Abstract

We consider a database machine consisting of n nodes connected by an O(n*processing speed) bandwidth network. Each node consists of a processor, a random access memory, and a slower but much larger memory such as a disk. In order to approach optimal (O(n)) speedup on this hardware architecture, we partition relations roughly evenly among the processors. We study the problem of optimizing multi-join queries assuming such a data distribution strategy. For us, optimization consists of minimizing the number of times we need to redistribute relations in the course of a query. The optimization problem is NP-complete for general queries but linear time for a subclass of tree queries. 1. Setting and Problem The goal of the New York University Ultrabase project is to discover architectures and algorithms that speed up relational query processing roughly linearly with the number of processors. 1 Fortunately for our choice of project title, we choose a symmetric parallel architecture (figure 1...

Citations