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.
Mark Hahn hahn at physics.mcmaster.caSat Aug 17 10:24:38 PDT 2002
- Previous message: Racks vs. pile of PCs
- Next message: Beowulf List archives.
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
> One advantage that comes to mind, although it may not be worth $6000, is > when you have 20 PC's stacked, according to Murphy, the bottom one will fail > first. The rack mounted version will be much easier to service a failed unit > than the stack version. wire shelving units cost very little and often have an option for wheels. I think the choice really boils down to the number of nodes you want. chances are good that your floorspace is limited, therefore density is an issue. we want O(150) CPUs and can only afford ~2 more racks of floorspace in our machineroom, so we need roughly .5U per CPU. further, with more than, say, 64 CPUs, you have to think seriously about hardware failures. that, in turn, biases you towards higher-end hardware, since a generic ATX power supply is definitely not in the same MTBF range as a good rackmount one. still, Dell/Compaq all make small uniprocesor desktops; for instance: http://www.dell.com/us/en/biz/products/model_optix_optix_gx260.htm 3.6*12.5*13.9=625.5 cuin, versus 19*21*1.75=698.25 cuin for 1U. wow, you can actually get 2.26GHz P4, 533MHz FSB, 512M ddr, 1000bT all for under $1k. 3-year warrantee.
- Previous message: Racks vs. pile of PCs
- Next message: Beowulf List archives.
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
