In the last several years we witnessed the proliferation of multimedia applications on the Ineternet. One of the unavoidable techniques to support thei type of communication is multicasting. However, even a decade after its initial proposal, multicast is still not widely deployed. One of the reasons is the lack of a solid business model. If the gain and the cost of multicast could be predicted, network operators might be encouraged to deploy multicast on a larger scale. In this paper we propose analytical expressions that could be used to estimate the gain of network-layer multicast. We show that the theoretical model matches extensive simulation and Internet measurement results remarkably well. Furthermore, we examine the reliability of traceroute data and of traceroutes-based conclusions. We investigate the node degree distributions in the Internet maps obtained from CAIDA and RIPE and we show the divergency of our results with those obtained by other researchers. We further focus on the analysis of multicast trees based on traceroute data. Only few results have been available on the node degree distribution of multicast routing trees which provided contradictory conclusions. Our results seem to indicate that the node degrees follow power laws only for a large number of multicast users@en

Application layer (AL) multicast protocols emerged, and in large number, as a response to the slow deployment op IP multicast. However, the gain of different AL multicast mechanisms, in terms of resources consumption over unicast, is questionable. In this paper, we investigate the efficiency of AL multicast protocol NICE, in terms of the number of hops a packet traverses on its path toward the destinations. We propose four alternative algorithms for the creation of NICE overlay and compare their efficiency to the efficiency of unicast, IP multicast, and another two previously evaluated AL multicast, and another two previously evaluated AL multicast protocols (MCAN and Scribe). Further, we investigate the influance of the underlying topology awareness on the efficiency. Our analysis has been performed via simulations, as well as via measurements on the PlanetLab network. We show that the NICE structure achieves better results than MCAN and Scribe.@en