*BSD News Article 61694


Return to BSD News archive

Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!nntp.coast.net!howland.reston.ans.net!news.sprintlink.net!news.sunquest.com!tony
From: tony@sunquest.com (Tony Jones)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: Linux and FreeBSD:  Differences?
Date: 9 Feb 1996 23:45:21 GMT
Organization: Sunquest Information Systems, Inc.
Lines: 32
Message-ID: <4fgmah$t8s@odin.sunquest.com>
References: <4drq9h$o2m@nnrp1.news.primenet.com> <4eijr5$mrt@venus.os.com> <4f5ge3$m0n@agate.berkeley.edu>
NNTP-Posting-Host: thing.sunquest.com
X-Newsreader: TIN [version 1.2 PL2]

Nicholas C. Weaver (nweaver@madrone.cs.berkeley.edu) wrote:
: 	"A Performance Comparison of UNIX Operating Systems on the
: Pentium"  Kevin Lai and Mary Baker.  USENIX technical conference, jan
: 96.

: 	Nuff said.
Depends, how did you read the conclusion ? That Linux is better ?

For those that were not present, or havn't read the paper ...

That presentation said that FreeBSD was still running approx 2X better at TCP
even after the recent Linux networking mods. (did I also hear at Usenix that
those Linux mods had to be pulled due to panics ??? maybe I'm confused)

Linux has a more optimised trip into top end of the kernel.
In addition, Linux's async metadata writes gave it a disk performance benefit. 
I believe FreeBSD now has or soon will have *optional* async metadata writes
(though the above paper used an earlier release of FreeBSD).

In the end, as I recall the presentation - it was judged a wash.
The factor that tipped the scales for the MosquitoNet project to use Linux was
better support for Linux.

The support metric was based upon the time taken to get a response to
the same question posted to comp.unix.linux.* and comp.unix.freebsd.*.
I don't think they were aware of the existance (and relative role) of
the freebsd-* mailing lists.

My main objection was their NFS benchmarks - why choose the O/S with the
worst UDP/IP performance (Linux) for your NFS server ?

tony