From owner-freebsd-smp Fri Oct 11 15:50: 1 2002 Delivered-To: freebsd-smp@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 6834137B401 for ; Fri, 11 Oct 2002 15:50:00 -0700 (PDT) Received: from swordfish.cs.caltech.edu (swordfish.cs.caltech.edu [131.215.44.124]) by mx1.FreeBSD.org (Postfix) with ESMTP id 1662143E6A for ; Fri, 11 Oct 2002 15:50:00 -0700 (PDT) (envelope-from chengjin@cs.caltech.edu) Received: from fast2.cs.caltech.edu (fast2.cs.caltech.edu [131.215.45.55]) by swordfish.cs.caltech.edu (Postfix) with ESMTP id 43D72DF260; Fri, 11 Oct 2002 15:49:51 -0700 (PDT) Received: from localhost (chengjin@localhost) by fast2.cs.caltech.edu (8.11.6/8.9.3) with ESMTP id g9BMnpl31160; Fri, 11 Oct 2002 15:49:51 -0700 X-Authentication-Warning: fast2.cs.caltech.edu: chengjin owned process doing -bs Date: Fri, 11 Oct 2002 15:49:50 -0700 (PDT) From: Cheng Jin To: Cc: Xiaoliang Wei Subject: SMP + dual GbE Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-smp@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org Hi, We are currently working on a high-speed TCP project. We have a couple of machines built on supermicro's P4DP6 board each with two 2.4 Ghz Xeon processors and two GbE cards from SysKonnect. We were able to get close to 1 Gbps throughput(loopback tests) on RedHat Linux 7.3 SMP kernel (2.4.18-3) with interrupt moderation on the GbE cards (5000 interrupts/s) and binding one CPU to each card. However, under FreeBSD SMP, we have only been able to achieve about 640 Mbps on each card. I suppose the binding of CPU to card is implicit (?) under FreeBSD SMP since we didn't notice any out-of-order packets. We tried interrupt moderation going as high as 10,000 per second and as low as 2500, but that didn't seem to help much. Furthermore, we were able to get about 560 Mbps throughput for each card under a single-processor kernel so it seems that the extra processor didn't help much under SMP. We don't know a whole lot about how SMP works under FreeBSD. Actually, we don't know anything about FreeBSD SMP. Any comments/suggestions are greatly appreciated. Cheng Lab # 626 395 8820 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message