Date: Sat, 15 Apr 1995 03:47:32 -0400 (EDT) From: Peter Dufault <dufault@hda.com> To: rcarter@geli.com (Russell L. Carter) Cc: phk@ref.tfs.com, rgrimes@gndrsh.aac.dev.com, freebsd-hackers@FreeBSD.org, julian@TFS.COM Subject: Re: Just how fast can we go... (was: Re: SCSI target) Message-ID: <199504150747.DAA05621@hda.com> In-Reply-To: <199504150455.VAA02893@geli.clusternet> from "Russell L. Carter" at Apr 14, 95 09:55:41 pm
next in thread | previous in thread | raw e-mail | index | archive | help
Russell L. Carter writes: > > > If you are going to do a "nice little cluster", there is nothing more > important than latency. Bandwidth doesn't matter so much, once you get > above several user megabytes per second. I am not able to judge how > gather-scatter affects things; I do know that nearly all modern numerical > algorithms are constructed so that this does not occur. CRI systems > for quite some time have had hardware support for gather-scatter and > I've never seen an intelligently written code that required it. YMMV. Cluster as a poor choice of words. There is a quite a difference between a clustered MP system and two pentia hooked together sharing partitions, devices, and running TCP/IP over SCSI. And maybe with 100 mbps ethernet and fs cache it isn't even an interesting idea. The latency of the SCSI in FreeBSD is pretty discouraging. The actual use of this code is similar to where Bakul used it: in an embedded system, in our case in an industry (graphics pre press) that needs the interface to be SCSI. It has been in use a few years now and I finally moved it into the fold. That particular application is moving to fast-wide. -- Peter Dufault Real Time Machine Control and Simulation HD Associates, Inc. Voice: 508 433 6936 dufault@hda.com Fax: 508 433 5267
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199504150747.DAA05621>