Distributed Computer System Design for Large Decentralized Organizations

This paper deals with the issue of designing distributed computer systems for large decentralized organizations. Specific problems addressed within this context include: determining the number of computer installations, choosing the sites that will receive these installations, deciding on the sizes of computers at different sites, configuring databases, allocating databases to computer installations and assigning users to computer installations. A class of decentralized organizations is identified for which decisions regarding the database configuration and the allocation of databases among processors can be effectively merged with decisions regarding the assignment of user nodes to processors. For these organizations, the design problem is formulated as an integer programming model. A brief outline of an effective solution procedure is provided and potential uses of the model as a design tool is demonstrated through a number of experiments. INTRODUCTION there has been a significant change in the nature and scope of information systems. During the last decade, computing and communication Decentralized organizations generally operate costs have declined dramatically. Communiover large geographical regions. Since decision cation networks having high levels of reliability making rights and the demand for decision aids and availability have become operational. These are distributed throughout the operating region, cost reductions and advances in the information information systems have to provide support at systems technology created new opportunities many points dispersed over a large geographic and changed the way information systems funcarea. Traditional centralized computer systems tion within organizations. Developments like permit collection and retrieval of data throughcomputer integrated manufacturing, storage and out the operating region but concentrate the distribution of goods, and point of sale systems processing and storage of data in a central site. reduced the cost of capturing and processing Such systems are becoming less and less vast amounts of data. Using this data, systems desirable for providing service to decentralized provide managers with timely and accurate inorganizations. This is mainly due to deterioratformation that was not previously available. In ing service levels resulting from increased loads many businesses, the mode of operation has on computer systems. These increased loads are changed, e.g., electronic funds transfer, partially caused by increases in size and comautomatic teller machines, long distance transplexity of organizations. More importantly, action processing in the banking sector, and air-