Linux Claims 4 of the Top 5 Supercomputer Spots

Anonymous Coward writes to tell us that the November 2005 list of supercomputers has been published. Certainly something to note is that four of the top five use linux. Relatedly Multiflow writes "CNET is reporting that the number of supercomputers on the Top500 list which use Intel Itanium 2 microprocessors has fallen by almost 50% in the past year. While new higher performance Itanium chips are in the pipeline, the article reports that 64 bit Xeons and Opterons have increased their representation on Top500."

niche market?

[OffTheLip (636691)]

It may be a niche market but what a market it is. Rock on Linux!

Re:niche market?

[miffo.swe (547642)]

I use it as a desktop and find it very usable. Thats enough for me. Linux has never been about snagging market share wich is something many Windows jockeys has a hard time understanding. Linux success doesnt stand and fall with the number of users. If it stays at 10% so what? There should really be 10 different OS out there competing and 10% of that is pretty good.

Re:Oh but they are

[Grishnakh (216268)]

There's probably at least two camps of Linux users, based on their belief on whether marketshare is important or not. I'm somewhere in the middle.

On the one extreme is people who don't give a rat's ass if other people use it or not, or at least say this. On the other extreme is people who want global domination.

The problem with the first extreme is that for the whole open-source concept to work, you have to have a critical mass of users so that you have a large enough base of developers (which are a subset of the users) to keep the project working. Since most open-source software isn't owned by corporations (though some is certainly supported to some extent by them), open-source requires a large number of users to help work on the various projects.

The problem with the other extreme is that, in order to make Linux (as a group of distributions) a viable choice for all current Windows users, certain concessions and changes might have to be made, such as providing an API for closed-source drivers, removing features from the most popular software because it's "too flexible" for many users, standardizing on one desktop (GNOME or KDE), etc. Many powerful people in the open-source community don't like these things, and it's quite debatable whether they might end up hurting or destroying the open-source movement instead of helping it. (For instance, if it became easy to distribute closed-source drivers, then while Linux might become more popular initially, it might suffer from the same problem as Windows where companies release crappy drivers for their hardware, which makes Linux systems unreliable, and the companies refuse to help any open-source driver efforts).

Personally, I don't want Linux to become a commercialized, closed-source OS with a few open-source bits, but all the important stuff closed as some companies are trying to do. I also don't care if "Aunt Tillie" uses it, as long as she doesn't ask me for free support for her Windows computer, so I don't really care about it becoming the dominant OS. What I do want is for it to gain enough marketshare so that it's taken seriously, most hardware is supported on it (by open-source drivers), most worthwhile application software is ported to it, and that there's enough business in it that the dominant distros can make very polished versions without any major shortcomings like we still see today. In a nutshell, I want to be able to use Linux at work and at home to do anything I need to do (including buying and using the latest TurboTax or AutoCAD, for instance), without ever running into any major problems because I don't use Windows. If Linux reached 50% marketshare, this dream would probably be realized. The Windows users could happily live with their BSODs, activations, client-access licenses, high license prices, etc., and me and the other Linux users could happily ignore all that crap without being hindered because some web site is "optimized" for IE, TurboTax doesn't have a Linux version, ATI cards have crap drivers for Linux, etc.

Already, we're getting fairly close: certain types of hardware still have serious driver problems (video cards and WLAN adaptors), most lower-end commercial software does not have a Linux version (although much high-end software, such as that by MentorGraphics, Cadence, etc., does), and we still have serious problems with non-HTML-compliant websites. But on the plus side, we have a very reliable kernel and OS, we have very functional desktop environments (GNOME and KDE, and apps from one will work in the other), we have tons of free software to satisfy most of your needs both on the server and the desktop, and we have tons of drivers for most popular and also much older and obscure hardware. We're at the point now where you can get a recent Linux distro and install it, easily and quickly, on the hardware of your choice, and probably not run into any problems at all. You'll get tons of included software (web browsers, CD burners, word processors, etc.), and be able to do just about anything you reasonably need to do with a computer, unless perhaps your raison d'etre is to play all the latest 3D PC games.

Re:niche market?

[bersl2 (689221)]

Ah, but most OSes (with the notable exception of you-know-who's) adhere to certain standards, such as POSIX. By standardizing what can be standardized, and by carefully abstracting, it becomes easier to develop for a wider range of OSes.

Anyway, user share only matters (for us) to the extent that we do not want to be excluded from doing something simply because we haven't enough users to be relevant. The actual number doesn't matter, only the effect that number has on consideration of our OS as a "first-class citizen".

Re:niche market?

[NetRAVEN5000 (905777)]

"You're obviously too young to remember the OS wars... C64, IMB-compatible, Apple (Mac), Amiga, etc. It used to be a real nightmare to buy and even more of a pain to develop software. As someone who was a geek during those days, I can say that things in the world of PC's are MUCH better today than they were when we actually had a lot of OS competition."

First of all, this is not the "OS wars" so much as the machine wars - these are all different machines, even if they do use different OSes - the only reason they have different OSes is because their OSes were written specifically with that machine in mind. This is no different today - you still can't run MS Windows on a Mac.

Second of all, each of these OSes had their own strengths and weaknesses, right?

And third of all, as long as they follow standards, there'd be no problem. I have yet to hear anyone say that they have trouble switching between Opera and Firefox. In fact many of the problems we have with computers today such as vendor lock-in and version incompatibilities are partially due to *certain companies* (AKA MS) not following standards.

well duh

[scenestar (828656)]

These aren't off the shelf desktops.

What else would you expect them to run, windows ME?

Re:well duh

[Eightyford (893696)]

These aren't off the shelf desktops. What else would you expect them to run, windows ME?

HP-UX, IRIX, Solaris, SCO UNIX, Mac OS X, free/open/netBSD...? Palm OS?

Less duh, more history

[fm6 (162816)]

The alternative is Unix, which is what most supercomputers used to run. Or more precisely, they ran proprietary OSs that had started out as ports of Unix to their particular hardware. Then in the late 90s everybody realized that they couldn't afford to keep developing their own processors, and started shifting to commodity processors, such as Itanium. Rather than go to the expense of porting their own OSs to the new processors, they just adopted Linux. A commodity OS for a commodity processor, if you will.

I was working at SGI in 1999 when they made their Itanium/Linux move. A lot of customers (and employees for that matter) would have liked SGI to port its version of Unix, Irix, to the Itanium. But that was just too expensive. Instead, SGI promised to continue selling the MIPS/Irix Origin line, in addition to the Linux/Itanium Altix line. So Irix is still alive — as a legacy system. If you check the Top 500 list you'll find several Altix systems but not a single Origin system.

Hooray for Linux!

[Elrac (314784)]

While this will not do much to encourage the Unwashed Masses to embrace Linux, it certainly shows that Linux is a serious operating system suited to high-powered computing (or at least to hosting high-powered computing applications). I hope at least a few Fortune 500 CIOs will take notice.

Re:Hooray for Linux!

[Decaff (42676)]

While this will not do much to encourage the Unwashed Masses to embrace Linux, it certainly shows that Linux is a serious operating system suited to high-powered computing (or at least to hosting high-powered computing applications). I hope at least a few Fortune 500 CIOs will take notice.

Actually, it doesn't show that at all. Supercomputing is a very specialised niche use of hardware. Generally, this sort of software wants the operating system to get out of the way as much as possible and allow the fastest possible access to memory and processors and (depending on the situation) I/O systems. In the past major supercomputer applications have required very little operating system functionality to back them up.

There is little comparison between specialised numerical supercomputing and general multi-processor mainframe use, which requires concurrent multiuser access to app servers, general filesystems, databases etc. This is where older OSes such as IBM operating systems and Solaris work very well, and where Linux is now making inroads.

It is rather like comparing a formula one racing car to a truck. I agree that Linux is suited to both purposes, but working well in one environment does not indicate usefulness in another.

Itanic, we hardly knew ye

[drinkypoo (153816)]

Itanium already has no popularity whatsoever. If it can't even be successful in the supercomputer market, it can't succeed anywhere (last I looked, itanium had truly awe-inspiring FP but was slow at everything else.)

MY HEART WILL GO ONNNNNNNN!

linux? Not exactly.

[daknapp (156051)]

Where, exactly, did you get the information that these systems "run linux?"

In the Blue/Gene system, for example, the user front-end nodes use linux, but the OS for the system itself is very definitely NOT linux. So acting as if the system runs off a linux kernel is misleading, to say the very least!

Re:linux? Not exactly.

[TFMReader (225915)]

The compute nodes of Blue Gene/L do not run anything that may be called an OS. They basically run a single application thread per processor, and they do not do any sophisticated system work at all (no context switches, the single user process has access to all the memory, etc). The system tasks are concentrated in the so-called I/O-nodes, and those run Linux. So all the system-related things there are Linux indeed. See this paper [ibm.com], for instance.

Note that I/O nodes and not "front-end" nodes. All the front-end machines (there are many) run Linux as well.

All the user-level stuff (the programming model, tools, compilers, etc) is standard Linux, too.

So, is it Linux?

[Disclaimer: I have worked on some system aspects of the beast, but this post is not sanctioned by BG/L team or IBM or LLNL. I am not disclosing anything proprietary here - all this is open info that can be found in many papers on the subject. Check out IBM Journal of R&D [ibm.com] for a wealth of information.

Re:linux? Not exactly.

[Wesley Felter (138342)]

May I suggest the following paper:

J. E. Moreira, G. Almási, C. Archer, R. Bellofatto, P. Bergner, J. R. Brunheroto, M. Brutman, J. G. Castaños, P. G. Crumley, M. Gupta, T. Inglett, D. Lieber, D. Limpert, P. McCarthy, M. Megerian, M. Mendell, M. Mundy, D. Reed, R. K. Sahoo, A. Sanomiya, R. Shok, B. Smith, and G. G. Stewart: Blue Gene/L programming and operating environment. [ibm.com]

Summary: It's not all Linux.

Re:linux? Not exactly.

[daknapp (156051)]

In fact, here is the most relevent passage from the article:

The computational core of 65,536 compute nodes is partitioned into 1,024 logical processing sets, called psets. Each pset consists of one I/O node running Linux and 64 compute nodes running a custom compute node kernel(CNK).

the scoop

[SebNukem (188921)]

Looking at this chart http://www.top500.org/lists/2005/11/l/Operating_Sy stem [top500.org] it actually appears that the OS ran on all system are:
- Linux: 72.2%
- Max OS: 1.0%
- Others 4.4%
- UNIX and Linux: everything else (~22%)

So it appears that Linux/UNIX* runs on about 95% of all super computers. The Story headline should have been:
Linux Claims Almost All Supercomputers Spots

What a scoop.

*Linux,UNIX, what's the difference really?

Why no Itanic

[Groo Wanderer (180806)]

If you want a good explanation for the Itanic drop-off, look to the funding side of things. Follow the money, and all will be explained. Read a lot into this.

                  -Charlie

In case you were wondering...

[vectorian798 (792613)]

...the one in the top 5 that is not running Linux is ASCI Purple, and it is running AIX. In case you haven't heard of it, AIX is a version of Unix developed by IBM:

IBM AIX 5L [ibm.com]
Wikipedia: AIX Operating System [wikipedia.org]

Japanese SUPER HYPER MEGA EFFICIENT ENGINEERING

[garrett714 (841216)]

I find it funny that the US's challenger to the Earth Simulator came out 3 years later, used almost twice as many processors, and only has a slight performance advantage.

6) Sandia National Laboratories
United States Red Storm Cray XT3, 2.0 GHz
Cray Inc. #Processors: 10880 Year: 2005 Rmax: 36190 Rpeak: 43520

7) The Earth Simulator Center
Japan Earth-Simulator
NEC #Processors: 5120 Year: 2002 Rmax: 35860 Rpeak: 40960

Linux Claims 4 of the Top 5 Supercomputer Spots

[by Anonymous Coward]

Well done linux, i yearn for the day when the headlines read

Female Linux Users Claim 4 of the Top 5 Supermodel Finalists

Re:One Supercomputer?

[Wesley Felter (138342)]

The definition of "supercomputer" these days seems to be "a collection of hardware that can run an MPI job". So BlueGene/L is a cluster of 64K computers, but it counts as one supercomputer.

Re:One Supercomputer?

[Kadin2048 (468275)]

Fairly good question. I'm not sure where you start calling something a "computer" and where you fall off into the grey area of "computational network" or "cluster" or "grid computing system." After all, isn't SETI@Home a pretty massive computer? By some (very loose) definition it should be.

I think most people consider a computer to be something that, at some level, runs a single operating system (which then can abstract other OSes on top of itself), or perhaps is capable of addressing a single logical range of main memory (although this might not be a good definition either).

I haven't read the article yet to see if they give their definition, but it does seem as if the line between 'this is a computer' and 'this is a bunch of computers working together' is fairly blurry, and perhaps where one draws it is completely arbitrary.

Re:Source?

[jbellows_20 (913680)]

Well, if Sony hadn't went off and got caught they would have the largest supercomputer and it would be running Microsft Windows!

Re:*yawn*

[FooAtWFU (699187)]

Oh, give them some credit. There's a lot of effort and design that goes into organizing such a big mess of supercomputing, keeping everything streamlined, keeping processes on thousands of different processors talking to each other, deciding what to do if one processor decides to fail, et cetera et cetera. There is real work and real innovation present- perhaps not as glamorous or even as useful as faster general-purpose microprocessor cores, but don't sell them short, either.

Re:*yawn* (Ahem...)

[Frumious Wombat (845680)]

If you look up how the top500 benchmark, and most of the others, slapping together a heap of boxes doesn't get you anything. To actually get a decent score on parallel DP linpack, or simulation codes used as benchmarks, you need a fast, very low latency interconnect between the nodes, excellent synchronization, and fast disk access.

Even the allegedly "off the shelf" systems contain an awful lot of not off the shelf hardware. Case in point would be PNNL's Itanium cluster http://www.emsl.pnl.gov/capabs/mscf.shtml/ [pnl.gov] (at 1000 or so nodes). At SC2003 I chatted with people I know from there, and they mentioned that they had four (4) Quadrics http://www.quadrics.com/ [quadrics.com] interconnect cards Per Node, plus extra switches, in order to get the bandwidth up high enough. Even a cheap cluster will add Myrinet (at about $1500/node when the switch is factored in), and start worrying about topology after the first few dozen nodes are installed.

There are clusters (basically networks of workstations), and then there are supercomputers.