"Bruce Momjian" <bruce@momjian.us> writes:
> Where are we on this?
Well Simon just sent the reworked patch yesterday so the answer is we haven't
started tuning this parameter. (Bruce's message is referring to the discussion
about what the optimal value of lsns per clog page would be.)
I intend to do some benchmarking on it and testing what the best value of this
parameter is one of the things I aim to determine with these benchmarks.
I'm not 100% sure yet what to measure though. There are two questions here:
1) What are some good workloads to test which will require larger values for this parameter or which will be hurt by
excessivelylarge values?
I think any short-transaction workload should be basically good enough. I'm
thinking of using just pgbench's default workload. Does anyone think there are
other workloads that we need to specifically test?
2) What metric do I use to determine if the value is large enough or too large?
The gross measurement would be to look at tps. I would prefer to actually
count events which we want to minimize such as xlogflushes because the clog
lsn is too old and how much extra work the visibility checks do. I'm not sure
exactly how much of this we can really measure though. Are there any other
events having an insufficiently large or excessively large value of this
parameter will cause which we can count?
-- Gregory Stark EnterpriseDB http://www.enterprisedb.com
