From owner-freebsd-net Wed Feb 26 23: 8:42 2003 Delivered-To: freebsd-net@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 95F7037B401 for ; Wed, 26 Feb 2003 23:08:40 -0800 (PST) Received: from daemon.kr.FreeBSD.org (daemon.kr.freebsd.org [211.176.62.31]) by mx1.FreeBSD.org (Postfix) with ESMTP id 7B37E43F85 for ; Wed, 26 Feb 2003 23:08:39 -0800 (PST) (envelope-from cjh@kr.FreeBSD.org) Received: from gradius.wdb.co.kr (daemon [211.176.62.31]) by daemon.kr.FreeBSD.org (Postfix) with ESMTP id 006868F612 for ; Thu, 27 Feb 2003 16:08:30 +0900 (KST) Received: from localhost (localhost [127.0.0.1]) by gradius.wdb.co.kr (8.12.7/8.12.7) with ESMTP id h1R77w8x027646 for ; Thu, 27 Feb 2003 16:08:00 +0900 (KST) (envelope-from cjh@kr.FreeBSD.org) Date: Thu, 27 Feb 2003 16:07:57 +0900 (KST) Message-Id: <20030227.160757.68128597.cjh@kr.FreeBSD.org> To: net@freebsd.org Subject: Re: Performance tuning hints of gigabit networking? From: CHOI Junho In-Reply-To: <20030226.220551.10329540.cjh@kr.FreeBSD.org> References: <20030226.220551.10329540.cjh@kr.FreeBSD.org> Organization: Korea FreeBSD Users Group X-URL: http://www.kr.FreeBSD.org/~cjh X-Mailer: Mew version 3.2rc1 on Emacs 21.2 / Mule 5.0 (SAKAKI) Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org 2. I noticed other parameters for network tuning is not-so-important in this time. Most problem comes with low kern.ipc.nmbclusters -- I failed to set it over 65536. 3. Usually thttpd use mmap() for caching contents in memory. Our service file(only static files) varies from 10k ~ 300Mbytes. Sometimes thttpd denies request with 500 internal error, resetting mmap() buffers. I patched current unused sendfile() patch for thttpd, it works nicely, but in this case, thttpd process suffers "sfbufa" state, with all requests delayed or even disconnected(total traffic is < 100Mbits then). So another problem comes with kern.ipc.nsfbufs -- I tried to increase arbitrary values but it also cause kernel panic. Final: What is a good math for calculating these values safely? kern.ipc.nmbclusters kern.ipc.nsfbufs From: CHOI Junho Subject: Performance tuning hints of gigabit networking? Date: Wed, 26 Feb 2003 22:05:51 +0900 (KST) > > Hi, > > I am looking for a good resource for kernel tuning on very high > bandwidth HTTP servers(avg 500Mbit/sec, peak 950Mbit/sec). Today I > faced very unusual situation with 950Mbit/sec bandwidth! > > > netstat -m > 16962/93488/262144 mbufs in use (current/peak/max): > 16962 mbufs allocated to data > 16952/65536/65536 mbuf clusters in use (current/peak/max) > 154444 Kbytes allocated to network (14% of mb_map in use) > 512627 requests for memory denied > 2614 requests for memory delayed > 0 calls to protocol drain routines -- CHOI Junho KFUG FreeBSD Project Web Data Bank Key fingerprint = 1369 7374 A45F F41A F3C0 07E3 4A01 C020 E602 60F5 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message