[Beowulf] OS for 64 bit AMD
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.caSun Apr 3 09:46:39 PDT 2005
- Previous message: [Beowulf] To run GAMESS parallel on Beowulf with LAM-MPI
- Next message: [O-MPI users] Re: [Beowulf] Alternative to MPI ABI
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
> (2) Reducing the finger pointing loop when things go wrong in complex IT > configurations. This has only happened to me once or twice but as an this is a great argument if you've actually got a cluster which has a single, fixed use. my world is a completely general HPC provider, and we certainly can't pander to some single application's bug-compatibility with a particular distro. this is simply part of the cost that an end-user must bear when choosing to invest in a commercial package. I think such users all too frequently overlook this: yes, they get something called "support", but that support drasticallly limits where they can run the app. > example: If I have to connect a cluster I/O node to an enterprise SAN > fabric I'd rather my linux host OS be something that the SAN/FC switch > vendor officially certifies and has qualified. The cost of the FC is one of my favorate peeves - a whole industry where the standard is so low that none of them actually interoperate. put commodity disks in company X's SAN shelf! heavens to betsy! that's unsupported... contrast this to a *real*, honest standard like ethernet, where everyone really does expect nics/switches/packets/udp/tcp even nfs/cifs to just play nice together. yes, there are occasional problems with some obscure corner like Cisco 10/100 autonegotiation - but Cisco is an FC-esque vendor... vendors love lock-in. ignorant or naive users don't notice lock-in. certain categories of admins love lock-in because of the job security that accrues from managing crappy fragile stuff. but as the computer industry matures, more people are realizing that lock-in is the antithesis to standardization. my only fear is that the legal/political profession will continue riding the fallacy of "intellectual property". that's a great analogy, that IP is like a huge, ongoing traffic disaster, with the lawyers and politicians chasing the ambulances.
- Previous message: [Beowulf] To run GAMESS parallel on Beowulf with LAM-MPI
- Next message: [O-MPI users] Re: [Beowulf] Alternative to MPI ABI
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
