Ok. I got two responses, and I don't think I need to read any more.
The answer is...the way UNIX does it, is the only efficient way to do it.
And I won't gain anything by changing the way UNIX does it.
Thanks for Dr Thomas Blinn and Marc Baudoin. :)
-Jonathan Williams
-----Original Message-----
From: tru64-unix-managers-owner_at_ornl.gov
[mailto:tru64-unix-managers-owner_at_ornl.gov]On Behalf Of Jonathan
Williams
Sent: Tuesday, June 28, 2005 9:55 AM
To: tru64-unix-managers_at_ornl.gov
Subject: Sleeping Processes
We had some problems with our new GS1280 last week, and since then the
question of system tuning in general has come up. One of the questions came
up was about our application server.
The Application server is an ES80, running Tru64 Unix 5.1b, pk4. It runs a
proprietary application that consists of many (well...about 50 or so)
individual programs that are each attached to their own queue. They
basically just wait around until something is in their queue, then do it
real quick, and go back to waiting.
While the queues are waiting, the processes are in a SLEEP mode (going by
top, and ps), and then switch to running for a bit while they are doing
something, and then back to sleeping.
The question is, is there any amount of time involved in switching from
sleeping to running. One of the programmers is thinking there is a way to
have the processes in a RUNNING state all the time--I'm not sure if that's a
UNIX thing, or a programming thing. But whether we want to change it
depends on whether or not there is any time (anything more than .1 sec)
wasted by switching from sleeping to running. Or is it like a millionth of
a second or something negligible, in which case we don't really care.
So is this something people know? Or something I can measure myself? Any
help, or pointers in the right direction would be greatly appreciated. TIA
-Jonathan Williams
Received on Tue Jun 28 2005 - 15:27:20 NZST