Josh Berkus <josh(at)agliodbs(dot)com> writes:
>> Um, where are the test runs underlying this spreadsheet? I don't have a
>> whole lot of confidence in looking at full-run average TPM numbers to
>> discern whether transient dropoffs in TPM are significant or not.
> Web in the form of:
> Where #test_number# is:
> Machine0, no patch:
> Machine0, patch:
> Machine2, no patch:
> Machine2, patch:
Hmm. Eyeballing the NOTPM trace for cases 302912 and 302909, it sure
looks like the post-checkpoint performance recovery is *slower* in
the latter. And why is 302902 visibly slower overall than 302905?
I thought for a bit that you had gotten "patch" vs "no patch" backwards,
but the oprofile results linked to these pages look right: XLogInsert
takes significantly more time in the "no patch" cases.
There's something awfully weird going on here. I was prepared to see
no statistically-significant differences, but not multiple cases that
seem to be going the "wrong direction".
BTW, I'd like to look at 302906, but its [Details] link is broken.
regards, tom lane
In response to
pgsql-hackers by date
|Next:||From: Josh Berkus||Date: 2005-07-22 23:29:57|
|Subject: Re: Checkpoint cost, looks like it is WAL/CRC|
|Previous:||From: Alvaro Herrera||Date: 2005-07-22 22:54:32|
|Subject: Re: Autovacuum loose ends|