Hi,
On 12/23/2015 03:38 PM, Robert Haas wrote:
>
> I think one thing that this conversation exposes is that the size of
> the working set matters a lot. For example, if the workload is
> pgbench, you're going to see a relatively short FPW-related spike at
> scale factor 100, but at scale factor 3000 it's going to be longer
> and at some larger scale factor it will be longer still. Therefore
> you're probably right that 1.5 is unlikely to be optimal for
> everyone.
Right.
Also, when you say "pgbench" you probably mean the default uniform
distribution. But we now have gaussian and exponential distributions
which might be handy to simulate other types of workloads.
>
> Another point (which Jan Wieck made me think of) is that the optimal
> behavior here likely depends on whether xlog and data are on the same
> disk controller. If they aren't, the FPW spike and background writes
> may not interact as much.
I'm not sure what exactly you mean by "optimal behavior" here. Surely if
you want to minimize interference between WAL and regular I/O, you'll do
that.
But I don't see what that has to do with the writes generated by the
checkpoint? If we do much more writes at the beginning of the checkpoint
(due to getting confused by FPW), and OS starts flushing that to disk
because we exceed dirty_(background)_bytes, that surely interferes with
reads (which is a major issue for queries).
regards
--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services