OPTIMAL ALLOCATION OF CLIENTS TO FILE SERVERS

The transparent way in which the remote resources are shared with impr oved performance justify the rapid replacement of single Large Compute r Systems by Local Area Networks (LANs). Information (data) sharing is one of the major goals of LANs and file servers are one of the means of supporting data sharing in LAN environments. A typical LAN might ha ve one or more file servers and in a network consisting of multiple fi le servers making effective use of them is an important design issue. One of the means of effectively utilizing the file servers is to restr ict the number of clients using the server (Client Server Allocation P roblem). The existing techniques for solving the client server allocat ion problem either treat the clients on the network as statistically i dentical or treat them as entirely independent. However, the evaluatio n studies conducted in the context of computer networks show that clie nts on a network are neither identical nor entirely independent, but c an be grouped into different classes based on their workload generatio n pattern. Thus, an algorithm which solves the client server allocatio n problem taking care of the workload generation pattern is required. In this paper, we propose a new iterative algorithm for solving such p roblems based on the concept of demand inflation. The time and space c omplexities of the proposed algorithm are compared with that of the ex isting algorithms. The application of the proposed algorithm is illust rated with an example. (C) 1997 Elsevier Science B.V.