Re: Planner avoidance of index only scans for partial indexes

Merlin,

Thanks!

> The takeaway is that for this query I'm seeing end to end query
> execution improvement of 58% in the worst case (warm cache) and 20x or
> more in the best case when the cache is cold: that 22 sec time is
> unpredictable and can run into the minutes as the index randomizes the
> i/o to the heap (the table is not clustered on this index nor can it
> be).   This 'best case' is extremely important to us as the generally
> the first thing the user sees when firing up the application the first
> time and with 9.2 it's a 'night and day' experience.  Here is the
> bottom line in 9.1 vs 9.2 IOS:

So I could easily say: "An up to 20X speedup on data warehousing
queries,", no?  Or "A between 2X and 20X speed increase on data
warehousing queries".

Of course, I'd love to see you able to blog the detail ...


-- 
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com