From: | Chris Hoover <revoohc(at)gmail(dot)com> |
---|---|
To: | PGSQL Performance <pgsql-performance(at)postgresql(dot)org> |
Subject: | Estimating hot data size |
Date: | 2011-02-16 20:51:36 |
Message-ID: | AANLkTikxfB099wX3TtnXvKPoFLCcXtMCYpNvCSygpNPb@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
All,
I'm trying to estimate the size of my hot data set, and wanted to get some
validation that I'm doing this correctly.
Basically, I'm using the sum(heap_blks_read + idx_blks_read) from
pg_statio_all_tables, and diffing the numbers over a period of time (1 hour
at least). Is this a fair estimate? The reason for doing this is we are
looking at new server hardware, and I want to try and get enough ram on the
machine to keep the hot data in memory plus provide room for growth.
Thanks,
Chris
Example:
*Time*
*Total Blocks*
2011-02-16 11:25:34.621874-05
123,260,464,427.00
2011-02-16 12:25:46.486719-05
123,325,880,943.00
To get the hot data for this hour (in KB), I'm taking:
(123,325,880,943.00 - 123,260,464,427.00)* 8 = 523,332,128KB
Correct?
From | Date | Subject | |
---|---|---|---|
Next Message | Tomas Vondra | 2011-02-16 21:13:41 | Re: Estimating hot data size |
Previous Message | Greg Smith | 2011-02-16 20:36:01 | Re: high user cpu, massive SELECTs, no io waiting problem |