Simulation of distributed object oriented servers

Abstract

Distributed object oriented (OO) computing such as RMI, COBRA and SOAP etc is fast becoming the defacto standard for software development. The aim of the system designer is to determine the optimal deployment strategy for the system to perform efficiently. This is an enormous task especially when multiple object servers are fielded on hardware of different specifications. The number of possible deployment strategy of object servers to hardware grows exponentially with increase number of object server and machine. For example, with 3 machines and 10 object servers there are 59049 possible deployment patterns. Eventually, the number of possible deployment makes it impossible for system designer to setup test bed to determine the optimal deployment strategy. The main goal of the simulation model is to analyze the object server deployment, verify an existing optimization model and to determine the optimal deployment strategy that will reduce the client response time. In one of the experiment conducted with the simulation model, in an environment with 3 machine and 10 object servers, it will take 53 years to attempt all deployment patterns in the lab environment. The simulation model will take only 13 days, which is an improvement of 1480%.