From owner-freebsd-current@FreeBSD.ORG Mon Feb 11 17:33:37 2008 Return-Path: Delivered-To: current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id DCDBD16A419 for ; Mon, 11 Feb 2008 17:33:37 +0000 (UTC) (envelope-from me@lexasoft.ru) Received: from mmx.lexasoft.ru (mmx.lexasoft.ru [92.241.160.6]) by mx1.freebsd.org (Postfix) with ESMTP id 8FB2813C448 for ; Mon, 11 Feb 2008 17:33:37 +0000 (UTC) (envelope-from me@lexasoft.ru) Received: from [10.100.0.2] (petrovich-telecom-gw.wahome.ru [77.91.225.38]) by mmx.lexasoft.ru (Postfix) with ESMTPSA id C11DE2842C for ; Mon, 11 Feb 2008 17:33:33 +0300 (MSK) Message-Id: <79D93B0C-589E-41C6-A952-6D573297D9F9@lexasoft.ru> From: Alexey Tarasov To: current@freebsd.org In-Reply-To: <9bbcef730802110921i6af516b9t855a4065884cc64b@mail.gmail.com> Content-Type: text/plain; charset=UTF-8; format=flowed; delsp=yes Content-Transfer-Encoding: quoted-printable Mime-Version: 1.0 (Apple Message framework v915) Date: Mon, 11 Feb 2008 20:33:33 +0300 References: <9DA6FFCD-11DB-4580-9314-52B0885351D8@lexasoft.ru> <0EF3CEE6-6F38-4713-A245-4354D34AFE75@lexasoft.ru> <9bbcef730802110921i6af516b9t855a4065884cc64b@mail.gmail.com> X-Mailer: Apple Mail (2.915) Cc: Subject: Re: Disappointing speed with ZFS X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 11 Feb 2008 17:33:38 -0000 `--# dd if=3D/dev/zero of=3Dtest.tmp bs=3D1024k count=3D10000 load: 0.42 cmd: dd 30773 [zfs:(&tx->tx_quiesce_done_cv)] 0.01u 3.56s =20= 5% 1996k 2612+0 records in 2611+0 records out 2737831936 bytes transferred in 65.497687 secs (41800437 bytes/sec) 10000+0 records in 10000+0 records out 10485760000 bytes transferred in 211.643188 secs (49544519 bytes/sec) `--# dd if=3Dtest.tmp of=3Dtest2.tmp (AD6 is my device) tty ad4 ad5 =20 ad6 cpu tin tout KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s us ni sy =20= in id 1 238 0.00 0 0.00 60.22 18 1.04 64.00 982 61.37 1 0 =20 31 2 67 1 3349 0.00 0 0.00 64.00 7 0.43 58.53 757 43.26 1 0 =20 46 0 53 0 286 0.00 0 0.00 55.81 21 1.13 64.00 487 30.43 1 0 =20 31 1 67 0 269 0.00 0 0.00 64.00 8 0.50 64.00 21 1.31 2 0 =20 49 0 49 0 294 0.00 0 0.00 32.82 34 1.09 63.16 582 35.92 1 0 =20 49 2 48 0 278 0.00 0 0.00 60.00 15 0.86 64.00 983 61.45 0 0 =20 23 0 77 0 280 0.00 0 0.00 56.00 15 0.80 63.64 959 59.60 0 0 =20 23 2 75 0 297 0.00 0 0.00 62.55 11 0.67 47.94 197 9.21 1 0 =20 55 0 44 0 268 0.00 0 0.00 53.80 20 1.05 64.00 481 30.03 1 0 =20 52 0 48 0 290 0.00 0 0.00 57.71 14 0.79 63.23 597 36.89 1 0 =20 54 0 45 0 289 0.00 0 0.00 53.82 11 0.57 64.00 982 61.39 0 0 =20 27 1 71 0 276 0.00 0 0.00 57.18 17 0.94 63.32 963 59.56 0 0 =20 43 1 55 0 296 0.00 0 0.00 61.33 18 1.06 64.00 983 61.44 1 0 =20 37 1 61 0 280 0.00 0 0.00 61.60 14 0.87 63.83 927 57.79 0 0 =20 3 2 94 0 306 16.00 1 0.02 57.82 11 0.62 59.20 486 28.10 1 0 =20 46 0 53 0 251 0.00 0 0.00 62.77 13 0.79 63.17 933 57.57 1 0 =20 35 3 61 0 295 0.00 0 0.00 53.25 16 0.83 42.87 133 5.56 1 0 =20 51 0 48 0 293 0.00 0 0.00 58.00 10 0.57 64.00 501 31.34 2 0 =20 49 0 49 0 265 0.00 0 0.00 38.55 11 0.41 64.00 505 31.57 1 0 =20 33 1 65 0 299 0.00 0 0.00 54.67 18 0.96 62.65 76 4.65 0 0 =20 53 0 47 On 11.02.2008, at 20:21, Ivan Voras wrote: > On 11/02/2008, Alexey Tarasov wrote: >> I've done similar tests on the other machine, and all looks fine. >> >> But why on this machine ZFS works slower than UFS? When I make UFS >> file system on the same disk, rtorrent hashing works 10 times faster. >> And while hashing, HDD is used three times intensively with ZFS >> (noticed by flashing LED). >> >> I have an amd64 Core2Duo processor, 4 Gb of RAM, what is not enough >> for ZFS? >> >> What kernel tuning can help me? > > Ok, so you're saying that you've done similar testing (meaning > rtorrent hashing) on other machines, and they work fine, but only this > one is slow? > > There's no usual reason why would it be so. > > Can you try the suggestion I made (dd+iostat)? -- Alexey Tarasov (\__/) (=3D'.'=3D) E[: | | | | :]=D0=97 (")_(")