Research On Architecture Design And Transmission Optimization For Wireless Data Center Network

In recent years,with the rapid development of cloud computing,data center network as its infrastructure has expanded rapidly.In the face of highly dynamic data traffic,traditional wired data center network adopts static topology,and its complex cable architecture has brought great challenges to the network scale expansion,energy management,operation and maintenance costs,etc.To address these challenges,the emerging high-frequency wireless-based dynamic networking technologies have made it possible to implement a high-performance wireless data center network with dynamic adaptive topology.With the great features of high bandwidth,dynamic connection and controllable flexibility,the high-frequency wireless technology can effectively improve network transmission performance and reduce the costs on network operation.At the same time,it also has several problems,such as easy-to-block,short-distance transmission,and the complex interference due to the intensive wireless deployment,which introduces new challenges for its practical deployment.In response to the opportunities and challenges brought by wireless data center network,this thesis will take the network architecture design as a starting point,and perform the research from the underlying topology to the upper-layer application transmission.The main contributions of this thesis include:(1)Proposed a new architecture of wireless data center network named Diamond.It mainly includes the key wireless technologies of multiple reflection,precise reflection and hotspot elimination,the routing strategy based on dynamic topology,and the bucketbased architecture expansion solution.The multiple-reflection and precise-reflection technologies take full advantage of three-dimensional transmission space,and solve the wireless challenges of easy-to-block transmission and dense wireless interference.For the first time,the large-scale and high-performance server-level dynamic wireless networking is supported in the data center network.The routing and architecture expansion strategies further address the short-distance transmission problem of the high-frequency wireless technology.Experiments based on 60 GHz wireless devices verify the feasibility and efficiency of the above wireless technologies.Simulation results show that Diamond can effectively improve network throughput and ensure efficient fault tolerance.(2)Designed a hybrid routing solution in a wireless data center network.In the wireless data center network architecture,this thesis first proposed a traffic scheduling framework WiRo based on the hybrid routing technology.By combining flexible wireless links and fixed wired links,the wireless interference models under different scenarios are analyzed and optimized.Next,a hybrid routing algorithm WiRS with theoretical performance guarantee is proposed,which solves the key problem of constructing the dynamic wireless topology and routing strategy based on real-time traffic demands.Simulation results based on real measurement data show that WiRS achieves efficient load balancing of the entire network,effectively reducing the flow completion time.(3)Designed the optimization solution for virtual machine transmissions in the wireless data center network.In the wireless data center network architecture,for the first time,this thesis presents a technical mechanism using the high-frequency wireless-based topology adaption technology to optimize the transmission cost of virtual machine migration and communication process for the upper-layer virtual machine applications.Based on the capacity constraint of virtual machine placement,wireless interference model and routing conditions,an algorithm based on relaxation and progressive-decomposition strategies with theoretical performance guarantee is designed to achieve the joint optimization of virtual machine migration,wireless topology and routing solution.