>
> I saw some similar results comparing 9.0 and 9.2 pgbench tests. My tests were on a VM, but on a dedicate host/hardware with no other VM's running on it to minimize variables. I didn't have a lot of time to dig into it, but I do recall seeing more lock contention on updates on the 9.2 instance though.
Ah, good point. At that scale, Colin's test is mostly a contention
test. There's something there worth investigating, but it's not a
realistic use case.
Colin, please try your pgbench tests with -s 100.
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com