From owner-freebsd-hackers Sat Apr 15 02:35:42 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id CAA06880 for hackers-outgoing; Sat, 15 Apr 1995 02:35:42 -0700 Received: from netcom21.netcom.com (bakul@netcom21.netcom.com [192.100.81.135]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id CAA06868 for ; Sat, 15 Apr 1995 02:35:41 -0700 Received: from localhost by netcom21.netcom.com (8.6.12/Netcom) id CAA18801; Sat, 15 Apr 1995 02:35:26 -0700 Message-Id: <199504150935.CAA18801@netcom21.netcom.com> To: Peter Dufault cc: freebsd-hackers@FreeBSD.org Subject: Re: Just how fast can we go... (was: Re: SCSI target) In-reply-to: Your message of "Sat, 15 Apr 95 03:47:32 EDT." <199504150747.DAA05621@hda.com> Date: Sat, 15 Apr 95 02:35:24 -0700 From: Bakul Shah Sender: hackers-owner@FreeBSD.org Precedence: bulk > 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. TCP/IP over SCSI can be useful even if boring. You don't need a separate 100Mbps card and > 1MB/s is better than what we get with enet! High SCSI latency says you want the MTU to be large (let us see: with an MTU of 8KB on a 10MB/s max rate SCSI bus you can get 80% effective use even with 200us command overhead -- if the system bus cooperated). > 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. In my case I conviced my client to dump SCSI (this was for a closed system). I replaced their fancy Multibus II SCSI controllers with a pair of very simple FIFO boards built out of samples and scrounged parts. What we really need are *disks and tapes* that talk network protocols or a kind of simple distributed memory protocol. SCSI needs to die. --bakul