to who?
I demand an answer!
yep, I was slated to go, and then got told "no, we don't have the money in the budget"...and to top it off, even /. is rubbing it in...
puts head down and weeps as images of shiny, multi-colored SGI systems float through my head
PC moderators can suck my White pierced, tattooed dick. If you think pride == hate, s/dick/Aryan meat mallet/g.
(which runs Linux as its OS)
WRONG! It runs linux as it's kernel.
1: Fine-tune Linux to fit the platform
2: Design the platform to utilize linux
So it sounds fair to me.. Consider installing some propriatory OS instead.. where they cannot play around, change kernel design, drivers, VM or whatever they fancy. Would not that be a greater drawback ?
Probable impossibilities are to be preferred to improbable possibilities.
Aristotele
SGI was working on Linux (as in modifying the linux kernel and some other software components) for something like this.
Can you give me one good reason why Linux will not be a good supercomputer?
Hmmm... Pie...
Linux = kernel
GNU/Linux = operating system
It does not run GNU/Linux -- it runs OSCAR.
Get it right, Ho^Hemos!
The Los Alamos National Laboratory is building a supercomputer based on a Beowulf Cluster with 1024 nodes (2 processor in each node). You can read the story here or in this Slashdot thread.
Well it seems like a good choice for SGI. It allows them to come out with a supercomputer using "more" commodity parts (e.g. Intel processors) running an OS that is the hot topic and, one would assume, cheaper for them to sell. After all, what would their options be if they had not chose Linux? Irix only runs on MIPS afaik. Maybe BeOS ;)
This is more SGI schizo behaviour. They can't really decide what they want to do or be, what hardware platform they want to push or what OS they want to use. Another sad day in a long line of sad days for a once cool company.
Consider installing some propriatory OS instead.. where they cannot play around, change kernel design, drivers, VM or whatever they fancy. Would not that be a greater drawback ?
But IBM already have the source to AIX... they wrote it.
SGI has been working on this machine for a while now, and you can be sure the linux it runs has been tweaked for performance.
This machine is Intel based. SGI had to choose to port IRIX or make Linux run well. IRIX isn't going on their IA64 computers, period.
SGI has always focused on high speed internal communication in its machines, and SGI has tended to lag behind the curve in the MIPS processor's raw MHZ. The IA64 chips are much faster than any MIPS processor out there. This machine has some amazing performance--I think everyone will be surprised at what SGI has done. If you need a high end single image shared memory Linux or Intel solution, SGI has filled that role.
To be fair, if you don't need this machine.. don't buy it or think about it. SGI is a dying company and in the long term you'll get better support elsewhere. Sun and HP will have competing machines out, but they won't perform as well. I trust HP and Sun to be around longer than SGI though, and they won't fuck you over like SGI will.
What else can I say? Oh, the one I saw was painted in penguin colors.
The "open source developers" of which you speak now count among their number professional developers from companies like IBM and SGI who have been working hand over foot for the last few years to bring Linux to large computing platforms. Check the development mailing lists.
It's not like Linus has been sitting in his bedroom coding for a decade and now suddenly SGI is going to download the kernel and throw it at supercomputing hardware. Big companies are and have been investing development dollars in Linux in order to make Linux ready for platforms like this one. And the great thing about Linux is that whatever SGI or IBM adds, the community tends to get back in the form of permanent enhancements to Linux.
STOP . AMERICA . NOW
With all of the time and money they're putting into this, wouldn't it make sense to code their own operating system? I don't know much about coding OSes, but it seems that it would be the best way.
Regarding Linux: I don't know a whole lot about the kenel structure and what-not, but it seems that the good things about Linux are things that a supercomputer wouldn't care about (portability, a good GUI platform, etc.). Why not use IRIX?
You know, if they wanted a stable OS, they'd use Windows 98 and just pay the money for the second edition and then patch IE. That's pretty stable, right?
Well, damn as nearly. Linux is only in catchup when the manufacturor will not release spec on how to use thier hardware.
When it comes to NUMA machines, Linux is up there. It may not excel at everything (yet - I'm sure that it will get there if it's not already). I'm mostly talking about the 2.5 kernel series.
From the status list
New scheduler for improved scalability (Ingo Molnar)
Support for Next Generation POSIX Threading (NGPT team)
Syscall interface for CPU task affinity (Robert Love)
Hotplug CPU support (Rusty Russell)
NUMA topology support (Matt Dobson)
Per-cpu hot & cold page lists (Andrew Morton, Martin Bligh)
NUMA aware scheduler extensions (Erich Focht, Michael Hohnbaum)
The biggest performance changes in 2.5 seem to be in the many thread and many CPU region, including NUMA.
I'd trust it. (Yes, I do do scientific supercomputing).
But at the core it won't resemble anything you've seen. They use linux has a base and design it exactly to their needs. It won't be of much use to about 99.95% of the linux user base. The only reason they chose linux is because its already established and porting over IRIX would mean a near total rewrite. BTW they also own the original UNIX trademark so you can legally call IRIX "UNIX".
Only the State obtains its revenue by coercion. - Murray Rothbard
I wonder which is worse, Beowulf posts or goatse.cx links?
LINUX!? WHY LINUX!? Why not a stable OS...
Like Windows ME!
Just think how fast Windows 98SE would fly on this baby!!
Fascism should more properly be called corporatism, since it is the merger of state and corporate power.
...okay so Linux is being applied to all these terrific projects of scale both large and small. Is it because it's an open system with seemingly hyperactive development or is it because it's simply better than anything else out there?
I'm trying my best to maintain a level of respect for the MS operating system product so I'd like to know if anyone knows of any amazing projects MS OSs are being used for. For that matter, what about other OSs in general?
I think it's terrific that Linux is used this way but I wonder if it's because of its availability or because of its technology. I tend to think it's for its availability but I'm no expert. I think answers and other points of perspective from others in the Slashdot community would help to show some objectivity here.
what kernel version?
I've been hanging out at SC2002 all day, and I can tell you that nearly every booth on the show floor is showcasing Linux. Of course all the Linux cluster vendors have it, but so does sgi, Sun, IBM, Intel, AMD, HP, Compaq (separate booths - guess the merger isn't *really* done yet), and all the smaller vendors, to say nothing of all the research labs, etc.
Large linux systems and clusters are really all the rage right now in SC circles. I think the only booths I saw here not using Linux were the Apple booth (though they did have one gorgeous brand-new G4 running Xfree and twm, the sick bastards!) and the Japanese manufacturers NEC and Fujitsu (off in their own worlds, as always).
Linux isn't a big surprise to the SC set, though - this is a group that's used to UNIX. Hell, Microsoft doesn't even have a booth here, and they were at the last LinuxWorld conference.
-Isaac
I am not a lawyer, and this is not legal advice. For Entertainment Purposes Only.
Unfortunatly, I havn't found the time to test it on my Cray, so I might be talking out of my ass.
"A language that doesn't affect the way you think about programming, is not worth knowing" - Alan Perlis
You pay a lot of money to get a very large computer that can do very large tasks very fast.
Wasting 20-40% of the resources of your $2k desktop on your OS's feature bloat may not be too bad, but wasting 20-40% of the resources of your $5 mil supercomputer is a lot of money.
Or put another way, Linux is used in supercomputers because it can be set up to do exactly what you want it to, and ONLY that - which for most HPC applications is compiling and running custom code to solve Big Problems.
You're not going to use a 512 processor supercomputer to Save Christmas by being able to get those pictures off your digital camera without spending 3 hours trying to download the drivers.
paintball
A "supercomputer" just has a lot of computational horsepower. It seems that the whole idea of this thing is to expose lots of cpus and physical memory to a single OS image. That's relatively simple when compared to virtualized machines like z90's or E12K's.
Just think of it as your grandson's palmpilot.
There may be OS scalability issues. However, SGI should be well equipped to address them.
I'd be curious to see how a single Oracle instance might scale on such a beast.
A Pirate and a Puritan look the same on a balance sheet.
True. The more relevant question would be: is AIX worth the bother of maintaining and porting? What value does IBM derive from that R&D expenditure. Does any of that serve to distinguish IBM from it's rivals in the marketplace?
A Pirate and a Puritan look the same on a balance sheet.
I dunno. As much as processes bop from system board to system board on the bigger Sun machines, I am no longer certain that Linux would not be able to eat Sun's lunch in the +32 CPU box arena. Add IBM and SGI into the mix and you've more than enough engineering to bridge the gap.
A Pirate and a Puritan look the same on a balance sheet.
- No royalties. They can use it, hack it, sell it. Whatever they want, and never have to cut a check to anyone.
- The resources. The Linux development community is unlike any other. Using Linux means you have access to all sorts of development and product resources for absolutely free. The newsgroups are friendly, the documentation is deep. And if you're doing something weird, do it with Linux and chances are someone will help you.
- The name. If you need to impress the suits and get funding, Linux is a name you want to include. For a lot of people, Linux=cutting-edge technology. They don't understand it, but they know it's powerful, and they know it's gaining ground fast.
- The power. There's no two ways around it. Linux is a powerful and flexible system. You can push it, pull it, tune it and tweak it to do just about anything. Unlike some other OSes, the kernel was written to stand on its own, not necessarily part of any prefab package. There's no GUI code in the TCP/IP stack, and it's just as happy in a PDA as it is in a supercomputer.
Could you honestly immagine LLNL buying a Windows-based clustered supercomputer? Yeah. Me neither.
Using Linux helps companies keep from having to re-invent the wheel while at the same time keeping their options open and their money in their own pocket. It works so well it's a wonder more companies don't use it.For those afraid of the GPL, BSD presents a tempting alternative. But again, you lose a bit of the development resources and don't have the name to use to get your funding. For most people, though, GPL isn't a problem.
"With sufficient thrust, pigs fly just fine. However, this is not necessarily a good idea...."
RFC 1925
This isn't VMS or NT we're talking about here. The lack of good threading won't necessarily hamper a serious Unix application. Some of them don't even support threading very well. So, this particular Linux disadvantage may end up being a moot point.
You don't need threads to scale Unix apps.
A Pirate and a Puritan look the same on a balance sheet.
i want to SEE it. What does it look like, give me some jpgs to oogle over. Im sick of just imagining these supercomputers and beowulf cluster, lets see them now! Or dont they have digital cameras at these fancy affairs.
I want 2D games back.
It's possible that AIX is encumbered by technology that IBM licensed, but I don't know. I'd have to think IBM has ported AIX to x86. Likewise, I wonder why SGI hasn't chosen to port IRIX which has mature support for large scale multiprocessing.
-Kevin
Three reasons:
Recall that SGI is a platform vendor. You buy a package of hardware and software from them. If their software costs are lower, that translates directly into higher margins. Apple made a similar "don't reinvent the wheel" decision in the choices that led to MacOS X, which left them time to focus on their true focus of applications and system usability.
I disagree... if you need a decent server, especially a scalable one, this machine is for you. You can start out small and add stuff as you need it.
Wanna run a database? This machine is perfect. You can add to more CPUs to the system if you need them. You can also add more I/O to the system independently, w/o changing your CPU count unless you need to. For those reasons, this machine would also be great as a general purpose server for running apps of any sort that need server class I/O and CPU power.
Wanna run some crash simulation codes? Again, this machine is perfect. You can run a threaded version or an MPI version and get tremendous performance because of the massive memory bandwidths and low latencies that this system provides.
OTOH, if you wanna run Quake or UT you should probably get a Clawhammer or one of those hyperthreaded P4s.
SGI is a dying company
This machine is the most general purpose machine that SGI has put out in a long time, due to the fact that it runs Linux on Intel. This should mean bigger markets will be available to SGI, hopefully preventing its demise.
He proved me completely and totally wrong. :-D Mod him up, please.
...I had a visit from my friendly SGi representative and he was trying to sell me... this thing after I asked about Linux clusters. I didn't pay too much attention but he was all hush-hush about it, saying that it hadn't been announced yet. It seems impressive. The smallest machine will have 16 nodes and NumaFlex certainly beats the shit out of a Gb Ethernet. He couldn't quote prices, not that I wanted to know...
Non-Linux Penguins ?
I rather agree with u about SGI fscking customers...
but I don't think it will die because there will always be admins who want their data center look like the Entrepise deck...
You can buy this machine and run redhat out of the box. But if you want to kick some ass, you need to install all the tweaked out drivers and modifications SGI made in the kernel. This is the stuff that makes the difference (the secret sauce).
Recent benchmarks showed that SGI's customs modifications gave it almost 2x on throughput performance (versus IBM).
Don't forget that some people don't care how much it cost, they only care about real performance. A linux cluster just won't cut it...
-- Leeeter than leet
SGI has been coding for many years (3-4 at least). It released many code in the public domain such as the (defunct) Apache acceleration program which was simply amazing. Don't forget XFS, one of the best filesystem out there. I feel that for marketing reason, they will not release any "performance tuning" code anymore and keep it for themselve. That's good for customers but not so good for us. In a certain way, such code is probably proprietary to their architecture.
-- Leeeter than leet
Because if you take a kid fresh out of the college, he can run Linux. IRIX? hum not so sure.
Another reason? Oracle. You will see some kick ass performance and that'll bring big sales for SGI I'm quite certain! Oracle doesn't run on IRIX although IRIX on MIPS hardware has one of the best performance ratio in the supercomputing area.
It's not about the cost. Everyone knows very well that the cost of the OS isn't really helping sales. It's how familliar people are. Which one would you buy? an IBM box running AIX or Linux? I bet you don't want to learn AIX and will probably choose Linux if you can get pretty much the same performance..!
-- Leeeter than leet
...there will be a link to an SGI rep discussing what they brought to the show this year. He specifically discusses the Intel-based stuff they're working on.
http://www.sc2002.org/
in qt and real
Aaron
SC'02 webcasting committee
SGI's tweaking allows them to achieve very close to linear scalability.
Just remember when a few months/years ago, Linux wasn't scaling well over 8 cpu's. SGI has a 128 CPU solution already. I have to admit that these guys made a VERY good job!
-- Leeeter than leet
SGI = best throughput/bandwidth, Database = Need throughput, Oracle != IRIX but Oracle = Linux.
Obviously, SGI is in need of money. They know how to do the right thing with performance so this will open some doors that were closed before...
Also, you can bring machines closer to the admins w/o IRIX experience. Some people are scared to learn, you know...
Now remember, SGI offers up to 128 cpu in single system image. This is serious compute power and no one else can offer it. You're not talking beowolf, clusters, etc. It's *1* machine.
-- Leeeter than leet
You want to see images: http://www.sgi.com/features/2002/nov/hpc/images/lg _origin_3900_out.jpg
..that some people DON'T CARE ABOUT THE PRICE! All they want is serious kickass performance. They want the answer NOW!
This is what sets SGI apart, their performance...
Even if the Dodge Viper is a fast car, how come you don't see it racing against F1's...? It's simply because don't care about the price. They don't mind spending $20k on a friggin' steering wheel...!
-- Leeeter than leet
Windows can scale to what? 32 cpu's?
Are they any x86 solutions that offer 32 cpu's systems anyway?
-- Leeeter than leet
It's funny, I talked with an SGI rep there and he said "We haven't annouced this yet, so you don't see it sitting there." To which I replied "Why don't you not tell me about what's not sitting there." "Sure, I won't", he said, and proceeded to tell me all about it. :)
Seriously, it looks pretty sweet, but I was more excited by the Origin 3900 -- 16 processors in one C-Brick (4U).
Here
Maybe someone can describe the bricks?
-- Leeeter than leet
Having the chance to work with an IRIX version, I can tell you that you can actually do TCP over Numa at the speed of 3.2Gpbs so it's more than 3 times.
They also have a faster interconnection that allows 6.4 Gpbs so you can copy the equivalent of a whole CD rom per SECOND(!).
I wish my ISP could do that, unfortunately, my PC is decades behind that kind of performance...
-- Leeeter than leet
The current generation of SGI NUMAflex based machines use a mesh of full duplex 3.2 GByte/sec interconnects. That's 25.6 Gbit/sec.
:)
That's way more than 3 times. Plus the latency is several orders of magnitude less.
The tradeoff is cost. A fully populated rack (32 Itanium2 CPUs or 128 MIPS R1x000 CPUs) starts at $1M can can easily run upwards of $4M. If your task is CPU bound, then a homebrew cluster will be almost as good. If your task is I/O bound, you can't beat the Origin. Until the Cray X1 ships, anyway.
Also keep in mind that while an Origin system can be partitioned, they are typically run as one single image system. The beasts easily expand from 2 CPUs up to 512 (even 1024 with special support from SGI). The cross-system memory latency increases with the larger configurations, but the net cross-section bandwidth/thruput increases linearlly with the CPU count. Very efficent design.
Pretty sweet machine. Again, until the Cray X1 ships!
Well, considering that SGI had a perfectly free choice between using Linux (which already runs on Itanium), porting Irix to the Itanium and trying to make Windows64 (or whatever they (MS) are going to call their Itanium version) work on a NUMA architecture machine, I'd say they probably made a wise choice.
... now, thanks to SGI's and IBM's big-iron work, it does ... the Origin 3900 is a 128 processor monster, and it's my understanding that preformance under the NUMA-ized Linux scales linearly up to 64 processors. I haven't seen any reports on the bigger boxes, but Let's see Win2K server do that ...
SGI has been a BIG booster of Open Source software in general and Linux in particular for about the last 5 years. I was reading the changelogs for the 2.5 kernel tree a couple of nights ago and noticed that their code for running the kernel on NUMA-based machines has already been contributed to the tree and merged.
The major beef about Linux as a big-iron OS has always been "It doesn't scale well above 4 processors"
utter rubbish
Oh
SGI, like Cray, wandered away from doing what made the company special (godawful high performance computing) and both of them suffered the consequences. Cray has moved back to it's roots (cost-be-damned-make-it-FAST one-off boxes for government-sponsored research projects) and has returned to the land of the living. SGI is turning that corner now
The poster continues:
HP probably
and
utter rubbish
You seem to be completely out of your mind.
There are basically three ways to parallelize an application: processes, threads, and sprocs or other lightweight shared-memory process implementations.
Processes for multiprocessing suck. The reasons for this are well known and won't be re-stated here, but cross-reference "shared memory."
Threads are standardized, light-weight, and efficient.
Sprocs are an IRIX-specific thing, but there are similar dohickeys on other OS platforms. A sproc is basically another instance of a process that shares its address space. Sprocs are similar to threads in a lot of ways, but they're not as elegant for many applications and they're highly platform-specific. Programming with this is tough, and porting virtually impossible.
Threads-- and, almost as important, thread performance-- are critically important for application scalability under any operating system, UNIX or otherwise.
I write in my journal
No idea. You can learn more about sprocs by starting with the man page, here.
I write in my journal
1. Raw MHz means nothing. SGI's MIPS-based machines perform excellently, at the top of the Spec2000 benchmarks, and extensively blow away both x86 and Itanium I. There isn't enough data yet to draw any conclusions about Itanium II. I would take an Irix-based Origin system any day over the completely unproven IA64.
Huh? There's data out there, let's look at some of the SPEC results:
CPU specint specfp
600MHz R14k 483 495
800MHz Itanium 314 645
1GHz Itanium II 807 1356
So the MIPS CPU does pretty well considering its low clockspeed, but Itanium II has a much higher peak perf. When you combine that with the monstrous bandwidth of an O3K machine, you get something pretty powerful.
2. Having developed for NUMA architectures, I am confused as to why this machine is designed the way it is. Unless they've done extensive modifications to the kernel, and especially the brain-damaged Linux thread libraries, you're going to end up with what are supposed to be threads of the same process running with different memory access properties.
Linux 2.4 with the O(1) scheduler is pretty good at SMP thread scheduling, but since the latencies accross NUMAflex are so low, process and thread placement aren't as important as they would be on, say, an IBM NUMA box, which is much more 'non-uniform' wrt memory latency. The 2.5 kernel should be even more scalable, as it'll probably include a NUMA-aware scheduler.
3. Even more confusing, what little press there is on this machine claims constant data access to anywhere in the combined memory space. NUMA by definition is non-uniform memory access. What's with that?
I imagine they're referring to the fact that you don't have to use a message passing API to do your app development, i.e. all the memory of the machine is in one big address space. Of course accessing different nodes will result in varying access times, but that's obvious.
There are a large number of reasons going back to stupid internal politics (the offenders have either quit or been fired by now - see Rocket Rick as an example), but it basically came down to two basic reasons: 1) A conscious descision was made at some point (early '90's maybe?) that Irix wouldn't be ported to a radically different (*cough* little endian) architecture and so a port would be really hard (though not impossible) and probably more important 2) Linux for this thing can run any application that is written for IA64/Linux (which presumably will be at least a few) whereas Irix couldn't. You could add a Linux syscall layer, like the BSD's have done, but then you have to track changes to GPL'd software into something that SGI couldn't GPL even if they wanted to (as Irix contains source from some other vendors). So basically, it's a) difficulty of the port from a technical standpoint and b) application capture. Presumably, IBM had similar considerations for using Linux instead of AIX.
Go Badgers! -- #include "std/disclaimer.h"
I would say that AIX definitely does distinguish IBM from its competitors, just like showing up naked for a job interview would distinguish me from my competitors... What's their catchphrase? "It will vaguely remind you of UNIX?"
Lets feed the starving troll, then: The linux 2.6 kernel promises better overall scalability and performance on high-end machines whereas 2.4 is kind of lame in these respects (VM is worse than 2.2 infact) so maybe 2.4 really isn't the best (free) choice for a supercomputer? It remains to be seen exactly how good 2.6 is, coming 2003 to an ftp site near you!
What strikes me even more about the thing is that they actually chose a GPL kernel and not a BSD licence kernel... Maybe the BSD licence isn't as superior as some people would like us to believe.
True. The more relevant question would be: is AIX worth the bother of maintaining and porting? What value does IBM derive from that R&D expenditure. Does any of that serve to distinguish IBM from it's rivals in the marketplace?
IBM (and Sun and SGI) are hardware companies, so you might think that they would prefer a common OS and to compete on hardware. But that strategy has been shown to be disasterous for HP, Gateway, Compaq and even IBM themselves in the PC market. All these companies use their OS to position their hardware and leverage its capabilities to their target market.
Even today, the Linux community are working on the questions "how do we make a free Unix?", and the answer is of course Linux. But from day 1, IBM thought "how do we facilitate people running their data processing applications on our hardware?" and their answer was AIX. Sun thought "how do we facilitate people running their network applications on our hardware?" and their answer was Solaris.
So there is a lot of stuff in proprietary Unix implementations that adds value to them. I don't think anyone (not even IBM) seriously thinks Linux will ever replace AIX, but it does represent an avenue for them to penetrate back into the low-end space. (If you recall, IBM got their asses handed to them on a plate in the commodity x86 marketplace by Compaq and others). SGI have, for whatever reason, decided to make an Intel-based mainframe, and Linux is the way they can do that... but absolutely the last thing they can do is give away any IP that IBM could use to bolster its own Linux offering at their expense.
So help me and moderator's be kind, this is offtopic, but I'm pretty impressed that a Troll... which I am not... posted a message that got a Score:0, Troll was able to stem a conversation like it did. So if getting trolled for starting a 19 reply thread is what should be done I guess I will take it. The question was a honest one and I really did not know the value of Linux as an operating system for a supercomputer. PS. NO I don't want to see Windows 98SE or any Windows for that matter on one, but I just did not know if it would be the greatest choice.
those people who think they know everything are a great annoyance to those of us who do. -isaac asimov
There is more to a big machine than simply which OS gets the most bogomips. Can Linux be partitioned across a Sun 15K like Solaris can? Can you hot-swap processors/harddrives with it? There are lots of questions like these that have to be answered depending on what you intend to do with the machine.
The IBM 390 machines are not fast. People didn't buy them for their MIPS or FLOPS. They bought them because of their stability and uptime. It's less important that OS1 can serve up X amount of web pages per sec and OS2 can serve up 1.1X web pages per sec when what you want is that the machine *doesn't* go down because a CPU board just burned itself out. *That's* what big iron is about.
They also have OS/390, OS/400 and tons of things patented. I think they have they resources to build about anything they want. It probably isn't worth it right now to them to go that route.
While that is true, most are either niche or embedded. Sun is a driving force in SPARC but they don't enjoy the leverage they once did.
- Linux is a lot like Un*x. The type of people who mess around with these type machines are already familiar with Un*x (and many frankly refuse to even try to learn anything else). Stick with what's familiar.
- Un*x type OSs already have time-sharing down (many users on a single machine). While this isn't an issue on a desktop (because everyone has their own), it is an issue when lots of people want to share the $3M machine.
- Un*x type OSs already have remote login capability down (like the previous point). The locality of a machine you are using doesn't matter except for possible latency/bandwidth issues.
- Developers at the respective companies tweak the kernel quite a bit I would imagine. They aren't using the latest RedHat distro ISO they downloaded last night on it.
I've seen a number of MPI distibutions that simply did silly things like
foreach machine in machinesfile
rsh machine command
because it was easy. Of course, these run into problems with many machines (timeouts and takes forever to get started) *but* they wouldn't use anything else because "rsh was standard". This kind of software doesn't work for large systems. The people/companies who support these large systems like this have to solve these types of problems.
The other area of concern is compilers. These type machines have lots of processors but if you have a compiler that produces less-than-efficient code, you are robbing yourself. In a number of cases, GCC isn't going to be your best choice here so you need to find/make good compilers. It doesn't do much good to have a single processor that is capable of 10GFLOPS when your compiler will never be able to get more than 3GFLOPS out of it.
IBM, SGI, and the other companies that are moving to Linux are tackling these types of problems (they have to in order to be successful) and they have their own distributions for their machines.
Well sorta. Notice that there are only two major computer manufacturers that support two major hardware architectures, IBM and HP. Everyone else sticks to a single primary arch. because it just gets too expensive to support any more than that. It makes it a lot easier when there is no overlap between the products as well. Now SGI has Irix/MIPS and Linux/Intel product lines and in many ways they compete with each other (one of the reasons they dropped the workstations, well that and no one was buying them). Now on the high end they've once again introduced product overlap.
Ditto for the OS. Though I guess one saving grace is that the nature of the software for these hpc systems is vastly different and simpler than those for a "general purpose" desktop machine, so it's not like they have to try to support a wide variety of apps across two os's.
You have to remember that SGI is a hardware manufacturer, not an integrator. They don't just pick and choose commodity parts and through a machine together (well ok, their Intel boxes were kinda like that). This is what I mean by schizo.
That's an over-simplification in the case of IBM; their consultancy arm makes more revenue (NOT profit) than the whole of Microsoft. They do hardware too but in recent years that has been seen as a method of increasing demand for their consultancy, which includes bespoke software.
If you recall, IBM got their asses handed to them on a plate in the commodity x86 marketplace by Compaq and others
But what they learnt from that is that allowing someone else to control the OS is madness. Since they can't get everyone else to use AIX it might make more sense for them to switch to Linux (or any Free OS, but Linux is probably the best option for them now) so that no one controls the OS.
I assume that you've heard the old joke "IBM and Microsoft got together to run a goldmine; Microsoft got the gold, IBM got the shaft".
TWW
"Encyclopedia" is to "Wikipedia" what "Library" is to "Some people at a bus stop"
For if there is a sin against life, it consists perhaps not so much in
despairing of life as in hoping for another life and in eluding the
implacable grandeur of this life.
-- Albert Camus
- this post brought to you by the Automated Last Post Generator...