Return to BSD News archive
Path: sserve!manuel!munnari.oz.au!spool.mu.edu!sdd.hp.com!usc!news From: merlin@neuro.usc.edu (merlin) Newsgroups: comp.unix.bsd Subject: Re: Take out fs cache! Message-ID: <l6dphiINN8fk@neuro.usc.edu> Date: 17 Jul 92 15:24:34 GMT Article-I.D.: neuro.l6dphiINN8fk References: <ndadocs@sgi.sgi.com> Sender: merlin@neuro.usc.edu (merlin) Organization: University of Southern California, Los Angeles, CA Lines: 31 NNTP-Posting-Host: neuro.usc.edu In article <ndadocs@sgi.sgi.com> rpw3@rigden.wpd.sgi.com (Rob Warnock) writes: >merlin@neuro.usc.edu (merlin) writes: >If you can stand horrible DOS-like disk write performance, that's probably >the right thing to do. [But if you can, why are you running Unix? ;-} ] Because I'm porting 32 bit codes which rely intensively on availability of bsd unix facilities like system calls, lex, yacc, sockets, X11R5, ipc, rpc, f2c, gcc, tcp/ip, and a whole bunch of other things I won't go into -- but I'm sure you get the point. Degraded write performance from the selection of a configurable write through cache is OK in exchange for the added security it would give to people who need to be able to "pull the plug" without damaging the filesystem. Most of my stuff is "compute" intensive anyway. The only I/O tends to be either greyscale image I/O (mostly about 1 MB files) or input of 3D model databases. I would run my development system with full caching enabled. But I am still thinking it might be nice to be able to turn pff caching for the end user sites with poor quality power and/or nonexistent system manager expertise. Thanks, AJ ------------------------------------------------------------------------------ Alexander-James Annala Principal Investigator Neuroscience Image Analysis Network HEDCO Neuroscience Building, Fifth Floor University of Southern California University Park Los Angeles, CA 90089-2520 ------------------------------------------------------------------------------