From owner-freebsd-hackers Thu Oct 19 12:20:22 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id MAA29197 for hackers-outgoing; Thu, 19 Oct 1995 12:20:22 -0700 Received: from brasil.moneng.mei.com (brasil.moneng.mei.com [151.186.20.4]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id MAA29174 for ; Thu, 19 Oct 1995 12:20:07 -0700 Received: (from jgreco@localhost) by brasil.moneng.mei.com (8.7.Beta.1/8.7.Beta.1) id OAA28484; Thu, 19 Oct 1995 14:18:55 -0500 From: Joe Greco Message-Id: <199510191918.OAA28484@brasil.moneng.mei.com> Subject: Re: Bragging rights.. To: dennis@etinc.com (dennis) Date: Thu, 19 Oct 1995 14:18:54 -0500 (CDT) Cc: hackers@freebsd.org In-Reply-To: <199510191752.NAA27991@etinc.com> from "dennis" at Oct 19, 95 01:52:29 pm X-Mailer: ELM [version 2.4 PL24] Content-Type: text Sender: owner-hackers@freebsd.org Precedence: bulk > >Now if you can find yourself a TA that can do 230.4 or 460.8, and the > >ET/5025 is able to do that in async, that might be a benefit. I've > >retrofitted 16550 ports to run at such speeds and the CPU does eventually > >reach a point where it has difficulty keeping up on a consistent basis > >(although this is more likely a driver issue...?) > > 115k minus 20% async overhead.....Mostly I've heard about 70k or so for > async links....If you don't think that 20-30% is worth an extra hundred > dollars, then I guess you're entitled to that. It is, however, a consideration. Really??!! I had a 386DX/40 that routinely chatted with a 386DX/16 at 115200 (UUCP over TCP/IP as a SLIP connection) and consistently hit > 10.5K/sec -- the number ran around 11K/sec during non-peak times here at MEI, and I attribute the difference to our network rather than any of the FreeBSD boxes involved (our network traffic peaks at wire saturation at times, and never falls below 10%). The 386/40 was also running the uucico for one side... These two systems had jacked-up serial ports. When I ran them at 230.4K, I started to hit serial overruns, but with TCP/IP retries they still were capable of hitting 19K/sec (which amazes me!!!) However at 19K/sec the 386DX/16 was definitely swamped, and the 386DX/40 was having troubles too (it was also maintaining the uucico, several uncompress/gzip's concurrently, another uucico running over ANOTHER SLIP link over a V.FC modem connection to sol.net). For those who are wondering what the hell I was doing :-) I was receiving a full news feed. Because I could not directly connect to the local network (firewall restrictions), I came in through a carefully gated, hardwired SLIP connection into the terminal server, and ran UUCP to gather batches of news. That was the high speed link :-) I then turned around and tossed it into sol.net's WAN structure with an ordinary POTS/V.FC modem. In the process, because the news was only compressed, I uncompressed it and used gzip -9 on it to achieve more virtual throughput :-) So I don't see an extra 20-30%, and I do see MORE than an "extra hundred dollars" (you have a product that sells for $139 with two ports?? I didn't think so). Everyone else please note: I am a satisfied ET customer, but I am just having a hard time swallowing this line of reasoning :-) ... JG