Subject: | |
From: | |
Reply To: | |
Date: | Wed, 23 Aug 1995 17:25:01 GMT |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Gavin mused:
>Jeff said:
>> Yes, but all 3rd-party solutions implement a periodic scan of the spool
>> queues looking for candidates.[...]
>
>This has been something that ought to be addressed via an AIF.
>It should be possible to ask the system to awaken your process as soon as
>there is a new spoolfile in a particular 'queue'. This
>would let all the 3rd party software that currently has to do polling
>switch to an interrupt style of processing, which would greatly reduce
>their overhead, and eliminate the delays that currently exist before a
>new spoolfile will be seen and processed.
This is one of the more annoying things for our users, who were recently
torn from their comfortable world of serially-attached terminals and
printers (that responded immediately) and flung into the hurly-burly of
networking (and Windoze). We use NetPrint92 to provide printing for them
and ourselves, and the wait for printing is just one more straw on their
already-tender backs. An interrupt-driven approach would certainly help
to alleviate this situation.
(This average wait is n/2 seconds, where n is the rescan interval set in
the configuration file.)
Incidentally, I have observed CPU consumption upwards of 70-80% for short
periods by NP92MGR as it scans the spooling queues near end-of-day, when
we often have several hundred spoolfiles generated by our end-of-day jobs.
--Jim
|
|
|