Tracy Johnson wrote:
>
> I have seen lower queues run wild within the CS through ES queue range.
> I was told once there was a bug in 4.0 that would cause it. I don't
> know if it was ever fixed in 5.0.
I don't know that I/O and other resources are "prioritized" by queue;
the queue primarily serves as a dispatcher criteria of which of the
ready processes are to be launched next. In fact, a heavily I/O bound
batch job at low priority (or serial read of datasets by a report
generator run in session) can drastically impact session response time
not only by contention for the I/O resources, but also by pre-fetches in
response to serial reads flushing pages from memory. The flushed pages
can be session stacks, code, buffers, pre-fetch, write cache (ouch!),
etc. The entire online transaction processing (OLTP) tuning model of
MPE can be thrown off-balance by an "inappropriate" batch job getting
enough CPU to impact other processes. This is especially true of
memory-constrained systems.
Jeff Kell <[log in to unmask]>