Date: Thu, 28 Oct 2004 04:56:11 -0700 (PDT) From: Arne "Wörner" <arne_woerner@yahoo.com> To: freebsd-fs@freebsd.org Subject: ffs permormance Message-ID: <20041028115611.96139.qmail@web41203.mail.yahoo.com>
next in thread | raw e-mail | index | archive | help
--0-1005613650-1098964571=:95408 Content-Type: text/plain; charset=us-ascii Content-Id: Content-Disposition: inline Hi! I did some tests on a freshly booted system (FreeBSD 5.2-CURRENT-20040408). I turned off hw.ata.wc via /boot/loader.conf successfully(?). Can somebody tell me, why 'atacontrol(8)' tells me, that "write cache" is enabled (yes)? Sometimes it is disabled (no), but not always. Sometimes atacontrol says, that ad0 has write cache disabled while atacontrol says that ad1 has write cache enabled... Is atacontrol wrong? At least in the beginning big writes to a new file (~64MB) are slow with hw.ata.wc=0 (with hw.ata.wc=1 such writes are fast). The subsequent fsync(1) to that new file did not take much time. But: After a long read (~64MB) from /dev/ad1 the following long write is as fast as with hw.ata.wc=1 before the test (see b.out.bz2). I gathered the sysctl variables (see diff in sc.diff.bz2) Can somebody explain me, why ffs behaves better after a long read? Maybe ffs simulates in kernel memory a hard disc write cache after a long read, so that the writes can be done in a more efficient order? Thank you. Bye Arne __________________________________ Do you Yahoo!? Take Yahoo! Mail with you! Get it on your mobile phone. http://mobile.yahoo.com/maildemo --0-1005613650-1098964571=:95408--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041028115611.96139.qmail>