From owner-freebsd-hackers Wed Apr 19 07:58:44 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id HAA21347 for hackers-outgoing; Wed, 19 Apr 1995 07:58:44 -0700 Received: from shell1.best.com (root@shell1.best.com [204.156.128.10]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id HAA21340 for ; Wed, 19 Apr 1995 07:58:42 -0700 Received: from geli.clusternet (rcarter.vip.best.com [204.156.137.2]) by shell1.best.com (8.6.12/8.6.5) with ESMTP id HAA02926; Wed, 19 Apr 1995 07:58:23 -0700 Received: (from rcarter@localhost) by geli.clusternet (8.6.11/8.6.9) id HAA02568; Wed, 19 Apr 1995 07:57:46 -0700 Date: Wed, 19 Apr 1995 07:57:46 -0700 From: "Russell L. Carter" Message-Id: <199504191457.HAA02568@geli.clusternet> To: G.Watson@gu.edu.au Subject: Re: Case for FreeBSD presentation docs? Cc: hackers@FreeBSD.org Sender: hackers-owner@FreeBSD.org Precedence: bulk I bought a 192 wide node SP-2 for NAS last year, and supervised the benchmarking. After which I quit to do more interesting things. The 590 boards are very fine, and the switch and message passing architecture is very good, but the disks are about 2x slower than what you can run on pcs, and from what I understand it is hard to deal with large amounts of IO. Which is not a lot different from a cluster. You don't have all wide nodes, right? And about the load-leveler: I'm a beta site for the Portable Batch System being developed at NASA. The schedule for supporting parallel jobs on workstation clusters (including FreeBSD pc clusters) is mid-summer, so that is the last mandatory piece of the puzzle, I think. (Some would say they can't live without a Fortran source debugger, sissies I say ;-) Things are very exciting in the world of computing these days. Cheers, Russell