At 09:07 AM 2/13/96 EST, Ken Nutsford wrote:
>On  Tue, 13 Feb 1996 14:48:17 +0500 Mohan Das wrote:
>
><snip>
>
>I have a major concern with the loss of the global limit. My vote goes for a)
>because its positive side outways the positive side of b) and the negative side
>of b) outways the positive side of a). Overall, a) wins on both counts.
>
 
I too have a concern with the lack of an overall global limit.  I can see
where I might want to have a number of logical queues with individual limits
yet still limit the total system.
 
On the surface, it seems to me that if the global limit is smaller than the
total number of jobs that the individual limits would allow at a point in
time, the various competing subqueues should be processed in some
round-robin fashion based upon the actual priority number.
 
I believe that the idea is not to force <n> jobs to run in a subqueue but
rather to keep more than <n> jobs running in a subqueue concurrently.  The
fact that no jobs might be running at some point in time seems ok to me, the
relative priorities of the jobs would allow you to cause hot queues (eg.
payroll on payday) to run.  Maybe it requires a relative priority for each
subqueue itself but this may be unnecessarily complex.
 
Brian Duncombe  [log in to unmask]
"Inside every large program is a small one struggling to get out."
          C. A. R. Hoare