Research On Dynamic Aggregated Multicast Considering Load Balancing

As the rapid development of INTERNET, there exists many emerging large scale multi-user applications, such as audio/video conference, network games, distributed interactive simulation (DIS), etc. All these applications involve multi-point group communications, that is, delivering data from one source to multiple receivers. IP multicast is introduced as an efficient scheme to support this sort of communication. Multicast technology is widely implemented and applied in INTERNET because it reduces redundant data in the network and saves network resource.Multicast also has disadvantages. When there are large numbers of concurrent multicast groups in the network, a lot of resources to keep forwarding states and management overhead to setup and maintain multicast trees will be required, which brings heavy burden for network routers. This problem brought by large-scale multicast is defined as multicast scalability problem.Aggregated multicast makes multiple multicast groups to share a single delivery tree, which saves node memory resource and at the same time decreases control overhead from setup and maintenance of multicast trees, since becomes a preferably solution to multicast scalability problem. However, as the number of groups aggregated to one single tree increases, more flows will be added to on-tree links, resulting in congestion, and bringing higher transmission delay, which makes it hard to meet QoS requirements. In addition, off-tree links in the network could not be used sufficiently and network load is unbalanced.According to this problem, the idea of applying load balancing in aggregated multicast is proposed and studied respectively on algorithm level and protocol level.We first proposed load balancing metrics for aggregated multicast, such as degree of load-balancing and congestion rate, and designed a dynamic Aggregated Multicast with Load Balancing Algorithm based on those metrics. AMLB algorithm constructs, aggregates, extends, shrinks and deletes multicast trees with groups join and leaving, and applies load balancing strategy throughout the processes of native tree generation and tree aggregation. Therefore it is more suitable for multi-user network applications in realistic network condition. Simulation results demonstrate that aggregated multicast based on load balancing can achieve more balanced network flow and better network performance while maintains aggregation efficiency.Then we proposed a dynamic Aggregated Multicast Protocol based on Modified AntNet. AMPMA based on modified AntNet model. This protocol reflects network link load via distributed positive feedback of artificial ants, and uses reasonable metrics in order to achieve dynamic load balancing in aggregation. AMPMA modeled and simulated by OPNET Modeler. Simulation results indicate that the protocol is effective in balancing network load and reducing network failure such as packet loss and congestion., performs preferably in load balancing.