Date: Mon, 15 Apr 1996 13:59:02 +0200 From: se@zpr.uni-koeln.de (Stefan Esser) To: Gerard Roudier <groudier@iplus.fr> Cc: hackers@freebsd.org, nash@mcs.com Subject: RE: Unices are created equal, but... Message-ID: <199604151159.AA19231@Sisyphos> In-Reply-To: Gerard Roudier <groudier@iplus.fr> "RE: Unices are created equal, but..." (Apr 15, 1:22)
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 15, 1:22, Gerard Roudier wrote: } > I also ran iozone using a 50MB file with 4096 byte records. Terry, } > I think your idea of a FreeBSD upgrade for Linux may be coming } > true... } > } > FreeBSD on UFS: } > 2770803 bytes/second for writing the file 18.1s/50MB } > 3908495 bytes/second for reading the file 12.8s/50MB } > Linux on ext2fs: } > 3220442 bytes/second for writing the file 15.5s/50MB } > 1950476 bytes/second for reading the file 25.6s/50MB } > FreeBSD on ext2fs: } > 2768517 bytes/second for writing the file 18.1s/50MB } > 3265638 bytes/second for reading the file 15.3s/50MB } It seems that something is magic for Linux in my system: No! Actually it's not ... :) } P90/16MB/NCR53c810/IBMS12 } FreeBSD-2.0.5 installed in notche 0. (should be faster) } Linux-1.3.87 installed near notche 1. } } iozone 100 4096 } } Linux-1.3.87 / 1k ext2 file system } 4257312 bytes/second for writing the file 23.5s/100MB } 3924311 bytes/second for reading the file 25.5s/100MB } FreeBSD-2.0.5 / file system created while installation } probably some UFS 8k group ??? } 3485269 bytes/second for writing the file 28.7s/100MB } 4438416 bytes/second for readind the file 22.5s/100MB } I have not time enough to write my comments in english. } It is your turn to take some dictionnary if you want to understand } my comments. Sorry, my french is lacking. How about German or Japanese, if you don't want to write in English ? ;-) But regarding those iozone results: According to some discussion in a Linux group, the write performance comes out to high and the read performance to low, since there is no flush() at the end of the write phase. Since some of the write time is accounted to the read phase, both values come out wrong, but the total time taken should be about right ... (Lets see: 23.5+25.5=49, 28.7+22.5=51.2. Hmmm ... The numbers from the first test with 50MB files are 30.9 / 41.1 / 33.4, which indicates that EXT2FS under FreeBSD -current is faster than under Linux, and only slightly slower than UFS ... :) Please don't take this comment too serious. I consider IOZONE numbers highly bogus and only accept Bonnie with large files for comparison purposes ...) } FRENCH COMMENTS: } ---------------- } He l'ami, [ Deleted everything I couldn't understand :) ] Regards, STefan -- Stefan Esser, Zentrum fuer Paralleles Rechnen Tel: +49 221 4706021 Universitaet zu Koeln, Weyertal 80, 50931 Koeln FAX: +49 221 4705160 ============================================================================== http://www.zpr.uni-koeln.de/~se <se@ZPR.Uni-Koeln.DE>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604151159.AA19231>