Design, Implementation and Evaluation of Data Rivers for Efficient Intra-Query Parallelism

Upcoming applications such as OLAP, DSS and object-relational DBMS stress the demand for high performance and thus implicitly for efficient intra-query parallelism. In this paper, we evaluate the data river paradigm that has been designed for the management of intermediate query result sets that are produced as well as consumed by operators in a parallel database engine. We point out some aspects related to this paradigm that have a serious impact on query processing and efficiency. In addition we present an implementation based on a stringent modularization concept in combination with a set of parameters that on one hand provide necessary flexibility and on the other hand contribute to significant performance improvements. Furthermore, based on a thorough performance analysis we come up with a comprehensive set of parameter combinations that are recommended for specific situations covering the necessary spectrum of communication patterns typically found in parallel database engines.     «

Upcoming applications such as OLAP, DSS and object-relational DBMS stress the demand for high performance and thus implicitly for efficient intra-query parallelism. In this paper, we evaluate the data river paradigm that has been designed for the management of intermediate query result sets that are produced as well as consumed by operators in a parallel database engine. We point out some aspects related to this paradigm that have a serious impact on query processing and efficiency. In addition...     »