From owner-freebsd-questions@FreeBSD.ORG Fri Sep 17 20:56:44 2004 Return-Path: Delivered-To: freebsd-questions@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id E3CD816A4CE for ; Fri, 17 Sep 2004 20:56:44 +0000 (GMT) Received: from fincher.users.accretive-networks.net (fincher.users.accretive-networks.net [216.127.44.4]) by mx1.FreeBSD.org (Postfix) with ESMTP id B59F643D41 for ; Fri, 17 Sep 2004 20:56:44 +0000 (GMT) (envelope-from atlev@flyingcroc.net) Received: from fincher.users.accretive-networks.net (localhost [127.0.0.1]) i8HKui5b005032 for ; Fri, 17 Sep 2004 13:56:44 -0700 (PDT) (envelope-from atlev@flyingcroc.net) Received: from localhost (atlev@localhost)ESMTP id i8HKuiqd005029 for ; Fri, 17 Sep 2004 13:56:44 -0700 (PDT) (envelope-from atlev@flyingcroc.net) X-Authentication-Warning: fincher.users.accretive-networks.net: atlev owned process doing -bs Date: Fri, 17 Sep 2004 13:56:44 -0700 (PDT) From: Atle Veka X-X-Sender: atlev@fincher.users.accretive-networks.net To: freebsd-questions@freebsd.org Message-ID: <20040917133748.W31880@fincher.users.accretive-networks.net> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Subject: Coda, OpenAFS, or other alternatives that can handle high load X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 17 Sep 2004 20:56:45 -0000 Hey all, I'm looking for feedback on distributed filesystems like Coda or OpenAFS that does client side caching. Our needs at the moment are read-only but experiences with multiple nodes doing writes on a distributed filesystem would be nice as well. Let me explain our setup briefly; We have 17 load balanced FreeBSD servers in the cluster. 5 of those backend servers do ~20 requests per second (PHP - each hit has the potential of accessing multiple files, anywhere from 1-10) while the other 12 do ~120 requests per second (graphics - 5000 B/request on average) at peak. Each individual backend server has an identical set of content/data distributed via rsync. The optimal setup would be if each cluster could mount a read-only filesystem which is replicated for redundancy at the backend. From what I have read about Coda and OpenAFS, both already support this and they also have good caching features. What I would like to hear about are stability and performance experiences in the FreeBSD community. I have read that Coda in the past has had corruption problems. Thanks! Atle - Flying Crocodile Inc, Unix Systems Administrator