From owner-freebsd-hackers  Sat Jun 24 14:07:34 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id OAA25468
          for hackers-outgoing; Sat, 24 Jun 1995 14:07:34 -0700
Received: from gndrsh.aac.dev.com (gndrsh.aac.dev.com [198.145.92.241])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id OAA25455
          for <hackers@freebsd.org>; Sat, 24 Jun 1995 14:07:31 -0700
Received: (from rgrimes@localhost) by gndrsh.aac.dev.com (8.6.11/8.6.9) id OAA13740; Sat, 24 Jun 1995 14:07:11 -0700
From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com>
Message-Id: <199506242107.OAA13740@gndrsh.aac.dev.com>
Subject: Re: FreeBSD as a router
To: davidg@Root.COM
Date: Sat, 24 Jun 1995 14:07:11 -0700 (PDT)
Cc: nc@ai.net, hackers@freebsd.org
In-Reply-To: <199506242048.NAA00597@corbin.Root.COM> from "David Greenman" at Jun 24, 95 01:48:25 pm
X-Mailer: ELM [version 2.4 PL24]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 1985      
Sender: hackers-owner@freebsd.org
Precedence: bulk

...
> >On a pentium, we are under the assumption that the bus and processor 
> >aren't the limiting factor, and just BSD is slowing things down. So what 
> >is it?
> 
>    I think we should assume high performance hardware. The difference in cost
> between a Pentium-90 w/PCI (bus mastering DMA) ethernet card and a 486/66
> w/ISA ether is fairly small these days (unless of course you already have the
> 486/66...). The limitation is definately software at this point. That's why we
> do well in bytes/sec, but poorly in packets/sec.

I would like to quantify that ``fairly small these days''.  AMD DX2/66 CPU
chip, $131.00.  Intel P54C-90 CPU chip $421.  Over 3X, something I defanitly
would not call a ``small amount''.

Then add to that the difference in MB cost, $100 for the lowend 486MB, vs
over $200 for the lowest price P90 MB I could find.  Again 2X.  Now onto
the ethernet card, $49.00 for a cheap ISA WD80xx clone, $99.00 for the
a PCI card.

So it looks like $280 vs $720.  I would call that ``significant'' amount
of money.  Also since this is probably going to be highly memory speed
dependent I suspect an ASUS PCI/I-486SP3G (PCI) could route packets just
about as fast as an ASUS PCI/I-P54TP4 due to the fact that thier main memory
speeds are *very* close.  (Note the ASUS 486SP3G costs as much as the
P54TP4 so the price scale is only the CPU delta, but you get a free NCR
SCSI controller in the deal :-).)

> >high. On higher utilized networks, I can't imagine 10-12 ms latency on a 
> >80 megabit stream of packets is a problem. 
> 
>    We're not talking anywhere near that much delay. ...More like 700-800us.
> Again, the problem isn't latency or 'bandwidth'. The problem is packets/sec.

Agreed, and that is probably what was stuffing up my values, I will do a
more complete test and see what I can get.  


-- 
Rod Grimes                                      rgrimes@gndrsh.aac.dev.com
Accurate Automation Company                 Reliable computers for FreeBSD