Stefan Moeding <pgsql@moeding.net> wrote:
> Have you thought about keeping the counters for each backend isolated?
> I think in the end it would be beneficial to be able to break down the
> response time for a critical business transaction in isolation instead
> of having all backends in one figure.
I think per-backend profiling is confusable because one backend might be
reused by multiple jobs if you use connection pooling. If we need more
detailed profiling, it should be grouped by query variations.
I have another plan for detailed profiling by extending pg_stat_statements
for such purposes. It'll include functionalities of log_{parser|planner|
executor|statement}_stats parameters. They are log-based profiler, but a
view-based approach seems to be more easy-to-use.
> Do you know the work of Cary Millsap at http://www.method-r.com/ who has
> been working on response time based tuning in Oracle?
I didn't know that. What is the point of the works? Are there some
knowledge we should learn from?
Regards,
---
ITAGAKI Takahiro
NTT Open Source Software Center