Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site unrvax.UUCP
Path: utzoo!watmath!clyde!cbosgd!ihnp4!mhuxn!mhuxr!ulysses!allegra!mit-eddie!genrad!panda!talcott!harvard!cmcl2!seismo!utah-cs!utah-gr!stride!unrvax!pjg
From: pjg@unrvax.UUCP (Paul Graham)
Newsgroups: net.lan,net.unix-wizards
Subject: TAC/serial performance
Message-ID: <398@unrvax.UUCP>
Date: Wed, 13-Nov-85 12:51:15 EST
Article-I.D.: unrvax.398
Posted: Wed Nov 13 12:51:15 1985
Date-Received: Fri, 15-Nov-85 05:22:24 EST
Reply-To: pjg@unrvax.UUCP (Paul Graham)
Organization: University of Nevada Reno
Lines: 20
Keywords: ethernet multiplexor performance
Xref: watmath net.lan:1147 net.unix-wizards:15714


We are considering buying bridge cs/100 or cs/1-a type devices instead
serial controllers.  (These are ethernet based terminal multiplexors,
sometimes know as tac's or tip's). Can anyone relate experiences which
would suggest this is a reasonable idea or an especially bad idea.
Although cost is an issue we are more concerned about relative performance
of kernel services for the tac versus the serial controllers.

The serial contr. can be considered equivalent to a dmf32 or other
dma device running under 4.2bsd with no specific optimizations know.

If it matters, the hardware is either a CCI 6/32 (from Sperry) or a Pyramid
90x.

Please respond by mail, I will summarize if interest warrants.

-- 
Thanks for your time.

Paul Graham seismo!unr70!unrvax!pjg