Sampling HTTP response packets for prediction of Web traffic volume statistics

Over a period of time the volume of network traffic generated by an Internet application is equal to the cumulative size of the packets that carried the traffic. We propose a method of estimating the expected volume of network traffic that will be generated by a population of World Wide Web users based on estimating the cumulative size of the expected hypertext transfer protocol (HTTP) packets. Assuming that the distribution of HTTP packet size remains constant we should be able to infer statistics of the expected cumulative packet size from a sample of HTTP packets observed traversing the network. This paper examines such an approach examining the estimation of the maximum expected network traffic that may result from present user browsing behaviour We take a sample of HTTP packets and use it to estimate the expected size of future samples and hence the expected future volume of network traffic. We find that confidence interval estimates with good coverage can be obtained. The best estimates are obtained after recognising that long range dependence exists in the size of HTTP response packets.