Distributed query processing and optimization techniques for a hierarchically structured computer network

Abstract

This paper describes several distributed query processing and optimization techniques useful for achieving efficiency in a hierarchically structured hardware and software architecture. The query optimization techniques take into consideration data flow and pipelining query processing strategies and data distribution in a hierarchically structure network. Four optimization techniques are introduced: (1) reducing query versions by selecting only those whose subtrees process data stored in the same subnetworks, (2) eliminating redundant execution of common subtasks by dynamically merging concurrent queries, (3) balancing the gain in parallel processing with the cost of data and message communications among processors using the notion of “maximal limb”, and (4) adjusting processor assignment strategies to take account of dynamic merging of concurrent queries. A bottom up algorithm for response time and execution time calculation and some cost formulas for query tree evaluation are also presented.