Empirical Model of WWW Document Arrivals at Access Link

The cable and telephone industries have already begun constructing the information super highway. The network capacity planning for data services, however, cannot start until a realistic traffic model is developed based on actual traffic data. Since the future data services are perceived to resemble today's World Wide Web (WWW) browsers, we propose in this paper a traffic model for the access network based on today's WWW traffic. The model is an ON-OFF two-state model with the ON period consisting of a sequence of document transmission requests from an individual subscriber. Actual traffic data with over 20,000 data points was used to fit distributions to the model. The ON and OFF periods are found to be of heavy-tailed Weibull and Pareto distributions, respectively. The inter-arrival times of requests within the ON periods can be described with another Weibull distribution. This empirical model can be used as a realistic basis for network capacity planning for the future access networks. Designation of Technical Areas: • broadband access and delivery systems • transmission and access systems • traffic characterization • modeling and performance analysis If accepted, the paper will be presented by the author. Empirical Model of WWW Document Arrivals at Access Link Shuang Deng GTE Laboratories, Inc. 40 Sylvan Road Waltham, MA 02254, USA ([email protected], +1 617 466 2165) Abstract The cable and telephone industries have already begun constructing the information super highway. The network capacity planning for data services, however, cannot start until a realistic traffic model is developed based on actual traffic data. Since the future data services are perceived to resemble today's World Wide Web (WWW) browsers, we propose in this paper a traffic model for the access network based on today's WWW traffic. The model is an ON-OFF two-state model with the ON period consisting of a sequence of document transmission requests from an individual subscriber. Actual traffic data with over 20,000 data points was used to fit distributions to the model. The ON and OFF periods are found to be of heavy-tailed Weibull and Pareto distributions, respectively. The inter-arrival times of requests within the ON periods can be described with another Weibull distribution. This empirical model can be used as a realistic basis for network capacity planning for the future access networks.The cable and telephone industries have already begun constructing the information super highway. The network capacity planning for data services, however, cannot start until a realistic traffic model is developed based on actual traffic data. Since the future data services are perceived to resemble today's World Wide Web (WWW) browsers, we propose in this paper a traffic model for the access network based on today's WWW traffic. The model is an ON-OFF two-state model with the ON period consisting of a sequence of document transmission requests from an individual subscriber. Actual traffic data with over 20,000 data points was used to fit distributions to the model. The ON and OFF periods are found to be of heavy-tailed Weibull and Pareto distributions, respectively. The inter-arrival times of requests within the ON periods can be described with another Weibull distribution. This empirical model can be used as a realistic basis for network capacity planning for the future access networks.