| From: | Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> |
|---|---|
| To: | Robert Haas <robertmhaas(at)gmail(dot)com>, Jeff Davis <pgsql(at)j-davis(dot)com> |
| Cc: | David Rowley <david(dot)rowley(at)2ndquadrant(dot)com>, Andres Freund <andres(at)anarazel(dot)de>, Heikki Linnakangas <hlinnaka(at)iki(dot)fi>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: Spilling hashed SetOps and aggregates to disk |
| Date: | 2018-06-11 15:38:09 |
| Message-ID: | dfe0ba61-93e1-e7ee-d0d2-2e9f23c65919@2ndquadrant.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On 06/11/2018 04:25 PM, Robert Haas wrote:
>
> ...
>
> Maybe that's not exactly what Tomas (or you) meant by eviction
> strategy, though. Not sure. But it does seem to me that we need to
> pick the algorithm we're going to use, more or less, in order to
> decide what infrastructure we need, and at least to some extent, we
> ought to let our choice of algorithm be informed by the desire not to
> need too much infrastructure.
>
I was using eviction strategy in a somewhat narrower sense - pretty much
just "Which groups to evict from the hash table?" but you're right the
question is more general and depends on which scheme we end up using
(just hashagg, hash+sort, something else ...)
regards
--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Alvaro Herrera | 2018-06-11 15:51:05 | Re: CF bug fix items |
| Previous Message | Peter Eisentraut | 2018-06-11 15:31:23 | Re: file cloning in pg_upgrade and CREATE DATABASE |