Slashdot Mirror
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."
Does someone have a source that tells what OS these things run? I'm not seeing it in either article.
-- Moderation in all things, exceptions to all rules --
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!
I'm sure Google is up there in terms of size, but this list is about performance. Google's computers do searching which is different in terms of processing requirements than raw number crunching like simulating a nuclear explosion. Google also probably wouldn't release any specifications anyway.
Well, there's spam egg sausage and spam, that's not got much spam in it.
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?
Religion for nerds. Stuff that really matters
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.
Summary: It's not all Linux.
As far as the Top500 is concerned, I think the answer is yes. (And this explains why SETI@HOME does not count as a supercomputer for the purpose of compiling the Top500 list.)
In fact, here is the most relevent passage from the article:
Well yes, but they could easilly go for other free options other then Linux. FreeBSD, for example.
...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
Wikipedia: AIX Operating System
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.
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/ (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/ 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.
the more accurate the calculations became, the more the concepts tended to vanish into thin air. R. S. Mulliken
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 for a wealth of information.