Everyone, what's the consensus on how the site did today?
The histogram looked pretty nice until about half an hour before the close and is just now starting to taper off. Even with that spike, it looked very tolerable, but I'd still like to find some more headroom before re-enabling full-text indexing.
I probably published about a dozen new versions of read_msg throughout the day. It doesn't appear I've caused any outages, and hopefully I didn't cause any problems to crop up in scenarios I didn't test.
I've gone through every line of code up to line 220 so far (out of 339 currently -- was about 400 this morning) in read_msg and held the magnifying glass up to each and every one that has any interaction with the database. Most of the inefficiencies I've found were quite small, but with how frequently the webserver sends queries to the db, it does tend to add up.
In some cases I've revamped routines to run a tiny bit faster and in the most recent change, I completely did away with a separate, very tiny and inexpensive (but remember -- called a hundred or so times per second) query that simply wasn't needed because I already had the info I needed from another query. And I think I'm going to be able to zap another one shortly by adding just one field to the main query.
