*BSD News Article 43637


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!bunyip.cc.uq.oz.au!munnari.oz.au!news.hawaii.edu!ames!night.primate.wisc.edu!sdd.hp.com!svc.portal.com!news1.best.com!shell1.best.com!not-for-mail
From: rcarter@best.com (Russell Carter)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: FreeBSD performance
Date: 26 Apr 1995 16:57:06 -0700
Organization: Best Internet Communications, Inc. (info@best.com)
Lines: 48
Message-ID: <3nmml3$6s9@shell1.best.com>
References: <3nc5p0$vj2@rcogate.rco.qc.ca> <3nhcpp$3he@park.uvsc.edu> <25APR95.01988170.0111@VM1.MCGILL.CA> <3nkfg2$hj5@park.uvsc.edu>
NNTP-Posting-Host: shell1.best.com

In article <3nkfg2$hj5@park.uvsc.edu>,
Terry Lambert  <terry@cs.weber.edu> wrote:
>Matthew <BP7K000@MUSICB.MCGILL.CA> wrote:
>] >] (A) Linux 1.1.49 Plato P90      32MB Seagate ST31200N ncr8150 ??
>] >] (B) Linux 1.1.49 ASUS PC/I-P54SP4 P90, 16MB HP C3323SE     ""
>] >] (C) FreeBSD 2.0  Plato P90      32MB Seagate ST32250N BT946C
>
>[ ... ]
>
>] Though not identical the hardware is close enough to raise an
>] eyebrow when the results differ by a factor of 2. Whatever. As
>] was pointed out the shared libs exact a performance penalty in
>] exchange for ease and flexibility of development. That's good
>] enough for me. In a mix of real world applications execl() in
>] one system or another isnt going to have a perceptible effect
>] so I consider the case closed.
>
>I have to disagree.  There are known problems with PCI on Plato
>boards and writeback caching.
>
>The Seagate differences could be substantial in skewing the
>results.  The use of the Bustek controller is extremely biased
>against FreeBSD.
>
>The Use of 1.1.49 for Linux while using a 6 month older 2.0 for
>the FreeBSD is highly skewed as well.  Wait 6 months and compare
>FreeBSD -current of that time with the Linux numbers of today...
>heck, compare the -current code with Linux.
>
>Compare BSD with the disk striping code with Linux -- both machines
>get to use NCR PCI SCSI controllers and 4 Quantum disks that
>support command queueing.  8-).
>
>
>] BUT I still dont know why with 8 MB i was getting 7 MB/sec
>] throughput from the BT-946C/Barracuda and only 3.7 MB/sec
>] with 32 MB !?
>
>It's probably not identifying the card correctly, or you have it
>in a weird mode, and it's bouncing the buffers because it thinks
>it can't DMA from the card with more than 16M in your box.
>

I get about 3.5 MB/s writing my st32550n, and > 6 reading.  It's
not the OS I think, it's the drive.  Have you tried this drive
with Linux?  Setup: TP4-PB-100MHZ, ncr53c810, 32MB memory, 032195-current.

Russell