Research On Energy-Efficient Architecture Optimization And Routing Technology For Wired-wireless Hybrid Data Center Networks

With the rapid development of big data,cloud computing,artificial intelligence and other technologies,the amount and scale of data centers are constantly expanding,which has also made data centers one of the world’s largest energy consumers.Therefore,how to improve the energy efficiency of data centers has become one of the hot topics in current data center researches.Data center network(DCN)energy consumption accounts for 10-20% of the total data center energy consumption.Therefore,improving the network energy efficiency will remarkably increase the data center energy efficiency.Data centers usually redundantly deploy a large number of network equipment to meet the peak demand of network traffic.However,the actual traffic load on the network is usually much lower than the peak load,resulting in much waste of network resource and energy.The high-speed wireless transmission technology has the advantages of low power consumption and reconfigurability,so it has been widely adopted in data center network performance optimization researches.However,how to make full use of the reconfigurable wireless resources to improve network energy efficiency remains to be an important problem for wired-wireless hybrid data center networks.Therefore,the following researches have been carried out in this thesis.This thesis summarizes the existing data center network architectures,and analyzes the relationship among data center networks’ energy efficiency,architecture and traffic,according to data center network energy efficiency metric.On this basis,the existing energy-efficient routing technologies for data center networks are discussed,and their advantages and disadvantages are analyzed.Aimed at hybrid data center network architecture,an energy-efficient hybrid Fat-Tree DCN architecture optimization method based on wireless utility optimization is proposed,to improve the topological energy efficiency of hybrid network architecture.Firstly,according to the limited coverage of wireless links,an optimization model for the energy-efficient FatTree hybrid network architecture is constructed,which aims at optimizing the uniformity of wireless coverage and maximizing the wireless utility between the switches at the top of racks.Then,focused on the energy-efficient Fat-Tree hybrid network architecture optimization problem,which is a non-linear integer programming problem,we design a greedy random adaptive solution based on the genetic algorithm.Finally,simulation results show that this method can effectively reduce the average path length as well as the number of necessary switches in the hybrid networks,thereby increasing the energy efficiency upper bound of the hybrid data center network.According to the relationship between routing and energy efficiency in hybrid data center networks,an energy-efficient routing method based on wireless offloading is proposed to improve the energy efficiency of the hybrid network.Firstly,with the traffic demand guarantee as the constraints and the minimized overall network power consumption as the optimization goal,a routing optimization problem for hybrid networks is constructed,which is proved to be NP-Hard.Then,based on the characteristics of wired-wireless hybrid topology,a twostage solving algorithm based on topology weight transformation and relaxation-rounding is designed.This algorithm innovatively relaxes the non-linear binary constraints into linear topological constraints through the topological weight transformation,thereby reducing the solving complexity.Finally,two routing algorithm performance testing platforms separately for the flow-level simulation based on Python and for the packet-level simulation based on OMNet++ are built.Simulation results show that,compared with existing algorithms,the proposed algorithm can reduce the network equipment power consumption while ensuring the network traffic demand,and therefore the network energy efficiency can be improved.