Case Studies in Performance Modelling and Simulation of Client/Server Systems

The performance of client/server systems is complex and highly dependent on a number of interacting issues such as workloads, processor interconnection, mapping issues, processor type and communications medium making the realisation of a fast, efficient implementation difficult and time consuming. This can be further complicated if computer systems are not exclusively dedicated to one processing activity. A key obstacle to the delivery of client/server systems is the inability to build a scalable system which needs the performance criteria set before development and which often change during development and commissioning of the system. Often, it would be useful if the performance of the system could be predicted in different scenarios such as increasing the number of clients, increasing the number of servers, changing the bandwidth of the network, varying the cache size, etc. In this paper, a simulation-based toolset to support modelling and performance predictions is presented. This development environment is composed of a graphical modelling tool, a simulation facility, and a design tool. Simulation allows client/server program performance to be predicted on target systems and design alternatives to be compared. The target system models a cluster of workstations interconnected by a local area and / or wide area network. The simulation model architecture is modular and extensible which allows re-configuration of the client/server system. The paper will emphasise case studies based on client/server systems developed for industry.