Re: [HACKERS] Built-in connection pool

2010-06-28 Thread Kevin Grittner
I wrote: > When preparing to deal with a new statement: > - acquire lockX > - if not a superuser > - and not in an active transaction > - and countX >= max_active_transactions > - place current process at tail of waitX queue, and block > - (lockX would be released while blocked) > - in

[HACKERS] Built-in connection pool

2010-06-28 Thread Kevin Grittner
The thread on Admission Control caused me to realize that one policy I'd been considering as a possibly useful second tier (execution) admission control policy was actually more appropriate as a first tier (simple) admission control policy. The difference (as defined in the Hellerstein, Stonebrake