> > Have you looked at sqlrealy.sourceforge.net? IT looks like it might do
> > what you need.
>
> SQL Relay (and friends) do what I'm looking for in a round about way.
>
> If you put it onto the webservers it would help -- but it would require
> deployment of additional webservers to accommodate the increased load.
> That can be accomplished if it helps drop the load on the DB machine.
> But still uses resources unnecessarily.
>
> I've not looked at sqlrelay but most of these things use a different
> interface. That would work with the inhouse code but puts a damper on
> the commercial software.
>
> As a temporary step these types of things help. But it's still doesn't
> really fix the problem of Apache not using real connection pooling.
You can try pgpool
(http://www2b.biglobe.ne.jp/~caco/pgpool/pgpool-1.2.tar.gz) if you
like. pgpool is a connection pool server for PostgreSQL. Pgpool
pretends as if PostgreSQL and you do not need to change applications
to use a special interface like SQL Relay.
Moreover, pgpool uses pre-fork technique to reduce overhead. Here is a
benchmark result using pgbench -S -C -c 10 -t 100 performed on my note pc:
Normal PostgreSQL 7.3.6: 25.6 TPS
with pgpool 1.2: 36.1 TPS
As you can see even with some overhead with pgpool, I got 40%
improvement over normal PostgreSQL using pgpool. In this case
PostgreSQL, pgpool and pgbench are running on same machine BTW.
Another point with pgpool is it can be run in replication mode.
--
Tatsuo Ishii
