Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!necntc!ames!sdcsvax!ucsdhub!hp-sdd!hplabs!motsj1!mcdchg!usenet
From: pat@seradg.Dayton.NCR.COM (Patrick Pesch)
Newsgroups: comp.unix
Subject: Setting accurate application timers under Unix/Xenix
Message-ID: <1759@mcdchg.UUCP>
Date: Fri, 18-Sep-87 00:52:50 EDT
Article-I.D.: mcdchg.1759
Posted: Fri Sep 18 00:52:50 1987
Date-Received: Sun, 20-Sep-87 01:40:02 EDT
Sender: usenet@mcdchg.UUCP
Organization: SE-Retail Advanced Development, NCR Corp, Dayton
Lines: 32
Approved: usenet@mcdchg.UUCP


	From time to time we come up with the need to set timers
	within user mode applications for such things as message
	response timeouts, etc...  While the SIGALRM signal used
	with alarm(2) is okay for some situations, it lacks a lot
	of functionality, such as:

		o  Less than 1 second granularity

		o  Multiple simultaneous outstanding timers

		o  Uses signals, which may not be reliable in all
		   Unix variants (true?)

		o  Not very accurate when there are *lots* of processes
		   that are runnable (not sleeping)


	What I would like to know is if anyone else out there had
	come across this sort of application, and if any solutions 
	were found.  ANY suggestions would be appreciated...

	It seems to me there should be a way to write a little 
	"timer pseudo driver" that would help processes set/unset timers,
	and allow processes with timers expiring to get scheduled "quicker" 
	than other normal processes by fiddling with there priorities or
	something...  I realize that Unix is *NOT* a real time OS, but I
	don't really need a true real time system... ?comments?

				Patrick Pesch
				NCR Corporation
				Dayton OH