Slashdot Mirror
Cray Co-Founder Joins Microsoft
ergo98 writes "Burton Smith, co-founder and chief scientist at Cray (The Supercomputer Company), has jumped ship. He's joining Microsoft to help them with their clustered computer initiative. Burton joins Microsoft as a technical fellow."
Burton was the co-founder of "Tera", the supercomputer company that purchased the old Cray division away from SGI in their 1999 restructuring.
Tera was founded to develop massively multithreaded machines. After their big purchase, they took the Cray name for continuity with Cray's old customers and products, along with the fact that it's a much more viable "commercial" supercomputing name.
The Cray XD1 uses Opteron processors and runs a variant of SUSE Linux, but uses a custom interconnect. The Cray XT3 uses Opterons and runs Linux on service nodes, and the Catamount lightweight OS on compute nodes. The Cray X1 series has proprietary CPUs, interconnect, and OS. So you're only partly right. Cray does not hesitate to use Linux where it is appropriate. However, when you are doing something like designing your own vector processor from scratch, porting Linux to it just doesn't make sense.
Linux has certainly proven itself to be a winner in lots of HPC computing applications, and Microsoft has a tough uphill battle to fight if they want to break into this market.
You do seem to be implying that Linux-based computers running commodity hardware always makes more sense than using things like proprietary interconnects. It can certainly be more cost effective, but if performance is your main goal (this is "high performance computing" after all), custom-designed hardware like the interconnect on the XT3 is always going to smoke the off-the-shelf stuff which does not exclusively target the high end.
Back when I worked at Cray, one project I worked on was the Fortran 90 compiler. The Fortran 90 compiler was developed on Sun SPARC machines and it cross-compiled to the Cray. Crays, even the mighty C-90 back then, weren't that great interactively, and were pretty slow to compile code. Not to mention the fact that Cray CPU time was far more valuable than the Sun machine's. Pre SGI/Tera Cray machines came in two flavors, the original vector processors (C-90 up to 16 or 32 processors?), and the later massively parallel T3 series (with HUNDREDS of DEC Alpha processors). Both were specialized machines which excel at particular tasks. Wickedly fast at those tasks.
Too many people these days work only on PC architectures, and have no/little exposure to other, superior architectures. The PC was and is designed as a cheap, mass produced general purpose desktop device. It in no way compares to supercomputers, mainframes, or true server architectures. A computing environment is more than the sum of the raw megahertz and bandwidth claims of its disparate parts.
Larry
but whatever it is, it will be interesting. Burton Smith is a very bright guy who pioneered multithreading computing first at Denelcor, and then Tera, which bought Cray from SGI and adopted its name. He is the founder of the company which is today called Cray, but the original Cray company was, of course, founded by Seymore Cray.
Burton always reads broadly and thinks broadly. When designing a supercomputer he deals with every issue, from VLSI technology, Architecture, Operating Systems, and Compilers and Applications. He enthusiastically interacts with many experts, in many areas, and attains a very deep understanding of the issues.
Burton, best of luck at Microsoft.
Jon Solworth
isn't a founder of CRAY. He's a founder of TERA Computer who aquired CRAY in the late 90's. He's a proponent of their multithreadhed architecture - an architecture which has abysmally failed commercially. Since 1988 they've had only one actual cash sale of their system. What this probably means is that CRAY is returning to it's strength of vector supercomputers, such as the CRAY1, CRAY2, XMP, YMP, J90, SV1 and SV2 or possibly massively parallel systems such as the T3E and T3F.
Chris Kuivenhoven is a thief, beware