Research On Congestion Control Mechanism For Named Data Networking

With the vigorous development of Internet scale and applications,the number of users and the amount of information in the network have increased dramatically.In many aspects,the traditional network designed for communication between two hosts has hardly met the needs of users in the big data era.A content-centric network architecture,Named Data Networking emerged as the times require,which gradually developed into research hotspots for next-generation network architectures and become one of the most classic content center networks.Compared with the traditional TCP/IP network,NDN has many advantages and differences,including the introduction of a wide range of intranet caching mechanisms,traffic self-balancing capability,multiple content source property and so on.Therefore,the existing congestion control methods design for the traditional network cannot be directly applied to NDN,also making contribution to the congestion control scheme that can be applied to NDN according to its natural features is of great significance for designing excellent transport control protocols for NDN.This paper focuses on the congestion control problem in NDN,and explores it deeply from the receiver delay measurement under the multi-source transmission characteristics and the flow control capability of the NDN router.The main innovations in this paper are:First,this paper uses the Raspberry Pi Board as the hardware platform of network node,and uses NFD forwarder,ndn-cxx prototype library to build the NDN running environment.Base on this framework,developed the receiver side program and the corresponding data source program.Furthermore,under a variety of common topological structures,from the bottleneck link and multisource transmission which may lead to network congestion,the specific state of network congestion and its impact are analyzed.Secondly,this paper starts with the problem of instability of RTTs measured within receiver side in multi-homed content transmission,then further discussed the flow control capability of NDN routers,and proposed a hybrid congestion control mechanism based on probabilistic forwarding(called as HCCPF for short).HCCPF maintain windows of interest packet allowing to send at the receiver side,and deploys a RTT measurement method which can change the base RTTin real time by using the jitter characteristics of RTT caused by the switching of content source in data retrieval stage.Hence,the RTT information is correctly measured,after that,the congestion status is used to shaping the Interest rate accordingly,while an improving to the Interest retransmission is further considered.At the router side,the flow is divided into fit grained levels based on the content name,so the congestion contribution coefficient of each flow in each interface can be analyzed.Then the coefficient per flow in each face is used to obtain the congestion status of each interface.Furthermore,the congestion status and the forwarding probability of outgoing Interests in each face are related linearly.On the other side,by controlling the size of the sending queue per flow,the proportion of the resources occupied by the flow to the link is adjusted,so the effective utilization of the bandwidth of the node is realized.Finally,the paper implements HCCPF mechanism in the NDN running environment built before.Experimental results show that the proposed scheme has better performance in congestion avoidance and congestion control than the existing solutions and has strong robustness.