Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.3 4.3bsd-beta 6/6/85; site topaz.RUTGERS.EDU
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!seismo!columbia!topaz!hedrick
From: hedrick@topaz.RUTGERS.EDU (Charles Hedrick)
Newsgroups: net.unix-wizards
Subject: IBM Ethernet interface
Message-ID: <3475@topaz.RUTGERS.EDU>
Date: Thu, 29-Aug-85 23:00:34 EDT
Article-I.D.: topaz.3475
Posted: Thu Aug 29 23:00:34 1985
Date-Received: Sat, 31-Aug-85 08:22:35 EDT
References: <1108@brl-tgr.ARPA>
Reply-To: hedrick@topaz.UUCP (Charles Hedrick)
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 31

Well, I think I know what Art was hinting at.  I'm sure he was
following the convention that manufacturers are not allowed to
advertise their own products.  Since we are just a customer, I think I
can talk.  ACC is building an Ethernet interface for IBM systems.  It
is intended to be high performance.  (The design goals have numbers
noticably higher than those that you see on a VAX.)  It is part of a
more general network front end effort that they are working on.  The
intent is to have a single interface that supports Ethernet, DDN, etc.
The version we are getting is to look just like the ACC 1822 interface
as far as the host is concerned.  Thus it should be able to use the
UCLA TCP/IP code for MVS without any change.  Their eventual plans
involve putting more intelligence into the front end than this.

I had not said anything in public about this for the last year, since
we had had problems getting a clear commitment to delivery time from
ACC.  However I have heard that our box will be shipped next week, so
it sounds like the thing is real.  I won't give a price, because it
may have changed since when we ordered it (over a year ago).  But I
think it is fair to say that it is going to be significantly more
expensive than a DEUNA.  It is also a much more ambitious and complex
piece of equipment than a DEUNA.  It has several processors (68000's,
I think) in it.

I'll say more about this after we get it and have some experience with
it.  I may have a hard time giving you good performance figures on it,
since I suspect it will run faster than anything we have to test it
with (though our primary Unix systems are Pyramids, and their TCP is
somewhat faster than TCP on the VAX).  Given the role (and cost) of an
IBM mainframe, one would like it to be able to feed several VAXen at
the same time, driving each of them at their top speeds.  Thus it
makes sense for ACC to have set high goals for performance.