Practical throughput estimation for parallel databases

Abstract

Methods for estimating the performance of DBMSs can aid the design of database systems by identifying potential performance bottlenecks or by predicting the relative performance of di erent designs. Performance estimation is critical in parallel database systems with distributed memory where an e ective overall performance depends on a good choice among a wide range of ways of placing data. This paper describes an approach to performance estimation for shared-nothing parallel database systems. It estimates system throughput for a given benchmark or set of queries, and can exercise di erent data placement schemes to determine the data layout which provides the best throughput value.