ALLOCATING DATA AND WORKLOAD AMONG MULTIPLE SERVERS IN A LOCAL-AREA NETWORK

Abstract

Local area networks (LANs) have become indispensable in today's highly competitive business environment. A LAN connects multiple file servers for many concurrent information retrievals. The system administrator has to resolve two allocation problems, workload and data. The allocation decisions are interrelated; a transaction is routed to the server that contains the data items requested. Affinity-based workload allocation is employed, and the database is thus fragmented into data files that are allocated across multiple servers. The paper's two primary objectives are (i) to provide a rationale for workload and data allocation and (ii) to present an analytical model which will attain effective allocation policies. The integrated problem is presented in the form of a nonlinear zero-one integer program. A special structure of the problem is employed to demonstrate the complexity of the problem, and to propose an algorithm.