Acknowledgements
Many people helped in the work described in this thesis.
First, the research was dependent
on the Cascades code
of Goetz Graefe
and our discussions with him
via mail and meetings.
We were very fortunate
to have
other discussions
on top-down rule-based optimization efforts in
industry.
Pedro Celis told us about Tandem's optimizer,
William McKenna about EROC and NEATO
and Cesar Galindo-Legaria
about Microsoft's new transformation-based optimizer.
Locally,
Dave Clay,
M. Muralikrishna
and others
helped with discussions of
subquery unnesting,
bit vectors and
other physical modeling issues
and other areas of optimization
that were extremely useful in our work.
From the Oregon Graduate Institute,
we benefited from
discussions with
Leo Fegaras (now at University of Texas, Arlington)
on his query optimizer for an object-oriented database,
Bennet Vance (now at IBM Almaden Research Center)
on join order optimization,
Quan Wang on cost models
and Prof. David Maier
on all of the above as well as various optimization strategies.
At PSU,
Prof. Leonard Shapiro
defined the problem of this thesis clearly
and laid out a clear roadmap of
the work to be done,
broke up the Cascades search engine and model code,
then provided guidance with
Model D while at the same time making significant
changes to the search engine.
Finally he endured an
infinite non-convergent series of drafts of this thesis.
Professors
Jingke Li,
Earl Ecklund and
Sergio Antoy also helped with our work and
with this thesis.
Graduate students,
Susanne Stamp (alternative models)
Xinhong Yuan and Limin Chen (TPC-D, Oracle and SQL-Server)
Hsiao-min Wu (COVE)
Yubo Fan (cost models, materialized views)
made substantial contributions.
Finally
Beth Phelps and
Cynthia Beretta-Loep
in CS Department Office
provided essential support for our efforts.
This research was funded from grants from NSF and DARPA.
