RE: Adding percentile metrics to pg_stat_statements module

Hello

While looking at slow queries on pg_stat_statements. I was looking for percentile fields..

If we are worried about CPU cost, maybe it could be useful to turn in on when you have a high stddev_exec_time for the query ?

Regards,

________________________________
De : Adrien Nayrat <adrien(dot)nayrat(at)anayrat(dot)info>
Envoyé : samedi 2 novembre 2019 10:23:49
À : Tomas Vondra; Igor Calabria
Cc : pgsql-hackers(at)postgresql(dot)org
Objet : Re: Adding percentile metrics to pg_stat_statements module

On 10/31/19 8:32 PM, Tomas Vondra wrote:
> IMO having some sort of CDF approximation (being a q-digest or t-digest)
> would be useful, although it'd probably need to be optional (mostly
> becuase of memory consumption).

+1, I like this idea. If we are afraid of CPU cost we could imagine some kind of
sampling or add the possibility to collect only for a specific queryid.

I dreamed of this kind of feature for PoWA. Thus, it could make possible to
compare CDF between two days for example, before and after introducing a change.

Regards,

--
Adrien NAYRAT