Slashdot Mirror


News From Super Computer 99

Donald "I fixed Rob's X Key" Becker wrote in to report from that other conference that is happening right now: Super Computing 99. Reports and picture of the freebies, the parties, and a somewhat surprising extra little sponsor.

2 of 33 comments (clear)

  1. BEER! by cronio · · Score: 5

    Just before the party, Eadline was approached by Tod Needham of Microsoft Research who suggested that he would like Microsoft to help underwrite the party. A quick meeting by the organizers was held on the trade show floor to discuss the various political and community consequences of accepting money from Microsoft until it was pointed out that "We can buy more beer with the money".

    So the secret comes out eh? It turns out Beowulf clusters are run on BEER!
    I've done some research, and it seems that: (Lots of comps) + (Lots of beer) + Beowulf > (Lots of comps) + Beowulf alone.
    The reason for this is that beer was shown to actually do the opposite for computers as it does for men. It INCREASES memory and speed, meaning male computers no longer have an excuse for not remembering that ugly looking female's phone number.

    --


    My plan is to pimp before they realize I'm a jackass. Hit 'em hard and fast.
  2. Beowulf, Beowulf, Beowulf. what about mosix? by Anonymous Coward · · Score: 5

    Hmm, a search of all these comments yields no hits on "mosix". Hmm. Never heard of mosix, eh? Well, take a look at http://www.mosix.org Mosix offers transparent mid-execution migration of processes and dynamic load balancing. The kernel is made aware of the cluster rather than the user processes. I tried it out the other day, (the installation was pretty simple, the only problems I had had nothing to do with mosix) and I have to say it's pretty impressive. I'm by no means an expert, but mosix does seem a lot cooler to me than beowulf, at least from what I've seen and read. Once installed and set up, you can just kind of forget about it and it automatically moves processes around from node to node *while they are running* to balance the load on the cluster, totally transparently. Really cool. You can of course manually direct process migration.