A Software Defined Network QoS Research Based On Network Calculus

With the development of Internet of Things,Social Networks and Smart Cities,network traffic is still increasing rapidly.Traffic engineering could realize precise control of traffic flow to provide on-demand Quality of Service(QoS)for different flows in complicated network environment.In Software Defined Networks(SDN)paradigm,control function is decoupled from switches and developed into an independent plane,which is much more convenient to retrieve a global view of networks.This provides a new opportunity for research of traffic engineering and End-to-End QoS.Network Calculus provides a theoretical framework for bound performance analysis based on min-plus algebra.Taking advantage of network global view and Network Calculus theory,this work proposes a realization of traffic engineering and corresponding theory system.From bottom up,this work could be divided into four sections.The foundation section is network calculus models in SDN,which analyze resource sharing in different queue disciplines.Based on global view of network traffic pattern,a traffic monitoring method is proposed to retrieve traffic pattern with little bandwidth overhead,an estimation algorithm is proposed to create whole traffic matrix.Furthermore a network calculus service matrix is proposed based on traffic matrix.Network calculus model and traffic matrix are utilized for routing optimization.The last section is a research of wireless sensor networks(WSNs)routing algorithm and its realization in SDN paradigm.(1)Network calculus analysis model in SDN paradigm.Researchers did lots of work on traffic,conflict and performance bound analysis in OSI network model based on network calculus.As SDN network paradigm is very different from classic OSI model,analysis of traffic flow,conflicts and performance bound is consequently very different too.This section analyzes protocol parsing and flow table searching in Openflow switches based on network calculus,a more specific discussion of flow table hit or loss is provided.In Openflow switches traffic aggregates in output port according to queue disciplines.This section discusses difference of queue theory and network calculus theory and followed by analysis models of FIFO,SFQ and HTB queue disciplines with network calculus theory.(2)Research of traffic matrix based on network calculusNetwork traffic pattern could be utilized for network planning,traffic engineering and routing optimization.This section provides a reactive traffic monitoring method in SDN network based on Openflow events information.In this proposal SDN controller monitors core links in SDN data plane by sending traffic query commands,and estimates aggregation links with an estimation algorithm.A complete traffic matrix is created from monitoring data and estimation data.Traffic on America Abilene network core links showed clear cyclicity,this section developed a traffic recording algorithm with reactive monitoring method.This recorded traffic matrix is able to locate traffic abnormality.This section develops network calculus service matrix theory to apply traffic matrix to network QoS and routing optimization.A service matrix is calculated from traffic matrix and a delay matrix and backlog matrix is provided based on service matrix and network calculus theory.(3)Research of SDN QoS routingThis section builds a SDN simulation environment on Mininet,with OVS as forwarding switch and Ryu as controller.Openflow protocol is installed as southbound protocol between OVS and Ryu,and OVSDB&ssh protocol to configure OVS switch and host client,RESTful web API to check and configure Ryu controller.With this environment,this section researched SDN network routing algorithm,QoS routing based on network calculus model,delay and backlog guaranteed QoS routing based on network calculus service matrix.(4)Research of wireless sensor network routing protocol in SDNThis section researches WSNs routing protocols for static and dynamic environments.AODV routing protocol starts a flooding process for route discovery,creates routes based on reply packet of flooding.AODV issues route discovery process each time before route creation,which is suitable in dynamic scenarios.An enhanced AODV routing is proposed,named AODV++.AODV++ evaluates sensor energy residual and traffic loads of each node and returns this information to source node.A route with a balanced residual energy and traffic loads is selected,and other appropriate routes are stored as backup.This section realizes WSNs routing in SDN environment and compares classic network and SDN network paradigms.To summarize,this work does a comprehensive research in network calculus theory,traffic matrix and QoS routing optimization in SDN,proposes methods or algorithms on SDN traffic matrix creation,network calculus service matrix and QoS routings.And more,this work does a comparison research on SDN wireless sensor networks routing protocols.