I’m back from three days in sunny (yes, really) San Francisco. As usual for a PostgreSQL conference, I had a good time and had my brain nicely melted. A few talk highlights:
Query Planner (R. Haas): When investigating problematic queries, the big thing to look for is differences between your row estimates from EXPLAIN, and the actual rows from EXPLAIN ANALYZE.
PCI Compliance (D. Patel): I want to try this out: http://pgfoundry.org/projects/tablelog/, even though it doesn’t seem to be currently maintained.
pg_upgarde (B. Momjian): 44 seconds to upgrade a 150G database with pg_upgrade.
System Tables (R. Haas): Columns stay in pg_attribute, even if you drop them; the attisdropped field gets set to true. (I have about 5 lines of notes from this talk. I think I was sitting there saying “wooooow” the rest of the time.)
Scaling with Postgres (R. Treat): Perfect relational modeling is not the goal when you’re thinking about scaling.
Temporal Datatypes (J. Davis): Scheduling conflicts will be enforced in real life if they are not enforced in the database.
Of course, the hallway track was where I really learned the most. I got some good comments on my talk, which I’ll be incorporating into its next iteration. I also met a lot of new people, mostly SF locals, and I finally got to meet Dave Page, Robert Haas, Jignesh Shah, and the elusive Greg Stark.
There were some glitches this time: there were no lightning talks ;), room and schedule updates weren’t communicated effectively to attendees, and we had a few speaker no-shows (and no apparent backup plan). I really want to thank Rob Treat for stepping up & filling one of the empty slots!