Scheme for scalable ALM architecture based on topology-awareness

Various ALM (application layer multicast) mechanisms have been suggested to remove the barriers that block deployment of IP multicast. However, most ALM mechanisms do not consider the host group model, which is not a problematic portion, but advantageous feature of traditional IP multicast in terms of scalability, robustness, security, and transparency. Furthermore, those mechanisms build an overlay multicast network for data delivery mostly based on a simple measurement, with a few metrics, without directly considering the IP network topology. Thus, they might suffer some inefficiency, such as long join delay and large control overhead. We propose a scheme to take advantages of both IP multicast and ALM, based on two principal design factors. By adopting the host group model, our scheme can inherently acquire network topology information used to get rid of application level routing inefficiency, and to gain additional efficiency for many group members. Also, we exploit topology-awareness to reduce join delay and localize group members so that controls for setting up and maintaining overlay networks can be limited in a domain. Overall network performance is enhanced with our scheme by building an overlay network based on performance factors associated with network characteristics using a hierarchical structure.