Path: utzoo!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!nosc!ucsd!ucbvax!decwrl!labrea!Portia!Jessica!morgan
From: morgan@Jessica.stanford.edu (RL "Bob" Morgan)
Newsgroups: comp.dcom.lans
Subject: Re: 802.3 a panacea ?
Message-ID: <3419@Portia.Stanford.EDU>
Date: 24 Aug 88 00:31:12 GMT
References: <10400002@osiris.cso.uiuc.edu> <4942@cos.com> <1988Aug18.165532.26169@utzoo.uucp> <20447@watmath.waterloo.edu> <1988Aug22.170009.4743@utzoo.uucp> <6896@umn-cs.cs.umn.edu>
Sender: news@Portia.Stanford.EDU
Reply-To: morgan@Jessica.stanford.edu (RL "Bob" Morgan)
Organization: Stanford University
Lines: 31


I heartily urge everyone who is interested in Ethernet performance and 
characterization to read the recently released paper

	"Measured Capacity of an Ethernet:  Myths and Reality"
	by David R. Boggs, Jeffrey C. Mogul, and Christopher A. Kent

which was delivered at last week's ACM SIGComm symposium [consider
this yet another plug to join the ACM!].  The work was done at DEC's
Western Research Lab; I imagine the paper may be available at some
point as a technical report from them.  Also, I believe the symposium
proceedings will come out as an issue of the SIGComm newsletter, which
should be at your favorite university library.

Quoting from the introduction [by permission of ACM, copyright ACM 1988]:

"We first summarize the theoretical studies relevant to Ethernet, and
attempt to extract the important lessons from them.  Then, based on
measurements of actual implementations, we show that for a wide class
of applications, Ethernet is capable of carrying its nominal bandwidth
of useful traffic, and allocates the bandwidth fairly.  We then
discuss how implementations can achieve this performance, describe
some problems that have arisen in existing implementations, and
suggest ways to avoid future problems."

Basically, you can get as close as you want to 10 Mbps, and the
exponential backoff algorithm really works to keep collisions low and
maximum delay reasonable, even on a maximum-length net.

- RL "Bob" Morgan
  Networking Systems
  Stanford