ECE 578 Challenge Problem 1                                          DRAFT 12-Sept-08
Technology, Open-Source, and Politics

In the late 70's Xerox decided to offer its Ethernet technology as a free and open
standard for local area networks.  The IEEE formed a committee, and was well on its
way to finalizing standard 802.3 when IBM, the dominant computer monopoly at the
time, decided to push their token ring technology as an alternative.  Opponents saw
this as an unecessary fork.  In any case, it led to delays and billions of dollars in
extra cost as industry tried for many years to support multiple standards.

The rationale for the fork was that Ethernet was flawed because of the way it handled
media access.  When an Ethernet is heavily loaded, collisions limit total throughput
to about 80% of the raw channel capacity, and there is no way to guarantee a fixed
bandwidth allocation to a station that needs a steady data flow with no gaps.  Even
with a large buffer at the receiving end, the statistical nature of Ethernet's
"exponential backoff" algorithm cannot guarantee the buffer will never run out.

a) What are your thoughts as to the merits of IBM's argument?  Remember at the time,
buffers were expensive, and devices like printers needed a steady flow of data.
Network cables were expensive too, and it was hard to forsee today's typical office
network with packet switches in every closet and less than 1% utilization of cable
capacity.

b) How would you modify the design of Ethernet to allow for a few devices with a fixed
share of the available bandwidth?  Remember at the time, Ethernet was fairly well
established, so a complete re-design would have less chance of acceptance than a minor
tweak.

c) How well does your modified design perform in comparison with a standard Ethernet?
Make a plot of total throughput vs load, similar to Fig. 1 in
http://ece.arizona.edu/~edatools/ece578/Ethernet/
Hint:  If your design mods are minor, you should be able to use the same simulation
program that was used to generate Fig. 1.

d) If you could jump in your time machine, and go back and put yourself in charge of
the standards committee, would you publish these design mods as part of the standard,
or an option for those that need it?  There is no right or wrong answer here.  We're
looking for the quality of your reasoning.  What are the tradeoffs?  What are your
assumptions?  Think about the last 25 years.  Hindsight is wonderful. :>)


=== Notes on the simulation programs ===

see READme.txt