Abstract: Intra-operator(or partitioned) parallelism is a well-established mechanism for achieving high performance in parallel database systems. However, the problem of how to exploit intraoperator parallelism in a multi-query environment is not well understood. This paper presents a detailed performance evaluation of several algorithms for managing intra-operator parallelism in a parallel database system. A dynamic scheme based on the concept of matching the rate of ow of tuples between operators is shown to perform well on a variety of workloads and con gurations. 1 This research was supported by the IBM Corporation through a Research Initiation Grant. 2 Author's current address: IBM Almaden Research Center, San Jose, CA, USA.
Manish Mehta 0002, David J. DeWitt