"Nicholson, Brad (Toronto, ON, CA)" wrote:
> 500k tps on read and 30k tps on mixed workload of a single server -
> wow... Do you have a comparison for the workload against 9.1? I'm
> curious about the factor of improvement.
Unfortunately, I was only able to grab a few days on this box before
it was put into production, and a large enough battery of tests to
have high confidence in the results took about 20 hours to run. I
focused on the impact of specific proposed patches against current
9.2 development HEAD; unfortunately, I didn't have time to do a run
comparing to the 9.1 production release. :-(
I may be able to get the machine for an occasional window of time on
a few more weekends. With dedicated time on that machine being a
fairly precious resource, I need to pick the tests to run pretty
carefully. Perhaps a 9.1 to 9.2 comparison will be a good one as we
near release time next year. Before that, I'm inclined to think
that any time I can grab would be more valuable evaluating proposed
patches.
With all of that said, I'd bet that Robert Haas has some overall
numbers from the big machine he's been able to use. Robert?
-Kevin