Re: performance of count(*)

On May 6, 2011, at 4:15 PM, Andrew Sullivan wrote:

> Yeah, in that case the HOT suggestions are very important. I strongly
> recomment you experiment in a test system with real data and
> pathological cases in particular, in order to see what happens when
> the outlier cases inevitably, Murphy willing, crop up. That's not to
> say you should arrange your plans for them, but forewarned is
> forearmed.

Again thanks. The HOT tip led me down the road of paying attention to my indexes, which led me to a nice realization about how to shrink the overall footprint of the materialized aggregates ;-) Which led me to a technique to seriously minimize updates...

I didn't have to worry about bloat too much--overall activity level is not huge; the possibility of collisions on updates is mostly because users tend to work on the same very small (but ever-shifting) subset of the data at the same time, but now I think I'm really set!

--
Scott Ribe
scott_ribe(at)elevated-dev(dot)com
http://www.elevated-dev.com/
(303) 722-0567 voice