Measuring and modelling the group membership in the Internet

In this paper, we measure and model the distribution of multicast group members. Multicast research has traditionally been plagued by a lack of real data and an absence of a systematic simulation methodology. Although temporal group properties have received some attention, the location of group members has not been measured and modelled. However, the placement of members can have significant impact on the design and evaluation of multicast schemes and protocols as shown in previous studies. In our work, we identify properties of members that reflect their spatial clustering and the correlation among them (such as participation probability, and pairwise correlation). Then, we obtain values for these properties by monitoring the membership of network games and large audio-video broadcasts from IETF and NASA. Finally, we provide a comprehensive model that can generate realistic groups. We evaluate our model against the measured data with excellent results. A realistic group membership model can help us improve the effectiveness of simulations and guide the design of group-communication protocols.