Date: Wed, 19 Jul 1995 13:43:55 +0100 (BST) From: Doug Rabson <dfr@render.com> To: Bruce Evans <bde@zeta.org.au> Cc: esser@zpr.uni-koeln.de, current@freebsd.org Subject: Re: slow nfsv3 writes Message-ID: <Pine.BSF.3.91.950719133900.22916A-100000@minnow.render.com> In-Reply-To: <199507182123.HAA25447@godzilla.zeta.org.au>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 19 Jul 1995, Bruce Evans wrote: > >This is NFSv2, a 486DX2/66 (16MB RAM, DEC 21040 PCI Ethernet), > >the server is a SparcServer 1000 running Solaris 2.4, measured > >when the NFSv3 patches had just been integrated: > > And for -current on a slow 486DX2/66 (16MB RAM, _loopback_): > > > -------Sequential Output-------- ---Sequential Input-- --Random-- > > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- > >Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU > >SP3G 20 577 35.3 579 4.5 356 7.5 874 58.6 985 10.4 53.4 16.1 > >Noname 1 111 8.8 132 3.8 140 7.1 964 76.8 9417 99.3 175.8 26.8 > > All combinations of interface, benchmark and benchmark block size that I tried > give about 100K/sec writes and 900K/sec reads. The disk on the server is on > constantly. This is fixed by a recent change to vfs_bio. The buffer cache was getting confused and was forcing synchronous writes for blocksizes smaller than 8k. I now get 400k/sec writes to my sgi with all blocksizes. I managed to hang my machine a couple of times testing a loopback mount. All the nfsds were blocked on nfsrcvlk. Any ideas? -- Doug Rabson, Microsoft RenderMorphics Ltd. Mail: dfr@render.com Phone: +44 171 251 4411 FAX: +44 171 251 0939
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.91.950719133900.22916A-100000>