| From: | Cédric Villemain <cedric(dot)villemain(dot)debian(at)gmail(dot)com> |
|---|---|
| To: | Konrad Garus <konrad(dot)garus(at)gmail(dot)com> |
| Cc: | pgsql-performance(at)postgresql(dot)org |
| Subject: | Re: shared_buffers advice |
| Date: | 2010-05-27 10:21:26 |
| Message-ID: | AANLkTinuS4hWw-Pg6RFpzsMo9qQj7zC-FulEgtn83Gkr@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
2010/5/27 Konrad Garus <konrad(dot)garus(at)gmail(dot)com>:
> 2010/5/27 Cédric Villemain <cedric(dot)villemain(dot)debian(at)gmail(dot)com>:
>
>> It works thanks to mincore/posix_fadvise stuff : you need linux.
>> It is stable enough in my own experiment. I did use it for debugging
>> purpose in production servers with succes.
>
> What impact does it have on performance?
pgmincore() and pgmincore_snapshot() both are able to mmap up to 1GB.
I didn't mesure a performance impact. But I haven't enough benchmarks/test yet.
>
> Does it do anything, is there any interaction between it and PG/OS,
> when it's not executing a command explicitly invoked by me?
pgfincore does nothing until you call one of the functions.
Reducing the mmap window is faisable, and I had start something to use
effective_io_concurrency in order to improve prefetch (for restore)
but this part of the code is not yet finished.
>
> --
> Konrad Garus
>
--
Cédric Villemain 2ndQuadrant
http://2ndQuadrant.fr/ PostgreSQL : Expertise, Formation et Support
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tom Lane | 2010-05-27 14:09:53 | Re: About Tom Lane's Xeon CS test case |
| Previous Message | Konrad Garus | 2010-05-27 08:59:50 | Re: shared_buffers advice |