Research And Design On Fault Tolerance Mechanisms In Data Center Network

The data center network is the key infrastructure for almost all Internet services.With the rapid increase in network traffic,the scale of data centers continues to expand,and network failures occur frequently.It takes a long time for the data center network to recover from the failure and find a new forwarding path,which seriously damages the performance of interactive real-time services and greatly reduces the overall availability of the data center.In this context,network faults have an extremely high impact since most routing algorithms have not been design to tolerate faults.Because of this,as few as one single link failure may stall message in the network,leading to deadlock configurations or even worse,prevent the finalization of applications running on the computing systems.Therefore,the fault-tolerant method of data center network and the fast recovery fault-tolerant method have been the research focus of academic and commercial circles in recent years.There are two difficulties in the traditional fault-tolerant method: it is more difficult to update the forwarding table during fault-tolerant rerouting and the fault recovery is generally slow.The dynamic update of the forwarding table in the data center network is difficult to achieve in the current IP network routing mechanism.There are two main reasons for slow recovery.Firstly,data center network with a multi-root tree topology,there is a lack of an immediate backup path that connects downward.Secondly,the fault-tolerant rerouting algorithm in the data center network is not efficient enough.In view of the above problems,from the perspective of simple fault recovery,this thesis implements the data center network fault tolerance algorithm on the routing architecture based on port addressing.The main work of this thesis includes:Firstly,a port-based source routing addressing(PSRA)scheme,which makes the table lookup operation unnecessary and decreases the switch complexity.This addressing scheme is coded by numbering the output ports of each network node.Secondly,routing mechanism plays an important role in utilizing path diversity,dealing with faults and balancing network traffic among multiple paths.Leveraging the characteristics of PSRA and the regularity of DCN topologies,an extremely simple routing mechanism is design,without switch involvement,control message interaction and topology information storage.The routing mechanism defines how to obtain the path(s)from a source to a destination applying some routing algorithms,how to express the path(s),and how to use the path(s)in a DCN.This algorithm combines the broadcast fault-tolerance idea with port-based source routing addressing(PSRA)architecture,and solves the problem of the difficulty of dynamically updating the forwarding table in the broadcast fault-tolerance idea.Lastly,a high-efficiency fault-tolerance mechanism is proposed for the addressing and routing architecture.A fast fault-tolerant method based on F10 topology is proposed.This method implements a local fast rerouting algorithm by improving the F10 tree network topology.In addition,in the routing addressing part,a routing addressing scheme based on port addressing is used to better realize the forwarding of data packets.In summary,this thesis has conducted research and experiments on the fault recovery and fault tolerance of the data center network,implemented the fault tolerance algorithm based on port addressing,proposed a fast recovery fault-tolerance method based on F10 topology,and verified the experiment through experiments.