Research Of The Optimal Path Allocation Algorithmoriented Load Balancing In Sdn

In recent years, with the increase of web applications, although Mobile internet, Cloud computing, Big data and such kind of services have emerged which have a profound influence on people’s daily life and have facilitated the users. But at the same time traditional network have been overwhelmed, it indicates that IT infrastructures construction is required the higher requirements of technology, which can provision what is needed; change based on demand, and be more flexible and robust. Network load balancing has been a hotspot in network research which can guarantee the performance of the network, improve the user experience. SDN(Software defined networking) adapt to the trend of network: the centralized control, the separation of data plan and controller plan, the programmability of the network, and the monitoring of network, whom played an important role in the network load balancing. This paper studies path selection algorithm, the network load balancing optimization has been divided into two parts, the link weight calculation and the path selection calculation. Main work and innovation are as follows:(1)The studies on network load balancing link weight calculation method are scant. The paper puts forward a new method for weight calculation----Based on the MC(Markov Chain transition probability calculation link weights) by using the centralized control of SDN. This method can effectively solve the problem of wide variance or the equal weight of Serial link, predictable link weights is a basis for the subsequent path calculation.(2)In order to get the best path base on load balancing, this paper proposes an Linear programming model by using the maximize the number of business flow request and ability of Network load balance as optimization objective, and Capacity and traffic of links as restriction, a method based on polynomial time approximation algorithm is put forward for its solution. The Simulation is conducted under the Fat-Tree topology of data center network. Extensive simulation results demonstrate that the proposed algorithm outperforms traditional SPF(Shortest Path First) approaches. In some cases, the proposed method can increase the network load balance rate of 30% and 20% Business flow request acceptance rate, which improve the network performance and achieve the effect of load balancing.(3)The paper propose a quantitative evaluation model because different operators have different sensitivity on different flow. Facing the different requests of flow, SDN controller can freely control the weights of those four factors: Average utilization of network, Utilization of optimal path, Network load fluctuations and Daly to get the optimal path allocation scheme, which can be best matched with different flow characteristics. Finally, using a case to specify how SDN controller chooses the best path in different path algorithm.