Newsgroups: comp.protocols.tcp-ip
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!barmar
From: barmar@think.com (Barry Margolin)
Subject: Re: How is UTP fault-tolerant?
Message-ID: <1991May21.215628.26766@Think.COM>
Sender: news@Think.COM
Reply-To: barmar@think.com
Organization: Thinking Machines Corporation, Cambridge MA, USA
References: <1991May20.133442.1309@banana.fedex.com> <1991May21.172018.11672@zoo.toronto.edu>
Date: Tue, 21 May 91 21:56:28 GMT

In article <1991May21.172018.11672@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes:
>Assuming "UTP" means "10BaseT", the crucial issue is that 10BaseT is a star
>topology, with each host connected directly to the hub, whereas other forms
>of Ethernet are bus topologies, with more than one host hanging off of each
>piece of wire.  Given good equipment design, star topologies have one big
>advantage:  a foulup on one wire does not mess up anyone else.

We use an intermediate solution here.  Our network is mostly thin Ethernet
(we considered twisted pairs when we were rewiring, but I don't remember
why we decided against it), with the large subnets implemented using a
number of segments radiating from an intelligent repeater (Cabletron IRMs).
Any one segment has about a dozen hosts on it, so a cable problem generally
is limited to affecting that many hosts.  The intelligent repeater isolates
the segments electronically and can selectively shut down an individual
port (either manually, or automatically upon certain triggers such as
excessive error rates).

-- 
Barry Margolin, Thinking Machines Corp.

barmar@think.com
{uunet,harvard}!think!barmar
