On Wed, 31 Jul 1996 j.n.bain@??? wrote:
> My worry would then be the case of a large (say 1000 subscribers) list. The
> above would generate 50 separate messages. So it follows that one message sent
> to a large mailing list could monopolise the MTA - something I guess is not
> good.
This is a possibility with any parallelization scheme.
This all reminds me of the good old days of trying to schedule jobs on
an IBM mainframe. One of the things that was very hard to deal with was
the 9am cliff. A naive scheme, when presented with a huge job at 8:30
am, saw the machine was empty and proceeded to run it. Large job then
monopolized the machine till well after 9am, by which time people had
arrived and were trying to run small jobs. I think similar things could
happen here; even if you have rules for sharing out your machine among
the messages currently trying to be delivered, it may start out grabbing
a lot of the resources for a huge delivery list because there's nothing
else going on at the time, thereby reducing resources available to a lot
of little messages that come in afterwards. Whether that would actually
matter in practice I do not know.
--
Philip Hazel University Computing Service,
ph10@??? New Museums Site, Cambridge CB2 3QG,
P.Hazel@??? England. Phone: +44 1223 334714
