home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.sys.sun.hardware:5349 comp.sys.sun.misc:5097
- Newsgroups: comp.sys.sun.hardware,comp.sys.sun.misc
- Path: sparky!uunet!pipex!warwick!fulcrum!igb
- From: igb@fulcrum.co.uk (Ian G Batten)
- Subject: Re: SparcStation 10 and multiprocessing
- Message-ID: <Bx8JD1.5nu@fulcrum.co.uk>
- Keywords: Sparc10 Multiprocessing
- Sender: news@fulcrum.co.uk
- Organization: Fulcrum Communications
- References: <11336@lochhead.ed.ac.uk> <1992Nov5.023821.27506@ra.msstate.edu>
- Date: Thu, 5 Nov 1992 08:53:25 GMT
- Lines: 13
-
- In article <1992Nov5.023821.27506@ra.msstate.edu> fwp@CC.MsState.Edu (Frank Peters) writes:
- > The multiprocessor support in 4.1.3 is very rough. It turns out that
- > Suns running a lot of little processes don't show as much of a
- > performance boost as you could hope for. That is basically because
-
- Our 4/690-120 running 4.1.2 plus all the obvious patches (Galaxy
- Performance, IPI Jumbo, and so on) acting as an NFS server plus some
- large amount of NFS (30-odd diskless ELCen) behaved worse (not no
- better, but WORSE) when it temporarily had an extra pair of processors
- in it making it a 4/690-140. The interactive use got very bursty.
-
- ian
-
-