SARA: A session aware infrastructure for high performance next generation cluster-based servers

Internet server clustering has been widely used by operators to improve the scalability and the availability of the rendered services under heavy load condition. Load balancing is a well known mean to optimize the usage of the cluster available resources by fairly allocating them to the offered network traffic. Most state-of-the-art research advocate either session oblivious or coarse grained session aware load balancing architectures. While guaranteeing session integrity, the coarse grained session awareness leads to an unfair allocation of the cluster resources. In this work, we advocate an innovative fine grained session aware load balancing architecture of an offered IP traffic to a cluster of Internet servers. The proposed architecture aims to maximize the cluster useful throughput in terms of completed sessions per unit of time. It provides high availability capabilities in case of failure of the legitimate entry point to the cluster. Finally, it is fully client transparent and is open to adapt to any multiple-flow based NGN service such as voice over IP or video streaming.