Design and analysis of a multi-backend database system for performance improvement, functionality expansion and capacity growth

The aim of the proposed research is to investigate whether for the management of very large databases the use of multiple mini-computer systems in a parallel configuration is feasible and desirable. By feasible we mean that it is possible to configure a number of (slave) minicomputers each of which is driven by identical database management software and controlled by a (master) minicomputer for concurrent operations on the database spread over the disk storage local to the slave computers. This approach to large databases may be desirable because only off-the-shelf equipment of the same kind is utilized to achieve high performance without requiring specially-built hardware and because identical database management software is replicated on the slave computers. The approach makes the capacity growth and performance improvement easy because duplicate hardware can be added and used with replicable software. In this report, we present a new approach to the solution of database management problems involving database growth and performance enhancement. A system which uses a multiplicity of conventional minicomputers, novel hardware configurations and innovative software design is presented. This extensible system tries to achieve the ideal goal of having the performance (both response time and throughput) be proportional to the multiplicity of minicomputers. Our first effort is to identify the major problems and bottlenecks involved in developing such an ideal system. Two major problems, one called the controller limitation problem and the other the channel limitation problem are identified.