H. Pirahesh, Joseph M. Hellerstein, and Waqar Hasan. Extensible/rule based query optimization in starburst. In Proc. of the ACM SIGMOD, San Diego, May 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....fB 1 ; Bng and the set of view definitions fQ 1 ; Qmg will be denoted by B and Q respectively. 5. 1 Traditional Approach Given a query as in Figure 3, most of today s optimizers represent this query as multiple query blocks with data flow among the blocks (e.g. QGM representation in [PHH92] Optimizing transformations are applied on this query representation. Examples of such transformations are that of propagating predicates from one query block to another [LMS94, PHH92] Note that if the views did not have any aggregates, then the query is reduced to a single block query. As ....

....represent this query as multiple query blocks with data flow among the blocks (e.g. QGM representation in [PHH92] Optimizing transformations are applied on this query representation. Examples of such transformations are that of propagating predicates from one query block to another [LMS94, PHH92] Note that if the views did not have any aggregates, then the query is reduced to a single block query. As stated earlier, when the query contains a join among aggregate views, the query cannot be reduced to a single block query by traditional preprocessing transformations. The traditional ....

H. Pirahesh, Joseph M. Hellerstein, and Waqar Hasan. Extensible/rule based query optimization in starburst. In Proc. of the ACM SIGMOD, San Diego, May 1992.

.... a per tuple cost of evaluation and a selectivity with every user defined predicate (as in [HS93] While the task of optimizing queries with user defined predicates is important, there are other interesting directions of research in user defined predicates, e.g. use of semantic knowledge, e.g. PHH92, CS93] As pointed out earlier, the traditional heuristic of evaluating predicates as early as possible is inappropriate in the context of queries with user defined predicates. There are two known approaches to optimizing queries that treat user defined predicates in a special way. The first ....

H. Pirahesh, Joseph M. Hellerstein, and Waqar Hasan. Extensible/rule based query optimization in starburst. In Proceedings of the 1992 ACM-SIGMOD Conference on the Management of Data, pages 39--48, San Diego, CA, May 1992.

....possibly efficiency concerns. For example, as discussed in Section 5, we sometimes find it convenient to use the notion of affiliated tables (instead of rewrite rules) in order to express the knowledge that a set of conditions need to be pushed down . In contrast to our approach, Starburst [PHH92] uses a procedural language to express the semantic knowledge. While making specification of semantic knowledge harder, such an approach enables rewrite rules to express any desirable transformation and leaves the design of search algorithms open ended. We should note that the ability to invoke ....

....functions is a limited form of extensibility. In contrast, the Starburst architecture has far more ambitious goal of providing extensibility. 9 Related Work Many extensible systems have been proposed [CH90] with varying degrees of support for extensibility in the optimizer [BG92, GD87, Loh88, PHH92, SJGP90] The query rewrite optimization [PHH92] in Starburst is most directly related to our approach. As discussed in the preceding section, the query rewrite optimization in Starburst relies on a procedural language. The rule programmer is responsible for ensuring termination and search ....

[Article contains additional citation context not shown here]

H. Pirahesh, Joseph M. Hellerstein, and Waqar Hasan. Extensible/rule based query optimization in starburst. In Proceedings of the 1992 ACM-SIGMOD Conference on the Management of Data, pages 39--48, San Diego, CA, May 1992.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC