| | |
Summary: Automatic Relationship Discovery in Self-Managing Database Systems
Ihab Ilyas1
Volker Markl2
Peter J. Haas2
Paul G. Brown2
Ashraf Aboulnaga2
1
Purdue University, West Lafayette, Indiana
2
IBM Almaden Research Center, San Jose, California
ilyas@cs.purdue.edu, {marklv,phaas,pbrown1,aashraf}@us.ibm.com
1. Introduction
Relational database management systems rely on the query-
optimizer component to choose a minimum-cost plan for access-
ing the data. The optimizer's cost estimates rest on "selectivity"
estimates, that is, estimates of the amount of data that will be ac-
cessed at each stage of processing. Such selectivity estimates are
in turn based on a set of summary statistics about the data (i.e.,
about the relational tables and columns) that are maintained in the
system catalog. In traditional systems, the user specifies the set of
|
