Tom Lane wrote:
"Joshua D. Drake" <jd@xxxxxxxxxxxxxxxxx> writes:
On Wed, 2006-12-13 at 18:36 -0800, Josh Berkus wrote:
Mostly, though, pgbench just gives the I/O system a workout. It's not a
really good general workload.
It also will not utilize all cpus on a many cpu machine. We recently
found that the only way to *really* test with pgbench was to actually
run 4+ copies of pgbench at the same time.
The pgbench app itself becomes the bottleneck at high transaction
rates. Awhile back I rewrote it to improve its ability to issue
commands concurrently, but then desisted from submitting the
changes --- if we change the app like that, future numbers would
be incomparable to past ones, which sort of defeats the purpose of a
benchmark no?
What is to stop us from running the new pgbench against older versions
of PGSQL? Any stats taken from a run of pgbench a long time ago
probably aren't relevant against a modern test anyway as the underlying
hardware and OS are likely to have changed or been updated.
