Font Size: a A A

Research On Energy-Aware Resource Scheduling Mechanism In Data Center Networks

Posted on:2018-01-24Degree:MasterType:Thesis
Country:ChinaCandidate:R Q WangFull Text:PDF
GTID:2348330518486515Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
In order to satisfy requirement of bandwidth and reliability of data center networks from new pattern network application of cloud computing and big data analysis,many topologies with rich connections have been proposed by academic community in recent years.These rich connections can improve network performance,but can also result in enormous energy consumption.On one hand so much operation cost and electric waste can hinder sustainable development of data center network,on the other hand it can destroy natural environment seriously.For problem of energy consumption optimization in data center networks,it has become an urgent problem for academic community and industry community.The paper analyzes typical topology structure of data center networks firstly;then,from the view of topology improvement,virtual machine scheduling and network flow scheduling,the paper studies energy consumption-aware resource scheduling mechanism.The mechanism can assure network performance and reliability,at the same time it can save energy consumption and improve efficiency.(1)For problem of invalid energy consumption from huge interlayer switches and links of typical topologies of data center networks,new type energy consumption-aware network topology called HyperDC has been proposed.This improved topology is based on HyperCube that is directed topology,which adds the account of switches in the single dimension and servers in the single switch.This topology keeps switches in same dimension in the full-mesh status with optical fiber as bus line.This strategy can ensure the nonblocking property,and it can improve network energy efficient.Use optical fiber as bus line,which can reduce energy consumption and extend network bandwidth.(2)For problems of lower server energy consumption efficiency and network latency from dispersive load in servers,energy consumption-aware virtual machine scheduling algorithm has been proposed.According to structure characters of HyperDC,this algorithm utilizes topology-aware virtual machine displacement mechanism with dynamic state that utilizes the method of reducing the dimension of topology to centralize virtual machines in minimal servers.Through this method,it should keep characters of the original topology and minimize account of servers to reduce the energy consumption of servers in data center networks.(3)For problems of lower link energy consumption efficiency and network congestion from centralized virtual machines displacement,energy consumption-aware network flow scheduling algorithm have proposed.According to structure character of HyperDC,this algorithm should utilize characters of orthogonality and full mesh of topology with network scheduling mechanism.It uses load rate of every switch in every dimension to schedule the routes in minimal network flow in the limited condition of certain network congestion,which can reduce network congestion and link energy consumption of data center network.
Keywords/Search Tags:Data Center Networks, Energy Consumption Optimization, Topology Improvement, Virtual Machine Scheduling, Network Flow Scheduling
PDF Full Text Request
Related items