Re: patch submission: truncate trailing nulls from heap rows to reduce the size of the null bitmap [Review]

From: Amit Kapila <amit(dot)kapila(at)huawei(dot)com>
To: "'Andres Freund'" <andres(at)2ndquadrant(dot)com>
Cc: "'Craig Ringer'" <craig(at)2ndQuadrant(dot)com>, "'Jameison Martin'" <jameisonb(at)yahoo(dot)com>, "'Tom Lane'" <tgl(at)sss(dot)pgh(dot)pa(dot)us>, "'Noah Misch'" <noah(at)leadboat(dot)com>, "'Simon Riggs'" <simon(at)2ndQuadrant(dot)com>, "'Kevin Grittner'" <kgrittn(at)mail(dot)com>, <robertmhaas(at)gmail(dot)com>, <josh(at)agliodbs(dot)com>, <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: patch submission: truncate trailing nulls from heap rows to reduce the size of the null bitmap [Review]
Date: 2013-06-24 14:23:39
Message-ID: 002e01ce70e6$6c861a80$45924f80$@kapila@huawei.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Monday, June 24, 2013 5:48 PM Andres Freund wrote:
> On 2013-01-25 09:06:12 +0530, Amit Kapila wrote:
> > On 01/25/2013 03:43 AM, Jameison Martin wrote:
> > > there have been a lot of different threads on this patch. i'm going
> to
> > take a stab at > teasing them out, summarizing them, and addressing
> them
> > individually.
> >
> > > Is this patch on the CF app? I can't seem to find it in 2013-01 or
> > 2013-next, and I
> > > wanted to add your summary.
> >
> > It is in previous CF (2012-11) in Returned with Feedback section
> > https://commitfest.postgresql.org/action/commitfest_view?id=16
>
> I'd argue that the patch ought to be still marked as "Returned with
> Feedback" instead of again being marked as "Needs Review". I don't
> really see anything new that has come up?

You are right that nothing new has come up. The major reason why this patch
could not go into 9.3 was that it is not clearly evident whether
Performance gains of this patch are sufficient to take risks (Tom points out
that bugs caused by such changes can be difficult to identify) of committing
this code.

I will summarize the results, and if most of us feel that they are not good
enough, then we can return this patch.

Observations from Performance Results for tables with less columns
--------------------------------------------------------------------
1. Approximately 13% space savings for table with 12 columns, out of which
last 4 are Nulls.
This table schema is such first 3 columns are integers and last 9 are
boolean's.

Refer below link for detailed results:
http://www.postgresql.org/message-id/00b401cde1d8$746c90f0$5d45b2d0$@kapila@
huawei.com

Observations from Performance Results for tables with large columns
--------------------------------------------------------------------
1. There is a visible performance increase when number of columns containing
NULLS are more than > 60~70% in table have large number of columns.
Approximately 12% for table (having 800 cols, out of which 600 are nulls
OR having 300 columns, out of which 250 are nulls)
2. There are visible space savings when number of columns containing NULLS
are more than > 60~70% in table have large number of columns.
Approximately 11% for table (having 800 cols, out of which 600 are nulls
OR having 300 columns, out of which 250 are nulls)

Refer below link for detailed results:
http://www.postgresql.org/message-id/6C0B27F7206C9E4CA54AE035729E9C382853AA4
0(at)szxeml509-mbs

Observation from original pgbench
----------------------------------

1. There is < 1% performance dip for original pgbench, this could be due to
fluctuation in readings or some consequences of addition of code which is
difficult to reason out.
Refer below link for detailed results:
http://www.postgresql.org/message-id/00bc01cde1f7$be4b4cb0$3ae1e610$@kapila@
huawei.com

With Regards,
Amit Kapila.

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Kevin Grittner 2013-06-24 14:29:43 Re: changeset generation v5-01 - Patches & git tree
Previous Message Andres Freund 2013-06-24 14:17:52 Re: [Review] Re: minor patch submission: CREATE CAST ... AS EXPLICIT