Racks vs. pile of PCs
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Josip Loncaric josip at icase.eduWed Aug 14 10:59:24 PDT 2002
- Previous message: Racks vs. pile of PCs
- Next message: Parallel scripts?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
David Mathog wrote: > > Racks better than piles: > > 1. Space efficiency. > 2. Aesthetics (racks look cool) I'd add another item to this list: Racks are a large assembled unit, much quicker to move and set up somewhere else than shelves and lots of small boxes plus shelves and their numerous wires. Our institute moves from building to building every couple of years. Our neatly wired cluster has to be taken apart then reassembled; this reason alone justifies at least a few percent extra in the price of a rack. If, in addition, we had space problems (which we do not) than racks would win; for now we're using mass produced parts. Regarding reliability: power supply fans indeed fail after about 3 years (we just replaced 4 units). Initially, out-of-spec NICs are also common, but once they are replaced, they rarely go bad. Virtually all other purely electronic parts have been OK for years. However, we've seen a steady stream of IDE drives needing replacement when bad blocks become unmanagable, so much so that diskless compute nodes are starting to look attractive. SCSI drives are more reliable but not bulletproof either (we just lost one that had seen 4 years of heavy use in our server). If local disks can be dispensed with, you'd want to build clusters using space efficient CPU+memory units w/good network capability (and easily replacable fans). If this kind of minimalist module could become an industry standard, a company capable of mass producing them at a good price could sell quite a few... even if they end up costing 20% extra. Sincerely, Josip -- Dr. Josip Loncaric, Research Fellow mailto:josip at icase.edu ICASE, Mail Stop 132C PGP key at http://www.icase.edu./~josip/ NASA Langley Research Center mailto:j.loncaric at larc.nasa.gov Hampton, VA 23681-2199, USA Tel. +1 757 864-2192 Fax +1 757 864-6134
- Previous message: Racks vs. pile of PCs
- Next message: Parallel scripts?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
