Supporting Collective Queries in Database System for Data-intensive Workflow Applications

概要

論文の詳細を見る
Workflow system becomes the most important and necessary tool for data-intensive applications, enabling the composition and execution of complex analysis on distributed resources. Due to simple and effective query expression and high performance on query processing of parallel DBMS, we proposed a database system for workflows in which task in workflows can be expressed as a SQL query and processed by our systems. To be effectively used in the support of workflows, our database system supports collective queries in which multiple clients collectively issue a set of queries operated on a collection of data set as if a large data-scanning query is sent. Collective queries are integrated into a single SQL query which will be transformed into a logical plan composed of relational operator and sub-query and parallelized into multiple underlying resources. Finally result data is distributed into multiple clients in a coordinated way which can reduce a lot of data communication cost. Experiments show that our current database system is feasible and scalable to some extends.
2011-07-20