Mixture Models of Endhost Network Traffic
In recent years there has been much interest in modeling internet traffic that comes from inside large networks, such as at routers or gateways. In this paper, the authors focus on modeling a little studied type of traffic - namely the network traffic generated from end-hosts. They study traffic data collected from hundreds of enterprise laptop users. They introduce a parsimonious parametric model of the marginal distribution for connection arrivals. They employ mixture models based on a convex combination of component distributions with both heavy and light-tails.