Blair Jensen wrote:

> Thanks for the reply. I'll be putting up a data warehouse later this year
> so I was attempting to test the impact of intra-parallel against all of my
> existing load tests. Do you have any ideas?

I typically use intra-partition parallelism for most reporting servers -
even non-DPF ones: basically any data mart/data warehouse where the
queries are getting 1000s of rows at a time. The results are nearly
linear improvements with up to four CPUs - according to my simple
benchmarks.

However, two marts just support custom web applications in which the
queries only hit highly indexed summary tables. One handles 60,000+
queries a day - and uses no parallelism: since that isn't necessary, and
cuts down on memory needed for additional concurrent applications.

On the OLTP side, max_querydegree of 'any' should typically result in 1
for transactional queries, and more for the occasional report. But I
haven't chanced it - just left intra_parallel completely off.

And load utility parallelism is handled by CPU_PARALLELISM parameter in
loads, and UTIL_HEAP_SZ in db cfg.


regards,

ken

DB2 & UDB email list listserv db2-l LazyDBA home page