Re: vacuum, performance, and MVCC - Mailing list pgsql-hackers

From Csaba Nagy
Subject Re: vacuum, performance, and MVCC
Date
Msg-id 1151074993.3309.184.camel@coppola.muc.ecircle.de
Whole thread Raw
In response to Re: vacuum, performance, and MVCC  ("Zeugswetter Andreas DCP SD" <ZeugswetterA@spardat.at>)
List pgsql-hackers
First of all, I base my assumptions on what I recall to have read on
this list, as I didn't try yet partitioning using inheritance. It's not
trivial to set up and I didn't have the time to play with it yet. So I
wouldn't know for sure that it won't work fine with our application, and
that will only change when I'll get a few days to experiment. The
experimentation will include the migration of existing data to the
partitioned schema, which will be probably the most difficult part of it
due to the size of the tables which need partitioning...

> You would query the parent (no union). Do you need order by's ?
> Without order by it is currently no problem.

It's clear to me that partitioning by inheritance is transparent to the
application, what worries me is that our application likely has a few
queries which will be equivalent to a union when planning, and I fear
bad performance there.

An I need order by on all queries with limit. The few exceptions where I
wouldn't need order by are when I want to delete/update chunk-wise, but
that's not supported right now... another feature I made noise about ;-)

[snip]

Cheers,
Csaba.




pgsql-hackers by date:

Previous
From: James Robinson
Date:
Subject: Webcluster session storage, was vacuum, performance, and MVCC
Next
From: Tzahi Fadida
Date:
Subject: Re: Planning without reason.