Re: WIP: cross column correlation ...

On Sun, Feb 27, 2011 at 5:17 PM, Josh Berkus <josh(at)agliodbs(dot)com> wrote:
>
>> I think there would be value in giving the DBA an easier way to see
>> which tables are hot, but I am really leery about the idea of trying
>> to feed that directly into the query planner.  I think this is one of
>> those cases where we let people tune it manually for starters, and
>> then wait for feedback.  Eventually someone will say "oh, I never tune
>> that by hand any more, ever since I wrote this script which does the
>> following computation... and I just run it out cron".  And then we
>> will get out the party hats.  But we will never get the experience we
>> need to say what that auto-tuning algorithm will be unless we first
>> provide the knob for someone to fiddle with manually.
>
> I'm not disagreeing with that.  I'm saying "first, we give DBAs a way to
> see which tables are currently hot".  Such a feature has multiple
> benefits, making it worth the overhead and/or coding effort.
>
> Whether we're shooting for autotuning or manual tuning, it starts with
> having the data.

Well, what we have now is a bunch of counters in pg_stat_all_tables
and pg_statio_all_tables. Making that easier for the DBA almost seems
like more of a job for a third-party tool that, say, graphs it, than a
job for PG itself. But if you have an idea I'm ears.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company