Re: Feature suggestions (long)

From: "Zeugswetter Andreas SB SD" <ZeugswetterA(at)spardat(dot)at>
To: "Martijn van Oosterhout" <kleptog(at)svana(dot)org>, <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Feature suggestions (long)
Date: 2003-05-19 08:46:04
Message-ID: 46C15C39FEB2C44BA555E356FBCD6FA4961FB2@m0114.s-mxs.net
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

> Partitions
> ==========

> Next stage would be teaching the planner. The conditions would be
> pseudo-constraints on the partitions. Hence if the conditions and the
> constraints form a non-intersecting set, you can skip that partition
> altogether.

Make that "normal check constraints", and make the planner consider
constraints,
and I think that by itself combined with the current featureset will
be much more powerful than any of the "partitioning" features out there.
(This is mainly needed to optimize selects on the big union all view)

Imho if a dba starts to partition, he usually needs to be more involved
than the average user, so I think he should be able cope with compexity.
What imho would help, is a tool that generates a suggested rule set,
indexes
and actions, which the dba can review and apply. I do not think new SQL
syntax
would really help, since that would somehow hide the great existing
power of
the rule system. A tool would teach the dba, and empower him to use it.

And yes, creating several smaller tables and adding the appropriate
rules
usually makes the VLDB life much easier compared to growing single
tables into
the TB range.

Andreas

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Martijn van Oosterhout 2003-05-19 09:32:11 Re: Feature suggestions (long)
Previous Message Mark Kirkwood 2003-05-19 07:55:07 Re: Feature suggestions (long)