Re: Checkpoint cost, looks like it is WAL/CRC

Tom,

> This will remove just the CRC calculation work associated with backed-up
> pages.  Note that any attempt to recover from the WAL will fail, but I
> assume you don't need that for the purposes of the test run.

Looks like the CRC calculation work isn't the issue. I did test runs of
no-CRC vs. regular DBT2 with different checkpoint timeouts, and didn't
discern any statistical difference. See attached spreadsheet chart (the
two different runs are on two different machines).

I think this test run http://khack.osdl.org/stp/302903/results/0/, with a
30-min checkpoint shows pretty clearly that the behavior of the
performance drop is consistent with needing to "re-prime" the WAL will
full page images. Each checkpoint drops performance abruptly, and then
slowly recovers until the next checkpoint.

Do note that there is a significant statistical variation in individual
runs. It's only the overall trend which is significant.

--
--Josh

Josh Berkus
Aglio Database Solutions
San Francisco