Re: Load Distributed Checkpoints, take 3

On Mon, 25 Jun 2007, Heikki Linnakangas wrote:

> Please describe the class of transactions and the service guarantees so
> that we can reproduce that, and figure out what's the best solution.

I'm confident you're already moving in that direction by noticing how the
90th percentile numbers were kind of weird with your 150 warehouse DBT2
tests, and I already mentioned how that could be usefully fleshed out by
more tests during beta.  That number is the kind of service guarantee I'm
talking about--if before 90% of transactions were <4.5ms, but now that
number is closer to 6ms, that could be considered worse performance by
some service metrics even if the average and worst-case performance were
improved.

The only thing I can think of if you wanted to make the problem more like
what I was seeing would be switching the transaction mix on that around to
do more UPDATEs relative to the other types of transactions; having more
of those seemed to aggrevate my LDC-related issues because they leave a
different pattern of dirty+used buffers around than other operations.

--
* Greg Smith gsmith@gregsmith.com http://www.gregsmith.com Baltimore, MD