Re: full text search to_tsquery performance with ispell dictionary

On 11.05.11 17:31, "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

>You really, really, really need to fix whatever is preventing you from
>using pooling. Opening a database connection to run one query is just
>horridly inefficient.

Very true. I did not mean that anything actually prevents us from using
pooling. We just have no idea, how it will interfere with our productive
pgcluster setup. I imagine the evaluation, testing and verification of
pooling systems in combination with our setup to be quite tedious.

Of course, I don't open a connection for each query. One is opened for
each service call. The services are designed to start an own process for
every call. Such a process - for now - needs an own connection. It usually
handles dozens of queries. Until now, we never ran into performance
problems. The time consumed by DB operations is usually negligible
compared to the rest.

First tests with a simple persistent connection setup seem to work fine
and solve the performance issue. I tend to put some thorough testing on
this setup and see if I step into a pitfall.

Big kudos to you and this list. You were a great help, as always.