Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site poseidon.UUCP
Path: utzoo!watmath!clyde!cbosgd!ihnp4!poseidon!brent
From: brent@poseidon.UUCP (Brent P. Callaghan)
Newsgroups: net.arch
Subject: Re: IOCALL results and problems
Message-ID: <1376@poseidon.UUCP>
Date: Fri, 3-Jan-86 09:51:41 EST
Article-I.D.: poseidon.1376
Posted: Fri Jan  3 09:51:41 1986
Date-Received: Sat, 4-Jan-86 05:11:31 EST
References: <380@ncr-sd.UUCP>
Distribution: net
Organization: AT&T Information Systems, Lincroft NJ
Lines: 24
Keywords: IOCALL, Multiuser

The IOCALL and similar system benchmarks are fine for 
COMPARING computer systems, but if you want some numbers
for ACTUAL degradation in user response time, you have
to bite the bullet and get some real user activity
on those ttys.

Rounding up 10 or 20 "typical" users is not all that
easy, and its even harder to have them perform the
same activity repeatedly over a week or so of performance
measurements.

If you can afford the luxury of another computer, you can
connect it's ttys through null modems to your test system
and run a user process on each line.  Each process reads
a script of things to do.  The script can also specify
"think" times, typing rates etc.  The "user" processes
are untiring, repeatable, and very accurate in response
time measurement. 
-- 
				
Made in New Zealand -->		Brent Callaghan
				AT&T Information Systems, Lincroft, NJ
				{ihnp4|mtuxo|pegasus}!poseidon!brent
				(201) 576-3475