Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83 (MC830713); site edcaad.UUCP
Path: utzoo!linus!philabs!cmcl2!floyd!vax135!ukc!edcaad!pete
From: pete@edcaad.UUCP (Peter Sykes)
Newsgroups: net.unix-wizards
Subject: Upgrades and Tuning (A question)
Message-ID: <649@edcaad.UUCP>
Date: Thu, 14-Jun-84 01:24:02 EDT
Article-I.D.: edcaad.649
Posted: Thu Jun 14 01:24:02 1984
Date-Received: Sat, 2-Jun-84 11:49:49 EDT
Organization: Architecture, Edinburgh U., Scotland
Lines: 26


We are running out of vax at the moment. Our small (but greedy) user
population is complaining about response time and I am looking at upgrade
paths. The relevant  details are.

Vax 11/750 2Mb (old controller therefore full). 2 Disks one on the unibus
one on the massbus swapping on both and both resonably quick (< 30ms).
We are running 4.1c bsd and the load factor at bad times is about 4.
The applications are mainly interactive graphics, hence the response time
is important though typically only 2-3 users out of 8 are using graphics.
We also use Prolog coupled to a graphics manager.
Vmstat reveals that idle time is frequently 0 for minutes at a time system
time varies between 5% - 50%. pi and po are usually 0 but go up to 10-12
when loading large databases and the number of disk faults is also 10-12

Can anyone supply any clues as to what the most effective upgrade would be
or what I can fiddle with to tune the system. 

One thing I have noticed is that system time goes up dramatically when we
run prolog graphics and even more so with emacs (Up to 90%). The common
factor is the use of multiplexed files. Are they known to be inefficient
in 4.1c or can this be normal?

Thanks in advance
Pete Sykes          vax135!ukc!edcaad!pete