Re: [PERFORM] MySQL vs PG TPC-H benchmarks

Josh Berkus wrote:

> Folks,
>
> I've sent a polite e-mail to Mr. Gomez offering our help. Please, nobody
> flame him!
>

Please keep in mind that the entire test has, other than a similar
database schema and query types maybe, nothing to do with a TPC-H. I
don't see any kind of SUT. Foreign key support on the DB level is not
required by any of the TPC benchmarks. But the System Under Test, which
is the combination of middleware application and database together with
all computers and network components these parts are running on, must
implement all the required semantics, like ACID properties, referential
integrity &c. One could implement a TPC-H with flat files, it's just a
major pain in the middleware.

A proper TPC benchmark implementation would for example be a complete
PHP+DB application, where the user interaction is done by an emulated
"browser" and what is measured is the http response times, not anything
going on between PHP and the DB. Assuming that all requirements of the
TPC specification are implemented by either using available DB features,
or including appropriate workarounds in the PHP code, that would very
well lead to something that can compare PHP+MySQL vs. PHP+PostgreSQL.

All TPC benchmarks I have seen are performed by timing such a system
after a considerable rampup time, giving the DB system a chance to
properly populate caches and so forth. Rebooting the machine just before
the test is the wrong thing here and will especially kill any advanced
cache algorithms like ARC.

Jan

--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck(at)Yahoo(dot)com #