13 July - 19 July

Tuesday, July 19, 10:15

The Impact of Heterogeneous Traffic on the Performance of Proxy Cache Servers

Tamas Berczes, Faculty of Informatics, University of Debrecen, Hungary

The focus of this talk is to examine the performance behavior of Proxy Cache servers when we use heterogeneous traffic. We describe the multi-class queuing network model of the Proxy Cache server, where we separate the requests in two classes by virtue of their size. If the size of the requested document is greater than average we put it into class a. In the opposite case, when the size of the requested file is small we put it into class b. We have calculated the overall response time with and without a Proxy Cache server. We analyzed how various factors affect the performance of a Proxy Cache server when we use heterogeneous traffic. In general when the arrival rate of requests increases, then the response time increases as well regardless of the existence of a Proxy Cache server. Increasing the percentage of the class a the response time increases too. When we use a higher percentage of the class a and we use a high arrival rate, then the response time gap is more significant between the cases with and without a Proxy Cache server. Using a low percentage of class a files, a low arrival rate and low cache hit rate probability we get higher response time in presence of a Proxy Cache server. Several numerical examples illustrate the effect of arrival, external arrival rate and the file size of class a and class b on the mean response times.