From: Danny Wyatt (danny@cs.washington.edu)
Date: Mon May 24 2004 - 09:31:35 PDT
This paper covers much of what we have seen in class over the past few
lectures: the creation of many plans for a query, the computation of
costs for ordering within the trees and for the physical operators used
at each node, and how these plans and operators interact with other
facets of the query and database such as aggregation and indices. They
also define metrics for which statistics can be kept and out of which
operator costs can be estimated.
I like the addition of "interesting orders" to their dynamic programming
approach. I hadn't thought about multidimensional optimality in such
problems before. I don't have much to comment on this paper beyond
that. It is interesting from a historical perspective, and I am curious
to find out how their estimated selectivity factors have been refined by
intervening research.
This archive was generated by hypermail 2.1.6 : Mon May 24 2004 - 09:31:40 PDT