Disk-based hash aggregate's cost model

From: Peter Geoghegan <pg(at)bowt(dot)ie>
To: Jeff Davis <pgsql(at)j-davis(dot)com>
Cc: PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>, Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>
Subject: Disk-based hash aggregate's cost model
Date: 2020-08-28 00:28:20
Message-ID: CAH2-Wzk1-JiW2heHg8LqJzV5Q4VtOBj7QDXTESd_o0Vzo+_Etw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

We have a Postgres 13 open item for Disk-based hash aggregate, which
is the only non-trivial open item. There is a general concern about
how often we get disk-based hash aggregation when work_mem is
particularly low, and recursion seems unavoidable. This is generally
thought to be a problem in the costing.

Tomas Vondra did some testing of the patch which led to discussion of
this on the hash agg GUC megathread, here:

https://www.postgresql.org/message-id/20200724012248.y77rpqc73agrsvb3@development

I am starting this new thread to deal with the open item.

Any progress on this, Jeff? Please correct or expand on my summary of
the open item if I got something wrong.

Thanks
--
Peter Geoghegan

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Stephen Frost 2020-08-28 00:30:23 New default role- 'pg_read_all_data'
Previous Message Ranier Vilela 2020-08-28 00:22:47 [PATCH] Explicit null dereferenced (src/backend/access/heap/heaptoast.c)