Distributed Control Plane:Adaptive Load Balancing for Parallel BGP Route Computing

The high scalability of next generation Internet supports deploying Internet services dynamically.With the deployment of more and more delay-and jitter-sensitive applications such as IPTV and VoIP,more route processing power is required to improve the performance of BGP route selection.Routers adopt distributed control planes and implement parallel BGP route processing,which is a potential approach to overcome the bottleneck of centralized control planes and improve the performance of route computation.However,current BGP parallel route computing schemes cannot keep load balance well,which degrades the performance of parallel route computing.In this paper,a load balance model for parallel BGP route computing is proposed based on hashing.Through accumulating prefix updates online and reallocating them among all processing nodes adaptively,the authors propose P-AP(Prediction-based Adaptive Partition)algorithm for parallel BGP route computation,design and implement the prototype of load balancing for parallel BGP route computation.Route Views BGP Update dataset is used to verify the performance of P-AP algorithm.Experimental results show that P-AP algorithm can balance load well among all processing nodes,minimize load adjusting,and have the maximum speedup of route computation.It can effectively improve the performance of parallel BGP route computing.