Xref: utzoo comp.unix.xenix:3487 comp.unix.questions:9506 comp.unix.microport:1689 comp.sys.ibm.pc:19833
Path: utzoo!attcan!uunet!seismo!sundc!pitstop!sun!amdcad!ames!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!ico!rcd
From: rcd@ico.ISC.COM (Dick Dunn)
Newsgroups: comp.unix.xenix,comp.unix.questions,comp.unix.microport,comp.sys.ibm.pc
Subject: Re: *nix performance
Summary: think about the disk subsystem too
Message-ID: <9902@ico.ISC.COM>
Date: 30 Sep 88 22:18:35 GMT
References: <1428@draak.cs.vu.nl> <12271@steinmetz.ge.com>
Organization: Interactive Systems Corp, Boulder, CO
Lines: 19

> | One of my main interests is whether a fast (25MHz with cache etc) 386 with
> | some users attached to it can compare itself with a VAX/750 or other minis.
> 
>   Benchmarks (both mine and others) seem to show that a 20MHz 386 is
> about 3:1 faster than an 11/780...

Davidsen (>) mentioned a smart serial card to reduce the interrrupt-per-char 
cost--a good idea, since that's a real CPU-waster.

Also, think very carefully about what sort of disk(s) you're going to put
on the 386 box.  The CPU itself has plenty of horsepower, but if you're
sitting around waiting for a single slow disk waving its head here, there,
and everywhere, all the CPU performance in Silicon Valley won't help.  The
importance of disk latency is more pronounced in a multi-user system
because you're more often servicing multiple requests at very different
places on the disk.
-- 
Dick Dunn      UUCP: {ncar,nbires}!ico!rcd           (303)449-2870
   ...A friend of the devil is a friend of mine.