From owner-freebsd-questions Sat Jan 11 08:01:50 1997 Return-Path: Received: (from root@localhost) by freefall.freebsd.org (8.8.4/8.8.4) id IAA22017 for questions-outgoing; Sat, 11 Jan 1997 08:01:50 -0800 (PST) Received: from vex.net (shell.vex.net [207.107.242.162]) by freefall.freebsd.org (8.8.4/8.8.4) with ESMTP id IAA22012 for ; Sat, 11 Jan 1997 08:01:45 -0800 (PST) Received: from vex.net(really [207.107.242.162]) by vex.net via sendmail with smtp id for ; Sat, 11 Jan 1997 10:57:05 -0500 (EST) (Smail-3.2.0.90 1996-Dec-4 #4 built 1997-Jan-8) Date: Sat, 11 Jan 1997 10:57:05 -0500 (EST) From: Brian Tao To: Nathan Stratton cc: FREEBSD-QUESTIONS-L Subject: Re: News server In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk On Thu, 9 Jan 1997, Nathan Stratton wrote: > > I am planing on adding to main news servers to feed a network of > news feeders. I plan on using P5 200 with 256 megs ram and 2 9 gig > WSCSI for spool and 1 4 gig WSCSI for /. My questions is would I get > any speed increase if I went with P6 200 and not the P5 200? No, I don't think there will be any noticeable performance increase. News servers are typically I/O-bound and not CPU-bound. I'll repost a reply I had just sent to inet-access, although the numbers will be different for you (unless you also want 40GB of spool ;-) ). >>>>> On Thu, 9 Jan 1997, DeWayne Ashcraft wrote: > > Are there any drawbacks to using a 45GB disk array for the news > server? Configuration would have 5 9.1GB ultra wide disks, 7200 > RPM. Space and expense considerations aside, I would instead buy three ranks of 7x2GB 7200 rpm drive arrays, one ultra-wide controller per rank. Stripe your filesystems *across* the controllers using an appropriate strip width (it may be non-obvious what the optimal width is, depending on your OS). 24 drives (3 spare), 3 nice rackmount cases and 3 Adaptec 2940UW's will cost around $15000 US if I've done my currency conversion correctly. This setup is more for concurrency than for capacity. I would aim to limit your spool filesystem to about 50% just before an expire runs. Don't worry about the unused space... it will eventually fill. ;-) Correct striping will tend to localize disk accesses related to article retrieval to a single disk, leaving the remaining drives free to service other concurrent requests. <<<<< -- Brian Tao (BT300, taob@risc.org) "Though this be madness, yet there is method in't"