On Wed, 1 Dec 1999, Shaul Karl wrote:
> I understand why "This is often a waste of CPU time...".
> What I am failing to understand is that the "OS can never put these processes
> to sleep".
the notin here is not "technically cannot", but rather "would not". these
processes will go to sleep due to context switching. the idea is that they
will use every free CPU cycle, because of the way they are written, and
because on a (normal) unix system, one cannot assign fixed priorities to
processes (this is not true when dealing with real-time extentions a-la
posix threads, i beleive). tuhs, other processes, who have actually usage
for the CPU, will suffer the need to constantly share it with these 'busy
poll' processes, and that is probably the idea that the paragraph was
trying to illustrate.
> If that is true then what about the kernel total control of the sys
> resources? What about the kernel always dividing sys resources among users
> according to sys policy?
"normal" unix has no notion of dividing resources based on policy. it has
support for upper limits on resources (such as total CPU time a single
process may consume before being killed), but not for "quality of service"
issues...
guy
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
