Scalable Application Layer Multicast

This paper presents an overlay-style protocol for scalable, application-layer multicast delivery. The primary focus here is on developing a hierarchical node organizational structure that minimizes control overhead. Even though the paper focuses on low-bandwidth applications, the authors claim that applications with different requirements can be handled by varying the overlay paths and metrics. The protocol employs a tree structure to avoid loops and data duplication, with a careful explanation of joining mechanisms for nodes. However, this tree model has several clusters at each layer and that may present problems - nodes at the top are overloaded compared to those at the bottom, which effectively lose bandwidth in one direction.

The cluster-based model may get complicated in a real network, especially merging and growing the tree. To an extent, this paper presents general approaches towards application-layer multicast rather than a specific protocol that can be readily implemented. It's more interesting than the previous one but once again, it may not really belong in the syllabus. The previous paper provides a better basic exposure to the requirements for multicast delivery.