From: Lucas Kreger-Stickles (lucasks@cs.washington.edu)
Date: Mon May 24 2004 - 11:51:02 PDT
In thier paper the authors present System R, a system which allows the
user to querry a databse without explicitly specifing access paths (as
would be the case when using a language such as sql). In particular
system R allows the user to express the desired data as a boolean
expression of predicates.
In their system users express querries and the system automatically
attempts to generate 'optimal' plans for executing that querry. At the
heart of this is the idea that the system can quess the cost of various
plans so as to select the best one. Furthermore, inorder to limit the
amount of potential querries to consider, certain stuctural constraints
were impossed on the potential querries.
The system used the RSS (Research Storage System) for datastorage and as
a means to estimate the cost of certain querry exectuions. In addiiton,
the system keeps around statistics relating to various aspects of the
database (such as the cardinality of certain relations and the number of
pages that hold relations of that type). All of this is used to help
the optimizer decide on the path to be executed.
Overall I found the paper relativly easy to read and understand (perhaps
given the reading from Query Evaluation Techniques for Large Databases
which covered or alluded to much of this.) If this idea was novel then
this paper must have been absolutly groundbreaking. I would imagine
that such optimization techniques (and a way to evalutae their
effectivness) is what took relational databses out of the realm of
intereting reserach topics and into the realm of practical solution.
-Lucas Kreger-Stickles
This archive was generated by hypermail 2.1.6 : Mon May 24 2004 - 11:51:07 PDT