Very Computer

by **William Hawthor** » Sat, 25 Mar 1995 10:53:52

Has anyone done any classic queueing theory modeling of DB2 requests? I'm
researching a scenario where hundreds of interactive users of an
application will be performing SELECTs against a single table, sharing an
(as yet) indeterminate number of open threads. For simplicity, assume
that only DB2 and the relevant application are running on the processor,
and that the only DASD contention is due to inter-thread conflicts. Also
assume that each user SELECT is no more complex than any other, and the
requests are Poisson distributed.

Some questions come to mind:

Is M/M/1 queue modeling appropriate?

Does DB2 ever decide that a maximum number of threads may be active at
one time -- or are they all multitasked with equal priority? If DB2
imposes a maximum, is the dispatching arranged in a queue?

Are the only "servers" (in the queueing modeling sense of the term) DASD
and the threads?

Newsgroups: bit.listserv.db2-1
Subject: DB2 Queueing Models
Summary:
Followup-To:
Distribution: usa
Organization: Texas Metronet, Internet for the Individual 214-705-2901 (info)
Keywords: