*BSD News Article 20332


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!elroy.jpl.nasa.gov!swrinde!cs.utexas.edu!not-for-mail
From: swanton@moose.usmcs.maine.edu (George P. Swanton)
Newsgroups: comp.os.386bsd.misc
Subject: NetBSD IO performance
Date: 2 Sep 1993 21:43:03 -0500
Organization: UTexas Mail-to-News Gateway
Lines: 34
Sender: daemon@cs.utexas.edu
Message-ID: <9309030243.AA03945@moose.usmcs.maine.edu>
NNTP-Posting-Host: cs.utexas.edu

A while back there was some file-system performance discussion and
a suggestion that someone try iozone on their system. I gave it a try
today with some odd results.

for reference, a Decstation 5000/120 (Ultrix) w/ SomeFlavor Seagate SCSI 1G

Writing the 30 Megabyte file, 'iozone.tmp'...51.027342 seconds
Reading the file...31.406224 seconds

IOZONE performance measurements:
	 616478 bytes/second for writing the file
	1001625 bytes/second for reading the file

meanwhile, turning in fair read rates but sucking the proverbial
bog water on writes for some unknown reason, my 486DX/33 w/ Seimens MegaFile
(who dreams up these names?) ~600M SCSI on 1542B on ISA running NetBSD-0.8

[iozone autoconfigured to Posix]

Writing the 30 Megabyte file, 'iozone.tmp'...167.950000 seconds
Reading the file...46.033333 seconds

IOZONE performance measurements:
	187301 bytes/second for writing the file
	683358 bytes/second for reading the file

So what's up? Could be iozone is confused (doesn't seem likely)
How come such poor write performance?
 
Comments? Suggestions?
gps
swanton@moose.usmcs.maine.edu