Date: Fri, 2 Feb 1996 12:11:55 -0500 (EST) From: Peter Berger <peterb@telerama.lm.com> To: dennis <dennis@etinc.com> Cc: hackers@freebsd.org Subject: Re: Multi-Port Async Cards Message-ID: <Pine.BSI.3.91.960202115948.8158A-100000@ivory.lm.com> In-Reply-To: <199602011704.MAA19726@etinc.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 1 Feb 1996, dennis wrote:
> Peter Burger writes....
That's "Berger".
> This is wrong on all 3 counts. Experience tells otherwise. A similarly
First off, you never address the support question. Second, my
"experience" only consists of helping support a network of 2 Cisco
7500's, 2 AGS+'s, a 7000, routing a T3 to the global Internet and
multiple Ethernets, FDDI rings, and metropolitan T3s and T1s, as well as
supporting a couple of unix routers. So perhaps you're right, and i'm
just haven't had enough "hands on" experience to be able to understand
these things the way you, with your oh so much more complex network, can.
> priced unit
> with a Web server will easily outperform a Cisco. Cisco's method of handling
> dumb serial controllers for hi-speed lines takes much away from its apparent
> architectural advantages, plus the use of a slow processor (which is consumed
> handling serial/ethernet traffic) makes any processing task (like filtering)
> very slow
> compared to a (say) Pentium solution. With a PC, basically, you get
> flexibility and power for a much lower cost. For Example, you can't add a
> 100Mbs port to a Cisco
> for $134!
That's right; and you can't add a 100Mbp/s port to a PC that will
actually route that many packets for $134, or for any price. Reference
the very interesting TCP performance tests at Usenix which showed that at
Ethernet MTUs, Pentium boxes running TCP/IP over the loopback interface
could only reach about 40Mb/s (this number went up if you increased the
MTU ... the cost is in the packet processing, not the raw byte speed).
The simple fact is that the relatively slow processor on the Cisco is not
a major problem if you're serious about routing. If you hit the
processor (i.e. if you're doing ip level filtering), you are a poor
network designer -- you should not be using your router as a firewall.
Pure routing through ciscos takes the fast path (a.k.a. the silicon) and
doesn't load the CPU substantially, if at all. The slow path is there
for people who insist on shooting themselves in the foot. And people who
run a web server on their router are not making a wise decision. I can
cook steaks on my engine block, but I'd be dumb if I tried to do it while
my car was going 65.
Incidentally, not only didn't you address the support issues, but you
also neglected to counter my observation that ciscos give a port density
beyond anything even remotely possible in a PC.
Anyway, this is a stupid argument. I like FreeBSD. FreBSD makes a good
T1 or Ethernet router for simple networks. Anyone who tries to use it in
a production situation for multiple very high speed connections is
begging for trouble, not because of anything specific to FreeBSD but
simply because the i386 busses and unix kernel combination in general
hasn't been optimized enough, yet.
> Synchronous Communications Cards and Routers For
> Discriminating Tastes. 56k to T1 and beyond. Frame
> Relay, PPP, HDLC, and X.25 for BSD/OS, FreeBSD
> and LINUX
Oh. So that's why you're being obnoxious. I should have
guessed. Tell you what -- come back and tell me about what your
"experience" tells me when you're routing DS3.
"The law locks up both man and woman/Who steals the goose from off the common
But lets the greater felon loose/Who steals the common from the goose." -anon
-------------------------------------------------------------------------------
Peter Berger - peterb@telerama.lm.com - http://www.lm.com/~peterb
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSI.3.91.960202115948.8158A-100000>