Date: Mon, 15 Apr 1996 13:15:46 +0200 From: se@zpr.uni-koeln.de (Stefan Esser) To: "Marc G. Fournier" <scrappy@ki.net> Cc: hackers@freebsd.org, nash@mcs.com Subject: Re: Unices are created equal, but... Message-ID: <199604151115.AA15755@Sisyphos> In-Reply-To: "Marc G. Fournier" <scrappy@ki.net> "Re: Unices are created equal, but..." (Apr 14, 18:48)
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 14, 18:48, "Marc G. Fournier" wrote: } Subject: Re: Unices are created equal, but... } On Sun, 14 Apr 1996, Gary Palmer wrote: } } > Alex Nash wrote in message ID } > <199604142039.PAA04761@zen.nash.org>: } > > FreeBSD on UFS: } > > 2770803 bytes/second for writing the file } > > 3908495 bytes/second for reading the file } > > } > > Linux on ext2fs: } > > 3220442 bytes/second for writing the file } > > 1950476 bytes/second for reading the file } > } > ^^^^^^^ } > } > Is that 2nd figure for reading the file right? Seems a bit dubious } > ... unless they're really doing something screwey, you should get } > higher speeds READING than writing ... Linux does account part of the time spent on writes to the reads (i.e. it fails to flush all buffers before the timing for the read test starts). This leads to "good" write and bad read performance being indicated. Bonnie is a by FAR better file system benchmark! } You might think so, but: } } > iozone 22 } } IOZONE: Performance Test of Sequential File I/O -- V1.16 (10/28/92) } By Bill Norcott } } Operating System: POSIX 1003.1-1988 -- using fsync() } } Send comments to: norcott_bill@tandem.com } } IOZONE writes a 22 Megabyte sequential file consisting of } 45056 records which are each 512 bytes in length. } It then reads the file. It prints the bytes-per-second } rate at which the computer can read and write files. } } } Writing the 22 Megabyte file, 'iozone.tmp'...10.703125 seconds } Reading the file...11.023438 seconds } } IOZONE performance measurements: } 2155321 bytes/second for writing the file } 2092693 bytes/second for reading the file } > df . } Filesystem 1K-blocks Used Avail Capacity Mounted on } /dev/sd0s1d 224174 108770 97470 53% /home } } } (ncr0:0:0): "QUANTUM FIREBALL1280S 630C" type 0 fixed SCSI 2 } sd0(ncr0:0:0): Direct-Access } sd0(ncr0:0:0): FAST SCSI-2 100ns (10 Mb/sec) offset 8. } } } Not much of a difference, but still slower to read then to } write. (-stable kernel as of March 26th) Please use Bonnie for these tests (with a filesize of more than twice RAM) and report those numbers. I've seen block read/write numbers of 3MB/s to 5MB/s reported for your drive ... (Or use bigger block sizes for iozone. 512 bytes is highly untypical for real world applications. Try 4KB or 8KB instead.) 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?199604151115.AA15755>