I'm considering using PostgreSQL in an environment where there is a
middle tier framework running the queries, which has no way of knowing
whether the query it is about to run will return one row or millions of
rows.
When it comes to returning millions of rows, I don't much care about
20us. Some of the one row queries run in as little as 265us (and I'm
thinking of submitting changes which would bring that down), so 20us
might creep into the range of significance. The test which showed a
20us difference were running across the LAN, while the production middle
tier will be running on the same box as the database; so, if much of the
20us is in network latency, it becomes much less significant.
-Kevin
>>> Oliver Jowett <oliver@opencloud.com> 07/19/05 6:38 PM >>>
You are considering using resultset streaming in a query that takes only
~600us (from your 3% number)? How big is that resultset anyway?