From owner-freebsd-stable Thu Jul 25 23:56:22 2002 Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.FreeBSD.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 62B5137B400 for ; Thu, 25 Jul 2002 23:56:18 -0700 (PDT) Received: from frontend2.aha.ru (bird.zenon.net [213.189.198.215]) by mx1.FreeBSD.org (Postfix) with ESMTP id 054F443E31 for ; Thu, 25 Jul 2002 23:56:17 -0700 (PDT) (envelope-from uitm@zenon.net) Received: from [195.2.83.132] (HELO backend2.aha.ru) by frontend2.aha.ru (CommuniGate Pro SMTP 4.0b5) with ESMTP id 132944116; Fri, 26 Jul 2002 10:56:15 +0400 Received: from uitm.zenon.net ([195.2.69.86] verified) by backend2.aha.ru (CommuniGate Pro SMTP 4.0b5) with ESMTP id 25697008; Fri, 26 Jul 2002 10:56:15 +0400 From: Andrey Alekseyev Message-Id: <200207260656.g6Q6uBY35654@uitm.zenon.net> Subject: Re: RE: Abominable NFSv3 read performance / FreeBSD server / Solaris client In-Reply-To: <200207252314.g6PNEPUR035516@apollo.backplane.com> from Matthew Dillon at "Jul 25, 2002 04:14:25 pm" To: Matthew Dillon Date: Fri, 26 Jul 2002 10:56:11 +0400 (MSD) Cc: Jaime Bozza , "'Barney Wolff'" , "'Olaf R'" , stable@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL61 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG > Your solution seems reasonable though it will not completely solve the > problem. A permanent solution, yielding maximal streaming, would be > to increase solaris's TCP receive window. With properly sized TCP > buffers I can run NFS over TCP between two FreeBSD boxes at the line > rate (11-12 MBytes/sec) > > test1# mount_nfs -T apollo:/FreeBSD /mnt > test1# nfsiod -n 4 > test1# dd if=test of=/dev/null bs=32k > 2048+0 records in > 2048+0 records out > 67108864 bytes transferred in 5.795125 secs (11580227 bytes/sec) > > > -Matt In turn, the combination of FreeBSD client (4.6 as of mid-June) and Solaris 8 NFS server (latest patches applied) behaves quite well in my recent tests either with TCP or UDP NFSv3 mounts (I just can't see the difference in speed). I can get nearly 11MB/sec reads and about 10MB/sec writes on FE with different block size values to dd. When running multiple dd's (about 20-25) the bandwith gets equally used by all dd processes. I run 6 nfsiod processes on FreeBSD client and I have also increased threads number for nfsd on Solaris). The only settings applied to Solaris tcp stack are: /usr/sbin/ndd -set /dev/tcp tcp_conn_req_max_q0 1024 /usr/sbin/ndd -set /dev/tcp tcp_xmit_hiwat 49152 /usr/sbin/ndd -set /dev/tcp tcp_recv_hiwat 49152 I'm just curious if I'll ever encounter the problem being discussed :) -- Andrey Alekseyev. Zenon N.S.P. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message