Open Source Databases: As the Tables Turn

My July Article comparing PostgreSQL and MySQL
caused a stir that really surprised me. Of course, users of both databases
called the tests rigged one way or the other, even though I noted that
both databases had their strong points under various circumstances.

All that aside, we’ve been running into some rather grim and serious
MySQL problems on SourceForge –
most of which are directly related to MySQL’s table-level locking mechanism.
To give us some breathing room, we’ve been running a Beta version of MySQL
with pretty good success.

Those problems and other MySQL limitations have caused me to evaluate
database alternatives, first Interbase (apparently a dead end), then PostgreSQL.

It didn’t take long to rule out Interbase after discovering that it
lacked replication, full text indexing, and had serious community
and corporate issues around its “Open Sourceness”.

Postgres also lacks replication and full text indexing (MySQL’s latest
betas include both features), however the postgres people issued a press
release about replication (currently vaporware) and it will be coming
soon.

I’ve worked with Postgres quite a bit in the past and had occasional but
serious problems with reliability under some circumstances (Geocrawler.com
runs on Postgres and had a tendency to melt down every few months, losing
some data in its 10GB database). So I began evaluation of Postgres 7.1 with
some trepidation. My benchmarks in July showed PG 7.0 to be clearly slower
than MySQL and I outlined several limitations that I found annoying. Still,
Postgres has dozens of advanced features that MySQL lacks (triggers, foreign
keys, rules, subselects, views…) and was well worth investigating.