Reliability of Beowulf
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.caMon Sep 16 10:44:18 PDT 2002
- Previous message: Reliability of Beowulf
- Next message: Reliability of Beowulf (memory limits)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
> server with no problems. Also what about the 12Gb RAM MSI motherboards? the amount of ram you can stick in a board depends mainly on the chipset, not the label on the board. it's generally not possible to put much more than a few GB ram into desktop-type boards simply because the chipsets don't support more than, say, 6 rows of ram. it's no surprise that board-makers also tend not to put many dimm slots in them, either, considering that the market is still averaging well under 1G/system. but not all chipsets/motherboards are for desktops - Intel and serverworks both make chipsets that scale much higher. so there's no real problem sticking >4G ram in a box if you want. but ia32 is inherently limited to 32b address spaces for each process. but the real answer is: beowulf and clusters in general, do not share address space. at some point, if you want to use lots of ram, you MUST move to a message-passing interface. and that's what most Beowulf clusters use. > > Is this true that Beowulf provides only parallel > > processing jobs but it does not tackle a huge problem > > such a big array requirement/management? We have to > > turn to supercomputer again. the issue is only whether you require shared memory or not. if you do, then yes, you are going to be paying huge bucks.
- Previous message: Reliability of Beowulf
- Next message: Reliability of Beowulf (memory limits)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
