Re: Inheritance efficiency

From: Vincenzo Romano <vincenzo(dot)romano(at)notorand(dot)it>
To: Alvaro Herrera <alvherre(at)commandprompt(dot)com>
Cc: David Fetter <david(at)fetter(dot)org>, Bruce Momjian <bruce(at)momjian(dot)us>, PostgreSQL General <pgsql-general(at)postgresql(dot)org>
Subject: Re: Inheritance efficiency
Date: 2010-04-30 16:52:11
Message-ID: s2x3eff28921004300952i28259025n85aa2c52f8dba5dc@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

2010/4/30 Alvaro Herrera <alvherre(at)commandprompt(dot)com>:
> Vincenzo Romano wrote:
>
>> In this specific case, if you think about "inheritance for
>> partitioning" and you stick with the example idea of "one partition
>> per month", then the current solution is more than OK.
>> In the real world, that is not really the general case, especially in
>> the "enterprise grade" world, where maybe you partition with both a
>> time stamp and another column, like product code ranges and prefixes
>> ...
>>
>> Is there any planning about this improvement?
>
> Of course.  People is always looking to make improvements in many areas.
> There are very few things that people consider to be "more than OK".
> The partitioning features are among those being more examined for
> possibly improvements.
>
> This does *not* mean that PostgreSQL doesn't serve mission critical
> systems already, on enterprises large and small, some of them on very
> large systems.  What you see in these lists (people describing
> "partition by month" schemes) are not necessarily the most complex
> setups out there.

Hi.
I've nerver meant to say that PG is not mission critical!
I argued that O(n) stuff will keep it away from "enterprise grade" applications.
I've been told earlier that "It is fine for dozens of child tables,
but not thousands;
it does need improvement."

This is not enterprise grade.
And the same could go for (a large number of) partial indexes.
Any idea here?

Infact I have in mind also a different approach to partitioning which
could be useful (under certain constraints, of course).
Instead of partitioning the table itself, you can partition the indexes.
The data can still be in a single table (for the sake of some FKs for example).
Just the indexes get "partitioned"·
But, of course, a lot depends on whether the selection of the right indexes
(among thousands) is effective or not.

>
> --
> Alvaro Herrera                                http://www.CommandPrompt.com/
> PostgreSQL Replication, Consulting, Custom Development, 24x7 support
>

--
Vincenzo Romano
NotOrAnd Information Technologies
cel. +39 339 8083886 | gtalk. vincenzo(dot)romano(at)notorand(dot)it
fix. +39 0823 454163 | skype. notorand.it
fax. +39 02 700506964 | msn. notorand.it
NON QVIETIS MARIBVS NAVTA PERITVS

In response to

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Mario Wojcik 2010-04-30 16:53:02 Re: Nuevo sobre PGday Latinoamericano 2011...
Previous Message Glyn Astill 2010-04-30 16:49:15 Re: pg_restore: [custom archiver] dumping a specific TOC data block out of order is not supported without ID on this input stream (fseek required)