An automated performance evaluation method for local area network server applications

Abstract

The distributed processing capabilities of local area networks has led to the development of "client-server" applications for microcomputers. Two separate computer programs, the client and the server, comprise the client-server system. The server program receives and services requests sent to it by the client program. More than one copy of the client program may be connected at a time to the server via the network with each instance of the client program operating on a separate computer. Two difficulties are encountered when testing such a system. First, since each client station is operated by one or more human-users, extensive evaluation is costly when several clients are operated for a substantial amount of time. A second problem is an inherent inaccessibility to information about internal system states. This thesis presents a method for automating performance evaluations of the server component of the client-server software system. Two changes to the original system are required. First, the human-user inputs to the system are generated using probability distributions. Second, an additional computer program, the test monitor, is added to the system in order to automate the accumulation and storage of test data.