d11wtq wrote:Before I start, I'm looking for a discussion, not a flame-war, although I'm not sure this lends itself to flaming in any case

Anything is capable of flames.. but hopefully this one won't be too bad. For my part, I will try to focus on light, not heat.
d11wtq wrote:I think it'd be interesting to read what your opinions on the pros and cons of various DBMS's are in the real world.
I have solid, extensive experience in PGSQL, MySQL, and MSSQL. I've dabbled in others, but those are my focal areas.
I find PGSQL to be the best balance. It is much more compliant to the SQL standard than others. It is easy to code for, and by default it has support for the things I need (transactions, atomic commits, row-locking, utf-8 support). The speed is (imho) not quite as good as MySQL (at least in my testing it isn't for several key areas I looked at), although it gets *dramatically* better every release. Its also under a sexy opensource license, developed openly, well-documented, and - as if that wasnt enough - its free.
MySQL is easily my second choice. It uses some fugly SQL-hacks, but it does so to help speed the engine up. Its the most common db used, so its easy to copy/learn code for it. The speed is industry-leading, although it seems to be decreasing with each new major release (again, at least in my testing, for several key areas I looked at). Its license caused some arguing and difficulties for PHP (which has its own license issues), but that has been settled for the most part. It is reasonably well-documented (although I find the documentation often glosses over MAJOR issues, and is poorly written), and its free.
MS-SQL is a blight upon humanity. It requires the sacrifice of small children to accomplish simple tasks, and it eats kittens and puppies just to keep running. It is only available under the black moon, in exchange for your first born. Avoid it at all costs, lest ye be judged.
d11wtq wrote:
One thing I'm particularly interested in is the server load. Postgres is drastically lower on memory usage from what I've seen so far in comparison to MySQL which guzzles at my RAM. That's a big plus when I'm running a VDS server where costs go up if I require more RAM allocation.
Without further ado.... discuss!
I suspect you are getting skewed results from a small sample size. Try increasing the number of simultaneous attempts to 30-40. Then you should see the two use fairly similar amounts of cpu and ram (at least I did). MySQL is simply more aggressive by default about its cache, memory use, and pre-emptive access patterns. PGSQL tries harder to be "per-use", and as a result does (far) better on memory use at the low-end, while being slower to return results. In my experience, they change places at higher levels.. PGSQL (in my experience!) does extremely well with heavy loads, while mysql needs considerable tuning to keep up.