Re: pg_stat_statements

On Mon, Jun 16, 2008 at 10:34 AM, Robert Treat
<xzilla(at)users(dot)sourceforge(dot)net> wrote:
> Given that PostgreSQL relies on the operating
> system for a majority of it's instermentation (ie. we have nothing like v$
> tables in oracle), we should really be thinking of dtrace as the ultimate
> tool for DBA's to figure out what is going on in thier systems.

Over a year ago, we added an Oracle wait-event-like interface
(counters and timers) to the core of Advanced Server which not only
provides everything DTrace currently does, but is also cross-platform
and is < 300 lines of code in the core.

While DTrace has a little less overhead due to its use of a user-mode
high-resolution timer, neither benchmarks or customers have found our
implementation to cause any performance degradation when timing is
enabled.

It wouldn't be too hard to write the probes in such a way as they
could be used by DTrace or by a loadable timing/counter implementation
for platforms which don't support DTrace.

--
Jonah H. Harris, Sr. Software Architect | phone: 732.331.1324
EnterpriseDB Corporation | fax: 732.331.1301
499 Thornall Street, 2nd Floor | jonah(dot)harris(at)enterprisedb(dot)com
Edison, NJ 08837 | http://www.enterprisedb.com/