GEQO is an attempt to provide a near-optimal join order without using an exhaustive search. "An exhaustive, deterministic search of a subset of the search space" has a non-zero probability of finding only a local minimum in execution time.
I'm not sure what you mean. By an "exhaustive, deterministic search of a subset of the search space", I was referring to using the normal planner, but restricting the search space to only left-deep plans. Since GEQO will also only consider left-deep plans, ISTM there is no issue of "local minima" that does not apply to an equal degree to GEQO itself.
Since purely random selection, with learning, works so well, it may be worth developing a new random join-order optimization algorithm
Yeah, I agree. I've read a few papers on randomized algorithms for join order selection, but none of them really seemed to be clear winners. Do you have a reference for the paper you read?
-Neil
---------------------------(end of broadcast)--------------------------- TIP 1: subscribe and unsubscribe commands go to [EMAIL PROTECTED]
