Slashdot Mirror
Linux Chosen for IBM's New Supercomputer
Uhh_Duh writes "news.com is reporting that Linux will be the main OS in the Blue Gene - IBM's $100m supercomputer project. The Blue Gene will contain 65,000 processors and 16 trillion bytes of memory." Wow. That's a lot of nuclear weapons simulations.
Anyone else notice this?
The decision to adopt Linux came, in part, as a result of the growing size and strength of the open-source community. Thousands of developers around the world are participating in the evolution of Linux. Creating a new OS inside of IBM would require a massive engineering effort.
followed by
We chose Linux because it's open and....saw considerable advantage in using an operating system supported by the open-source community, so that we can get their input and feedback."
So, basically, IBM doesn't want to design their own proprietary system (smart) and plans to use the resources currently available. (also smart)
They want open-source to get them rich, right? Less initial cost by the company, etc etc. What are the odds they'll profit-share with people they're getting rich off of? (well, ok, attempting)
-- El Sacarino tiene gusto de la chocha
In 5 years, there will be only Linux, BSD and Solaris - with BSD and Solaris being binary and source compatible to Linux.
Linux has reunited Unix, this is a good thing because it didn't happen by monopilzation from one company. There is lot of diversity within Linux (lots of different vendors and supporters) but it's all compatible.
Building a computer, to tell you how to build another, larger, more complex computer. Hrmmm..
Uh, that's how it works in general. Or did you think modern CPUs were laid out by hand?
Continuing on that theme, it's written (at least next to the Apple 1 and Cray machines at the Science Museum in London) that Seymour Cray used an Apple to design his super computers while Apple used a Cray to simulate one of their designs.
errr. I thought this beast had ~65k processors?
32*64*8*64 = 1048576.
"The new wave is not value-added; it's garbage-subtracted" - Esther Dyson, Dec 1994
That's a curious number. Because it's about the amount of memory needed to perform the matrix operation involved in using the Number Field Sieve to factor a 1024-bit number. It would still take a (long) while to do, but given enough time, this machine could do it.
The unedited copy/paste from the article goes like this:
The chip itself will extend an IBM design philosophy that will emerge in coming years with IBM's Power4 processor. That processor will package four CPUs on a single chip, IBM has said.
Blue Gene will use 32 CPUs in a single chip, Goyal said.
So, is that a one-million processors machine they're really building, a 32768 chips machine or what ? ZDnet reporters are on crack today :)
I wish someone found an article from IBM's PR site with more details and less confusion.
The one with ~1 million cpu:s is the final Blue Gene. The one which this article talks about with 65000 cpu:s is Blue Gene/L, which is supposed to be a prototype of the final design.
AIX does not suck, it just costs a lot of cash to maintain (someone posted $10 billion over its life). If IBM can use the opensource community to help defray some of their development costs then it is wise of them to do so. IBM does sell NT, but NT is not an OS you will ever run on IBM mainframe or SP class hardware. Try as MS might they are a small and mid sized server OS vendor. You are correct though that IBM could just has easily used the BSD's but they just never got the critical mass of developers that linux has achieved in the last couple of years.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
YEAH! :P Heh heh. I admin AIX and it is the best UNIX I have ever worked with. You can do everything from the command line, thru smit, or for the really bad off WSM. SMIT, while it's a CURSES based (Does IBM use NCURSES??) admin tool, you can do almost everything from it. Because you usually buy hardware from IBM, everything just works, or you have diagnostic info to tell you it isn't working including LED codes that tell you why you won't boot (Corrupted BLV, JFS Volume, Bad superblocks....it's all there). In any case, AIX is here to stay and just because IBM chooses to use Linux on their super computer means nothing. It may mean that the government wanted Linux because with THAT many nodes, your AIX support bill would be outrageous! :) That and it maybe the Beowulf stuff just works better then the AIX SP stuff. I wonder if this is using the new Power4 blade servers?
Gorkman
If SPARC is to survive, someone outside Sun will have to make it so. After the development of the original UltraSPARC, many (most?) of the talent that made it happen went bye-bye. There's been an ongoing brain drain from the design groups since then. Some people who used to be thought of as a waste of air are now considered top contributors.
In the meantime, successor projects (to UltraSPARC) have spent too much time redesigning and precious little time getting a competitive product out the door.
The performance of the software running on my server farm's fastest Intel/AMD machines is far superior to the performance of the same products running on the fastest SPARC boxes. On the other hand, every SPARC box we've ever purchased is still running in some capacity. I can't say that for the PC-platform servers.
I'd like to see Sun get its in-house design process straightened away and become competitive again. But somebody high up is going to have to take ownership of that process and make some major changes if it's going to happen, IMO. And since things have languished this long, it's hard to figure how somebody's going to wake up at this late date and put full effort into fixing what's gone wrong. I sure hope it happens, for some of the same reasons you shared!