Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!ll-xn!ames!aurora!jbm
From: jbm@aurora.UUCP (Jeffrey Mulligan)
Newsgroups: comp.unix.questions
Subject: Re: real-time UNIX
Message-ID: <931@aurora.UUCP>
Date: Tue, 25-Aug-87 13:44:34 EDT
Article-I.D.: aurora.931
Posted: Tue Aug 25 13:44:34 1987
Date-Received: Wed, 26-Aug-87 07:11:35 EDT
References: <8976@brl-adm.ARPA>
Organization: NASA Ames Research Center, Mt. View, Ca.
Lines: 33

We are running a real-time system that noone has mentioned yet;
it is called VENIX, and is sold by a company called Venturcom,
which is located in Massachusetts.  We run it on a PDP-11/73,
but they support a variety of machines.  There are a variety
of additional system calls:

nice(-100)			hog cpu
phys(...)			map I/O page (for instance) into user space

This system is OK, but there are a few bugs.


In my previous lab, we ran an old version 6 kernel (on an 11/23) which
had an enhancement (from Berkeley?) in the form of an rtp() system
call (real-time process).  This basically inhibited swapping.  As
we had kernel source, I was also able to do some things like squelch
lightning bolt clock interrupts.  I must admit I was never interested
in running a spreadsheet while doing a realtime task.  We did sometimes
run experiments with a second user editing at the same time, but
the performance for the second user was pretty poor.


I also had some experience with a Bell Labs in-house hack which provided
features similar to VENIX, i.e. mapping the I/O page into user space.
Another interesting feature it had was that you could set device
interrupts to trap to user space routines.


-- 

	Jeff Mulligan (jbm@ames-aurora.arpa)
	NASA/Ames Research Ctr., Mail Stop 239-3, Moffet Field CA, 94035
	(415) 694-5150