A Novel Load Balancer for Scalable Multi-NAT

Due to the upcoming IPv4 address exhaustion, the transition from IPv4 to IPv6 becomes an urgent problem restricting the growth of Internet. Multi-NAT, which is desired in large scale IPv4-IPv6 coexistent network, has inherent difficulties in the stateful traffic balancing and failure recovery. The existing schemes cannot handle them due to the absence of state synchronization. In this paper we propose a novel Load balancer (LB) to build a Scalable multi-NAT (SMNAT) in large scale network for various IPv4-IPv6 coexistent scenarios. The LB is specifically designed to have a translation pattern related hash keys and load-balance bi-directional traffic in two modes. Additionally an Adaptive reassigning algorithm (ARA) running in LB is presented to schedule flows adaptively to reduce the cost of state synchronization as well as guarantee the performance in load balancing. Comparing SMNAT with the existing load balancing schemes, the simulation result shows that our SMNAT outperforms other schemes and meets the goals of large scale NAT.