Joint Optimization Of Energy Saving And Load Balancing For Data Center Networks Based On Software Defined Networks

As a result of the Explosive development of Artificial Intelligence and Cloud Computing,the data center scale is rapidly expanding in recent years.With it comes,various data center operation problems are emerging,such as higher Data Center Networks(DCN)energy consumption and unbalanced network load.Nowadays,the data shows the energy consumption has became the main operation cost of data center.Many researches apply service migration or service hibernation to achieve DCNs energy optimization;The network load balancing is closely related to the Quality of Service(Qo S)of data center.Some links are under congestion and more link resources are idle in DCNs.How to improve the Qo S of data center with network load balancing is an area needs constant attention.Under traditional network architecture,the realization of network energy optimization and network load balancing require different middleware.There are a series of problems,such as manual configuration and devices loss.Generally,the network manager cascade the middleware or introduce the complicate middleware to joint optimization network energy consumption and load balance.As a new network architecture,Software Defined Networks(SDN)can merge these two aspects in one module which simplify the joint optimization process.The main work and innovation are as follows:(1)Due to the complicate middleware and difficulties in manual configuration,the unification of network energy optimization and load balancing is hard to achieve simultaneously.We propose a SDN based traffic flow strategy of DCN.This strategy can realize the network energy optimization and load balancing at the same time.We joint adopt multi-path routing and service hibernation to saving the network energy.And we schedule every subflow in the most available routing path to balancing network load.These two aspects can unification in one module with SDN'characteristics such as the control concentration and externally programmable.(2)We discover it has a high solution when the strategy find the optimal solution.We propose a heuristic algorithm to solve the problem.The heuristic algorithm reduces the time complexity by storing and preprocessing the inherent information of the network.We implement a experiment to verify the effectiveness of the heuristic algorithm through comparing with other typical methods.The experiment result shows our method can realize the network energy optimization and load balancing simultaneously.And it shows a better performance.(3)To study the performance of the traffic scheduling strategy in different application scenarios,the thesis combined the data characteristics in different application scenarios to modify the system model and heuristic algorithm.Then we design an experiments to research the strategy performance with the differences of network loads and data characteristics.We finally have found that the performance of the proposed scheme shows better than typical schemes.