"David W. Van Couvering" <[EMAIL PROTECTED]> writes:

> There appear to be definite improvements here in terms of the
> consistency of throughput and CPU usage.
>
> Are the regular massive spikes in throughput and CPU usage related to
> the issues that Oystein was raising about checkpointing?

Thanks for committing, David.

Yes, the spikes are related to Øystein's checkpointing issues. Every
time a checkpoint is initiated both throughput and CPU usage drop. If
the graphs had shown throughput per client, you would probably have
seen long periods where one or a couple of clients had very low or
zero throughput (I actually logged a transaction with 86 seconds
response time when testing the patch, so there's some serious
thrashing going on).

-- 
Knut Anders

Reply via email to