Research Of Optical Switches Scheduling And Routing Algorithm In Cloud Data Center

Cloud service has several features like diversification,facilitation and large scale development.All these characters have brought new challengesfor the data center network performance optimization.Around the data center network performance optimization problems,we carried out the following research works:(1)We proposed two optical switches scheduling algorithms including SDF(Stringent Delay First)and m-SDF(m-order Stringent Delay First)which are based on Input Queue(IQ: Input Queue).In this thesis,the proposed algorithms considered the influence of speed-up to the system delay,and scheduledpackets with different delay constraint reasonably.By this way we can achieve the optimal goal of maximum delay-satisfying rate.The SDF algorithm can make each individual configuration matrix achieve scheduling optimization.The other algorithm(m-SDF)is better than SDF.The difference is that we let sevral configuration matrix joint together to optimize,so as to reduce the system clock wasting further.It can reduce the average data transmission delay,and increase the rate of delay guarantee.Based on a large number of experimentswe verified the effectiveness of the two algorithms.The experimental results show that m-SDF has better scheduling performance and can achieve higher delay satisfaction rate.(2)What we studied of the two kinds of data center network model are Multiple Independent Data Center Network which is called MI-DCN for short and Integration of Distributed Data Center Network(ID-DCN).It assumed that multiple independent data centers have been deployed in the network inthe former model.So we only considered routing problem.The latter assumed that there is only one data center in the network which is logical integrating and the physical resources in itaredistributed in the network reasonabley.With the limitations of the user data flow from each network node,we combined the network project,resource deployment and routing algorithm design to reduce the total network cost.Through the simulations,it compared the two kinds of design scheme under the network model and verified that ID-DCN model can achieve lower total network cost.(3)With the rapid increase of user data and the fixed data centers,some serversmay bear heavy load while others is idle if routing algorithm is still obeying the principle of getting serve from the nearest data center.To this end,we used Software Defined Network for reference with data center and studied balancing resources utilization of data centers.By designing transmited routing algorithm,the system achieved server load balancing.In this thesis,an Integer Linear programming(ILP:Integer Linear Program)and heuristic algorithm are proposed to design corresponding routing scheme.At last,we verified the effectiveness of the proposed routing algorithm through simulations.The experimental results show that the proposed algorithms can achieve higher load balancing than others.