Cache Partitioning for Multiple Sessions in Local Loss Recovery of Reliable Multicast

Local loss recovery for reliable multicast can provide significant performance improvement in terms of loss recovery latency, bandwidth consumption and network throughput. Active Reliable Multicast (ARM) is a novel loss recovery scheme for largescale reliable multicast. In ARM, local loss recovery is realized by retrieving repairs at near active routers. In practice, active resources are limited and active resources allocation strategies are required for optimizing performance in terms of the loss recovery and/or network bandwidth consumption. An active router may need to support the local loss recovery for multiple multicast sessions simultaneously. How to efficiently partitioning the cache resource to these sessions is crucial to the performance of local loss recovery. In this paper, we propose a cache partitioning method called "Layered Greedy Cache Partitioning" to deal with this dilemma. For this method, the active routers are first grouped into different layers according to the tree topologies and the design begins from the lowest layer. The cache partitioning of an upper layer can be performed separately by making use of relevant design information of the lower layer. An optimization for cache partitioning at each layer can then be conducted and a "local" optimal solution can be achieved. On global sense, however, the solution is only suboptimal. The performance of using the proposed cache partitioning method is compared with that using uniform cache partitioning and proportional partitioning. A significant performance improvement is found.