Dispatches From The Geeks

News and Announcements from the MCS Systems Group

Chiba City, R.I.P.

During a cooling emergency in February, we turned off the remaining racks of Chiba City, and they haven’t been turned on since. Power, cooling and space requirements in the machine room necessitate that we bid farewell to our first big cluster. The old web pages can be seen at http://www.mcs.anl.gov/chiba for those of you who are unfamiliar with the beast. The picture links seem lost to the ages, but the video still works.In October of 1999, the whole division got together to assemble the towns that made up Chiba City. Each town had a team consisting of MCSers and a VA linux rep to get the thing put together. We labeled cables, strung them, mounted machines, power distribution units, and network switches into the racks… it was a lot of fun.Each team named their town. I remember my first suggested name for our team’s town being rejected (http://tinyurl.com/29x8he), and I had to go with a backup, Goobies (http://www.roads.gov.nl.ca/cameras/goobies.stm –seriously, we have some great town names in Newfoundland).It took us two days and some banjo accompaniment — a phrase you almost never get to use in the field of Computer Science — but we had our 256 nodes of dual CPU 500MHz Pentium III machines and 32 visualization nodes, gigabit ethernet and Myrinet interconnect, and 3.4 TB of storage up and running. Each node was jam packed with 512MB of RAM and a whopping 9GB of local scratch space. We were hard core cutting edge.Aside from some funky experiments in Windows space, and a proof of concept 8 node testbed, it was our first serious cluster, and it started us down the path that got us Jazz, TeraGrid, and all the other clusters currently crunching numbers in the machine room.On Thursday, June 7, what’s left of it will be leaving our hallowed halls. It’s heading out to pasture to become an educational tool, so it will live and breathe again. Still, I’m going to bow respectfully as I pass by on Thursday. We’ll miss you, Chiba![Edit: Chiba was our first big linux cluster.  The old IBM SP, Quad, had a similar architecture, but was running AIX and was specifically designed to suit that purpose.  Chiba was made up of commodity hardware.]

Written by Craig Stacey

June 5, 2007 at 10:56 pm

Posted in Uncategorized

Tagged with