Postgres Pro
Downloads
Home   >   mailing lists

Re: [REVIEW] Re: Compression of full-page-writes - Mailing list pgsql-hackers

From Michael Paquier
Subject Re: [REVIEW] Re: Compression of full-page-writes
Date
Msg-id CAB7nPqR651FJHGOW4iiZy9-pPj_5MqqmmuC63MHYXRAGz11MhQ@mail.gmail.com
Whole thread Raw
In response to Re: [REVIEW] Re: Compression of full-page-writes  (Michael Paquier <michael.paquier@gmail.com>)
Responses Re: [REVIEW] Re: Compression of full-page-writes  (Rahila Syed <rahilasyed90@gmail.com>)
Re: [REVIEW] Re: Compression of full-page-writes  (Fujii Masao <masao.fujii@gmail.com>)
List pgsql-hackers
Tree view


On Wed, Dec 17, 2014 at 12:00 AM, Michael Paquier <michael.paquier@gmail.com> wrote:
Actually, the original length of the compressed block in saved in PGLZ_Header, so if we are fine to not check the size of the block decompressed when decoding WAL we can do without the hole filled with zeros, and use only 1 bit to see if the block is compressed or not.
And.. After some more hacking, I have been able to come up with a patch that is able to compress blocks without the page hole, and that keeps the WAL record length the same as HEAD when compression switch is off. The numbers are pretty good, CPU is saved in the same proportions as previous patches when compression is enabled, and there is zero delta with HEAD when compression switch is off.

Here are the actual numbers:
           test_name           | pg_size_pretty | user_diff | system_diff
-------------------------------+----------------+-----------+-------------
 FPW on + 2 bytes, ffactor 50  | 582 MB         | 42.391894 |    0.807444
 FPW on + 2 bytes, ffactor 20  | 229 MB         | 14.330304 |    0.729626
 FPW on + 2 bytes, ffactor 10  | 117 MB         |  7.335442 |    0.570996
 FPW off + 2 bytes, ffactor 50 | 746 MB         | 25.330391 |    1.248503
 FPW off + 2 bytes, ffactor 20 | 293 MB         | 10.537475 |    0.755448
 FPW off + 2 bytes, ffactor 10 | 148 MB         |  5.762775 |    0.763761
 FPW on + 0 bytes, ffactor 50  | 582 MB         | 42.174297 |    0.790596
 FPW on + 0 bytes, ffactor 20  | 229 MB         | 14.424233 |    0.770459
 FPW on + 0 bytes, ffactor 10  | 117 MB         |  7.057195 |    0.584806
 FPW off + 0 bytes, ffactor 50 | 746 MB         | 25.261998 |    1.054516
 FPW off + 0 bytes, ffactor 20 | 293 MB         | 10.589888 |    0.860207
 FPW off + 0 bytes, ffactor 10 | 148 MB         |  5.827191 |    0.874285
 HEAD, ffactor 50              | 746 MB         | 25.181729 |    1.133433
 HEAD, ffactor 20              | 293 MB         |  9.962242 |    0.765970
 HEAD, ffactor 10              | 148 MB         |  5.693426 |    0.775371
 Record, ffactor 50            | 582 MB         | 54.904374 |    0.678204
 Record, ffactor 20            | 229 MB         | 19.798268 |    0.807220
 Record, ffactor 10            | 116 MB         |  9.401877 |    0.668454
(18 rows)

The new tests of this patch are "FPW off + 0 bytes". Patches as well as results are attached.
Regards,
--
Michael
Attachment

pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: Commitfest problems
Next
From: Heikki Linnakangas
Date:
Subject: Re: Too strict check when starting from a basebackup taken off a standby