[e2e] on local ethernet throughput?

[Vernon Schryver]

> From: Cannara <cannara at attglobal.net>

> ...
> Do read the Dec WRL report from 1988, which answered this issue quiet clearly
> and competently, by controlling one variable most folks just jabber (excuse
> the pun) about:  Offered Load.  If you're interested, I can email a copy.  In
> essence, depending on frame size, 20 or so nodes can maintain >90% throughput
> at the DLC level on a classic CSMA/CD segment.  The paper was written to
> dispel the then prevalent notion promulgated by IBM, for instance, that Token
> passing was superior to CSMA/CD in guaranteeing LAN throughput.  It also shows
> the fragility of the logic claiming a "capture effect" when all adpators are
> playing by the same rules (which Seeq in the '80s tried, until punished, to
> violate).  
> ...


While the WRL report was a great counter to the Token ring jabber
of the time, I think it predated concerns about the Capture Effect.
It may have predated the crystalization of the notion of the capture effect
and I don't think it mentions the idea except as "overload."
Mart Molle's investigations found some problems in 
Boggs et al. "Measured Capacity of an Ethernet: Myth and Reality"
I think the paper can be found at
ftp://gatekeeper.dec.com/pub/DEC/WRL/research-reports/WRL-TR-88.4.pdf

A message about "Performance problems on high utilization Ethernets"
by Wes Irish in comp.dcom.lans.ethernet from Feb. 1995 contains an
interesting list of problems in several flavors of Ethernet silicon.
Since he was "Network Scientist, Xerox PARC", I hope and think he
also published his note elsewhere.  Feb 1995 is just before
Dejanews/Google starts.  http://www.rhyolite.com/tmp/ethernet is
a temporary copy for anyone who cares and can't find it.

I don't know what SEEQ chips Mr. Cannara is talking about.  Wes Irish
mentioned only a bogus IFG for the SEEQ 8003, and not something that
I'd call anything about "claiming a capture effect."

Many Ethernet MACs after SEEQ and the LANCE had and no doubt still have
knobs to intentionally abuse various parts of 802.3 CSMA/CD.  One in
custom silicon I fought against and lost reproduced the old LANCE 7990
deferral or politeness bug as a way to avoid the capture effect without
the complications of BLAM.

The SEEQ chips I knew about were too stupid to do much intentional
violating of of anything.  I was introduced to the Capture Effect by
custom silicon extremely similar to the SEEQ 8003.  We were flummoxed
by newer, faster hardware could not push as many TCP/IP/Ethernet bits
as older hardware using 7990's and that encountered higher collision
rates.  It was almost enough to make us doubt the revealed truth of
Boggs, Mogul and Kent.  It turned out to be the politeness bug in the
7990 that avoided the capture effect and let the old systems go faster.
I could drone on about hacks to TCP ACK pacing, TCP window sizes, and
other things to prevent the capture effect with 802.3 compliant MACs,
but won't without more provcation.


Vernon Schryver    vjs at rhyolite.com