An Approximate Performance Analysis and Measurement of the Equivalent Model of Parallel Queues for a Web Cluster with a Low Rejection

In this paper, we propose an equivalent model by using parallel queues to estimate a Web cluster's performance. Primarily, we use the parallel queueing model to represent the Web cluster and then simplify this model by an equivalent queue. Under such a specific procedure, we try to merge the parallel queues into a single equivalent queue, not only to help us to reduce the system's performance analysis complexity, but also to help us to use an approximate queueing model to analyze the system performance of the Web cluster. In addition, we use a queueing network analysis tool (QNAT) to do performance simulations for the parallel queues; we find the mean system response time of two, or multiple parallel queues by simulation and by analysis an equivalent model. Moreover, in order to verify the equivalent parallel queues for estimating a system's performance, we set up a real Web cluster system, and use some software tools for the performance measurement. Therefore, we verify that from the analysis, the simulation, and the measurements, the mean system response time of the parallel queues has similar results