Return to BSD News archive
Newsgroups: comp.unix.bsd.freebsd.misc Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!simtel!swidir.switch.ch!scsing.switch.ch!news.belwue.de!news.uni-mannheim.de!thorin!news.gni.net!news.omnilink.net!news.maz.net!news.ppp.net!news.Hanse.DE!wavehh.hanse.de!cracauer From: cracauer@wavehh.hanse.de (Martin Cracauer) Subject: Re: SCSI - throughput Message-ID: <1995Sep21.090850.12239@wavehh.hanse.de> Organization: The Internet References: <43bufs$ekb@nx2.hrz.uni-dortmund.de> Date: Thu, 21 Sep 95 09:08:50 GMT Lines: 65 ab@distler.maschinenbau.uni-dortmund.de (Andreas Braukmann (NetAdmin)) writes: >hardware: i486DX2, 66MHz, 32 MB RAM > EISA (sis), AHA2742AT, > sd0: FUJITSU M2694ES-512 812A > sd1: IBM OEM DFHSS4F 4040 (DHFS 34320 4.3 GB) >os: FreeBSD 2.0.5 (950622-SNAP) >for the DFHS: >bonnie -s 512 >File './Bonnie.295', size 536870912 >[...] > -------Sequential Output-------- ---Sequential Input-- --Random-- > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- > K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU > 1379 98.3 3244 64.0 1448 47.9 1342 98.4 3226 52.3 74.7 9.1 > >for the M2694: >bonnie -s 96 >File './Bonnie.385' size: 100663296 >[...] > -------Sequential Output-------- ---Sequential Input-- --Random-- > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- > K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU > 1367 97.5 1855 28.8 998 34.1 1339 98.1 2669 42.6 57.7 6.8 >hmm, it seems, that the character i/o is constrained by the cpu. >I've switched from an AHA1742A to the 2742AT a couple of months ago. >Unfortunetly I have no exact 1742 based bonnie results for comparison, but >as far as I remember the 1742A(driver) gave better character i/o throughput, >especially it didn't eat that much cpu-cycles. >Have a look at the iozone-results. >(Same machine, 16MB RAM, disk is the Fuji. 2694) >1742: 1973790 bytes/second for writing > 2966137 bytes/second for reading >time: 0.5 u 21.9 s 0:28.72 78,3% >2742: 1760232 bytes/second for writing > 2542905 bytes/second for reading >time: 0.6 u 29.7 s 0:32.52 93,3% >Should I switch back to the 1742? As I need the two SCSI-channels, I would >have to find a second 1742, ... and that does'nt seem to be that easy these >days ... If you find the difference between the iozone numbers for the 1742 and 2742 significant for your needs, maybe you should consider buying a PCI-based 486 board. You should get about 5 MB/sec from your IBM 4.3 GB using an NCR controller, so your EISA equipment seems to limit it anyway. As you need more than one SCSI bus, I think an ASUS SP3G Mainboard (with one onboard NCR) and an additional SC-200 (NCR card) may be cheaper than buying two 1742, as you can reuse your CPU chip. Make sure the SP3G really has a decent enough PCI implementation to run a second NCR controller, many older PCI boards have not. Martin -- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Martin Cracauer <cracauer@wavehh.hanse.de>. No NeXTMail, please. Norderstedt/Hamburg, Germany. Fax +49 40 522 85 36. This is a private address. At (netless) work programming in data analysis.