Improved multobjective algorithm for dynamic load balancing of network traffic

In a stub network which is connected to Internet via more than one edge routers to Internet Service Providers, the users are classified into groups where each group is connected to one edge router. Each edge link of the stub network has its own maximum available bandwidth provided by the ISP. If the group of users assigned to an edge link downloads a total inbound traffic which exceeds the maximum available bandwidth provided by that edge link, then the link is said to be over-utilized. On the contrary some group of users might download very less, under utilizing the capacity of the edge link assigned to them. In this paper, we try to provide a suitable solution for this problem by finding a better possible way of assignment of users to edge links so that the inbound traffic of all the edge links is fairly load balanced and no link is over or under utilized. This problem is framed as a multi-objective optimization problem with three objectives - load balance the inbound traffic, minimize user movements and minimize the intradomain cost. We improve on an earlier approach [1] using NSGA II [2] to minimize the first two objectives and implement third objective. Our results show that our approach load balances the inbound traffic as well as minimizes the user movements and also show that intra domain cost affects other two objectives.