Well I hardly ever make Thursday meetings anyway, but a buddy of mine, Matthew Sottile who schooled here at UO, is now working with linux clusters at lanl. I got to spend a few days down in Sante Fe where he's living, and tour the lab where he works -- it's mostly unclassified stuff, twerking with systems that wouldn't otherwise be: they got folks in hardware, kernel land, visuals, and the rest. I got to see 'xed', the alpha-linux testbed. It's made of 104 nodes of Compaq DS-10 (462Mhz, 1GB RAM, DEC Tulip 100baseT) + 4 nodes of Compaq ES40 (4x833Mhz, 16GB RAM, Alteon Gbit) + 16 nodes of Compaq CS-20 (2x833Mhz, 2GB RAM, intel 100baseT) + switches + another ES-40 for the frontend. When I was there, every node was running folding : > well a few were out of commission, but hey... We checked the 'top 100' list while I was there, and they (lanl) had 4 of 'em! The 2 teraflop testbed at #6, #8 is the ASCI Bliue Mtn at 1.6 TF, the 690 Gflop 'Origin' was at #33, and the 344 Gflop 'Wolverine' was at #83... The dual Ghz G4's out now are (max-)rated at ~15 Gflops, but that's using the AltiVec. So anyway, they're working on something called ClusterMatic, which is basically their distro -- they give it away at conferences and shows like super computing... But Matt is coming through Eugene on March 28th, and offered to make an appearance at our Thursday evening meeting if there's interest. He talked about bringing their latest toy/creation, which they call the 'lunch box cluster'. It's more of a small toolchest, but sweet and a great proof of concept. I think it's a protoype for a new weather sampling system... linuxbios is developed there, too. Also: I heard lots of details that grounded me out of some notions I've been entertaining about linux. Fun! So Whaddya think?
http://www.acl.lanl.gov -- -- Ben Barrett Software & Systems Engineer counterclaim Phone: 541.484.9235 Fax: 541.484.9193
