Estimating the mean number of objects in M/H2/1 model for web service

In this paper, we estimate the mean number of objects in the M/H2/1 model for web service when the mean object size in the M/H2/1 model is equal to that of the M/G/1/PS and M/BP/1 models. To this end, we use the mean object size obtained by assuming that the mean latency of deterministic model is equal to that of M/H2/1, M/G/1/PS, and M/BP/1 models, respectively. Computational experiments show that if the shape parameter of the M/BP/1 model is 1.1 and the system load is greater than 0.35, the mean number of objects in the M/H2/1 model when mean object size of M/H2/1 model is the same as that of M/G/1/PS model is almost equal to the mean number of objects in the M/H2/1 model when the mean object size of M/H2/1 model is the same as that of M/BP/1 model. In addition, as the upper limit of the M/BP/1 model increases, the number of objects in the M/H2/1 model converges to one, which increases latency. These results mean that it is efficient to use smallsized objects in the web service environment.