Research On Network Virtualization Technology For Data Distribution Service

In recent years,with the emergence of various new types of network services such as big data and cloud computing,service types have become more and more abundant.And many new multimedia services such as VoIP and video conferencing have been added to the network.The users' demand for QoS(Quality of Service)is increasing day by day.The traditional network has a single function and poor scalability.There are many kinds of equipment vendors deployed in traditional network infrastructure,and no uniform stanards can be applied to all these equipments.Therefore,it's difficult to implement unified configuration and management.Besides,the traditional network is too complex to fully bear the new technology today.In this context,network virtualization technology emerged.As one of the most important achievements of network virtualization,SDN(Software Defined Network)proposed the idea of separation of control and forwarding.SDN extracts the conplex control logic from traditional physical devices,centralizes the management and the control of the underlying physical devices,and provides a unified programmable interface and global management view for the upper application layer through rich and flexible APIs.The emergence of SDN provides new ideas for people to guarantee the QoS of services.The thesis starts with the actual needs and applies SDN to the laboratory virtual enterprise network project.The service type in the virtual enterprise network includes video conferencing,voice,file transfer,DDS(Data Distribution Service),and etc.With the perpose of improving the data transmission efficiency and network resource utilization,the thesis designs and implements a Service QoS-based Routiong Method(SQBRM)using the SDN's global view of network resources such as bandwidth and delay,which implements the path selection according to the service needs of network resources and the system resource usage.And to a certain extent,SQBRM ensures the QoS of services and achieves load balancing.Bisides,SQBRM also considers link failures and reserves backup paths,which improves the fault tolerance of the system to some extent.As an important communication model in the virtual enterprise network,DDS adopts the publish and subscribe mode to distribute data,but does not consider the distribution of topic data resources and the topology structure of the underlying network.Therefore,this is a problem that the same data is repeatedly sent and queued on a single link.The bandwidth utilization is reduced and the queueing delay of the data is also increased.Therefore,this thesis improves the existing decentralized discovery of DDS,and proposed an Improved Decentralized Discovery Mechanism(IDDM)to obtain the topic data distribution information in the network,which will lay the fundation for follow-up research.In the design and implementation of SQBRM,the SDN controller first obtains the necessary information,and then processes the acquired information to obtain the network topology information,the link bandwidth information and the link delay information respectively.Then according to certain path selection strategy,an optimal path is calculated and backup paths are reserved at the same time.Different from the traditional shortest path calculation method in the SDN architecture,SQBRM combines the service requirements for bandwidth and delay and the current network status,and finally calculates a path that meets the service requirements and also improves the network resource utilization.In the design and implementation of IDDM,the current discovery mechanism of the DDS architecture is first summarized,and then the improved architecture is proposed.The current decentralized discovery mechanism of DDS adopts the RTPS protocol,which firstly conducts the discovery of the participants,and then conducts the discovery of the participants' communication entities.When the remote and the local entities match successfully,the entity information is exchanged and recorded.Throughout the discovery process,DDS does not care about matching failed entities.Based on this,the thesis improves the underlying discovery module of DDS,adds the processing of the entities that are not matched,and obtains their topic information,so as to obtain the distribution information of topic data in the network.Finally,the thesis carries out simulation experiments on the various functions designed and implemented above by using Mininet simulation tool,and tests the correctness of each module.And through the analysis and summary of the results,it proves the advantages and practicability of each scheme proposed by the thesis.