Jayant Kumar did some benchmarking of InnoDB vs. PostgreSQL and PG
came out 5 times faster. The benchmark isn't very thoroughly
described, but it turns out not to matter.
http://jayant7k.blogspot.com/2010/09/database-speed-tests-mysql-and.html
Apparently, the reason we're faster is that wal_sync_method =
open_datasync, which is the default on MacOS X, doesn't actually work.
[rhaas pgbench]$ pgbench -t 100000 -j 4 -c 4 pgbench
starting vacuum...end.
transaction type: TPC-B (sort of)
scaling factor: 25
query mode: simple
number of clients: 4
number of threads: 4
number of transactions per client: 100000
number of transactions actually processed: 400000/400000
tps = 1292.258304 (including connections establishing)
tps = 1292.281493 (excluding connections establishing)
Clearly we're not getting 1292 (or even 1292/4) fsync per second out
of whatever HD is in my laptop. So what happens if we change to
fsync_writethrough, which is the equivalent of what InnoDB apparently
does out of the box?
[rhaas pgsql]$ pg_ctl reload
server signaled
LOG: received SIGHUP, reloading configuration files
LOG: parameter "wal_sync_method" changed to "fsync_writethrough"
[rhaas pgbench]$ pgbench -t 100000 -j 4 -c 4 pgbench
starting vacuum...end.
transaction type: TPC-B (sort of)
scaling factor: 25
query mode: simple
number of clients: 4
number of threads: 4
number of transactions per client: 100000
number of transactions actually processed: 400000/400000
tps = 27.845797 (including connections establishing)
tps = 27.845809 (excluding connections establishing)
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company