Scalable Application Layer Multicast

Summary

This paper proposes a new scalable application-layer multicast protocol, designed specifically for low-bandwidth, data streaming applications.

They used 3 dimensions with which to evaluate application-layer multicast protocols:

• quality of the data delivery path
• robustness of the overlay
• control overhead

The NICE protocol arranges the end hosts into a hierarchy, which defines the multipath overlay data paths. The hierarchical structure is necessary for scalability and localizing the effect of member failures.

The hosts are assigned to different layers in the hierarchy and then split into clusters with other hosts close to them. The host in the geographic center of each cluster is designated as the cluster leader.

The authors performed several simulations of their protocol and Narada application-layer multicast protocol. Their results showed that they had lower link stress and improved end-to-end latencies and similar failure recovery properties, all with order of magnitude lower control traffic.

They also deployed their protocol in a wide-area testbed and found that nodes maintained low-latency paths and had a max packet loss rate of 1% or less. And, the average control overhead was less than 1 Kbps.

Criticisms & Questions

This was also another simple, but clever idea.  It would be interesting to see a direct comparison of this protocol to SRM.  I did that like that this paper did both simulations and an implementation and that they evaluated the results using multiple dimensions.