Research On Reliability Of The Data Center Network DCell

The arrival of the era of big data has a far-reaching impact on the development of human society.The generation of massive data promotes the transformation of data processing methods.In this development trend,cloud computing emerges as the times require.The mode of cloud computing is actually to provide online cloud services for network users by building a large-scale sharing platform,so as to achieve on-demand allocation of resources.Data center is the core facility of cloud computing,and data center network is the platform to connect large-scale servers of data center to realize online cloud services.Therefore,the performance of data center network determines the quality of service provided by cloud computing.With the large-scale increase of servers in data center network,server failures inevitably occur.Reliability is the premise to ensure the normal operation of data center network,and is an important factor affecting the quality of cloud computing services.Therefore,the re?search on reliability is very important to measure the performance of data center network.Fault-tolerant capability is an important metric for evaluating network reliability.The mea?sures of locating,diagnosing,detecting and repairing fault servers in data center network can improve the availability,reliability and robustness of the network.DCell is the representative of server-centric data center networks.It has good properties such as high scalability,large network capacity,small diameter and fault-tolerant routing,and can provide better services in network communication.This thesis studies the reliability of DCell based on its topological structure D_m,n.Firstly,this thesis evaluates the reliability of DCell from the aspect of fault tolerance capability.Fault tolerance measurement plays an extremely important role in the evaluation of network reliability.Connectivity is a landmark metric for evaluating the fault-tolerant performance of network.The greater the network connectivity,the stronger its fault-tolerant capability.The h-extra connectivity is the minimum number of nodes whose removal will disconnect a network,and every remaining component has at least h＋1 nodes.The h-extra connectivity overcomes the shortcomings of traditional connectivity and can better meet the needs of network reliability analysis.In this thesis,we obtain some fault-tolerant properties of D_m,n,and prove the h-extra connectivity of D_m,n。Secondly,this thesis studies the reliability on the fault diagnosis capability of DCell.When there are fault servers in the network and we want to ensure the normal operation of the network,it is necessary to quickly locate and detect the fault servers,and then repair or replace these fault servers,so as to improve the reliability of the network.Therefore,fault diagnosis is an important factor to ensure the reliability of the network.Fault diag－nosis strategies are usually classified into precise diagnosis and imprecise diagnosis.The g-good-neighbor conditional diagnosability and h-extra conditional diagnosability are two kinds of diagnostic metrics based on precise diagnostic strategy under different conditions.The conditions satisfy the actual situation of network failure,and can improve the fault di-agnosis capability of data center network.The imprecise diagnosis strategy can be divided into pessimistic diagnosis strategy and t/k-diagnosis strategy,both of which improve the diagnostic ability of the network by sacrificing a small number of misdiagnosis servers.In this thesis,we prove that,D_m,n,under the PMC and MM*models,the diagnosabilities based on the different strategies,the g-good-neighbor conditional diagnosabilities,and the h-extra conditional diagnosabilities,respectively.In this thesis,the D_m,n,has an h-extra connectivity of about h+1 times that of its traditional connectivity.At the same time,the diagnosabilities of different conditions and strategies of D_m,n under PMC and MM*models are obtained for the first time.The results of this thesis provide a quantitative analysis for the reliability and availability evaluation of DCell,and provide a theoretical basis for its application in the actual data center.