Proceedings of the 2005 ACM SIGMOD international conference on Management of data
Foundations and Trends in Databases
| Hi-index | 0.00 |
In this thesis, we present novel query optimization and execution techniques for both new and traditional applications. We show that if we are operating over streaming sources, it is more beneficial to employ a rate-based cost model for query optimization. With respect to query execution over multi-join queries the introduction of operators being able to accept more than two inputs allows us to maximize the output rate of the query. Multi-input operators are straightforward extensions of traditional binary operators and are more resilient to fluctuations of the execution environment. The streaming environments naturally raise the question of when to use adaptive execution frameworks and when to stay with the “optimize then execute” paradigm. As a first step toward answering these questions, we revisit the query optimization problem over relational, disk-resident data and address the issue of the quality of the plans the relational optimizer chooses with respect to the errors in the cardinality estimates in the cost model's inputs. Towards that end we introduce the concept of error-aware query optimization.