Oops I screwed up again. :)
I was actually right the first time my postgresql 7.0.3 was running with
fsync off. Due to my weird results I searched more thoroughly and found my
7.0.3's pg_options had a nofsync=1.
So 7.0.3 is twice as fast only with fsync off.
7.1beta4 snapshot - fsync.
./pgbench -i pgbench -s 5
./pgbench -c 5 -t 500 pgbench
starting vacuum...end.
transaction type: TPC-B (sort of)
scaling factor: 5
number of clients: 5
number of transactions per client: 500
number of transactions actually processed: 2500/2500
tps = 22.435799(including connections establishing)
tps = 22.453842(excluding connections establishing)
7.0.3 no fsync
./pgbench -i pgbench -s 5
./pgbench -c 5 -t 500 pgbench
starting vacuum...end.
transaction type: TPC-B (sort of)
scaling factor: 5
number of clients: 5
number of transactions per client: 500
number of transactions actually processed: 2500/2500
tps = 52.971997(including connections establishing)
tps = 53.044280(excluding connections establishing)
7.0.3 fsync
./pgbench -c 5 -t 500 pgbench
starting vacuum...end.
transaction type: TPC-B (sort of)
scaling factor: 5
number of clients: 5
number of transactions per client: 500
number of transactions actually processed: 2500/2500
tps = 7.366986(including connections establishing)
tps = 7.368807(excluding connections establishing)
Cheerio,
Link.
