3 ms·
This is a fine basic intro to the topic, but the whole argument is hung on a very sketchy benchmark (as noted by the article itself in a footnote). Some things
by karlmdavis 8y ago
This is a fine basic intro to the topic, but the whole argument is hung on a very sketchy benchmark (as noted by the article itself in a footnote).
Some things that'd help improve the argument:
1. Account for table/row locking in the benchmark: is it a factor or not?
2. Re-run the benchmark with for some of the proposed solutions, e.g. local connection pooling and PgBouncer.
- aidos 8y agoI didn't know about this global locking but I'm assuming the row / table locking is taken into account in the testing. Presumably just inserting into a simple table with explicit unique ids each time should avoid issues with locking (I guess you'd want to avoid indexes and sequences).