Re: Justifying a PG over MySQL approach to a project

On 12/21/09 11:23 AM PST, "Gauthier, Dave" <dave(dot)gauthier(at)intel(dot)com> wrote:

> I was wondering...
>
> In head-to-head comparisons, do DBs get stree tested, not only in terms of
> performance, but in terms of corruptions, down time, recovery time, lost data,
> etc... .?
>
> I've heard it said that MySQL is superior to MySQL in this regard. But if
> this were stated in an article from a 3rd party or something, THAT would make
> an impact on mgmt.
>
You are not going to find this kind of comparison (or at least a valid one)
for the same reason you would not find an off-the-shelf answer on
performance. It really depends on what your application does and how it
interacts with the database. Like performance testing, there's really no
substitute for trying it out yourself.

Some of the things you might try include crashing the database and seeing if
all the data come back, plus how easily can you run backups, how easily can
you upgrade, how hard it is to set up HA (MySQL master/master handy for
upgrades and to use as read-only slave, PostgreSQL warm standby far better
for full server replicas).

Hint: Putting a write load on MySQL with MyISAM table type and then
crashing the database is a really quick way to make MySQL look bad, as
you'll soon corrupt the tables and will typically get app failures until you
run myisamchk to repair them. Personally I like MySQL for a lot of purposes
but this one really drive me nuts. It's so easy to demonstrate it feels
like cheating.

Cheers, Robert Hodges