jonathan@comp.vuw.ac.nz (Jonathan) (11/03/87)
The Commerce faculty here is acquiring 50-100 PC-AT clones
which they want to network together and to an IBM mainframe (4381).
Currently there is an Ethernet on campus connecting various Unix and VMS
machines running TCP/IP (4.[23] bsd, and CMU/Tek on VMS), DECnet, and DEC's
LAT.
The Commerce faculty favours a token ring to connect the PCs to
3Com fileservers, apparently because they believe the Token Ring will perform
better under heavy loads (eg, booting a labful of ATs from the server
at the same time).
Given the current campus setup, we see Ethernet for the PCs and 4381 as
a possibly more viable alternative.
We would like to find out the following:
* are the Commerce Faculty's fears of saturating an Ethernet justified?
We have no experience with Token Rings, and do not know how their
performance compares with Ethernet. However, we do not think that the PCs
will generate sufficient Ethernet traffic to cause concern.
* What protocols would the PCs use to talk to the 4381 over Token Ring?
What facilities would these protocols provide?
How would they inter-operate with the existing protocols on the Ethernet?
* What is the availability/reliability/cost of Token Ring/Ethernet gateways?
* What experience have people had with Ethernet for IBM mainframes
(software and hardware) ?
* The Commerce faculty already has a number of Macs. How would they fit
into a Token Ring network? (We know this isn't a problem for Ethernet; we
are using Kinetics boxes.)
Any advice or information you can give us will be gratefully received.
--
|``Toto, I have a feeling we're
Sane mailers: jonathan@comp.vuw.ac.nz | not in Kansas anymore ...''
UUCP: ...!uunet!vuwcomp!jonathan | - Dorothy,arriving in Ozkerr@oodis01.ARPA (Grant Kerr) (11/06/87)
The novell "Lan Evaluation Report" page 53 compares several different types of network hardware, among them are Token Ring and Ethernet. The chart show that the ethernet has nearly the twice the throughput at all load levels. Given that the Ethernet boards are cheaper, faster and available for more kinds of computers (of any size) the conclusion should be obvious. Local testing here confirmed the same information as the novell chart. -- Grant Kerr Control Data Corporation INTERNET: kerr@oodis01.arpa UUCP: ihnp4!lll-tis!oodis01!kerr