Research And Implementation Of IP Backbone Traffic Scheduling Based On SDN

In traditional IP backbone network,generally adopts distributed scheduling and static routing strategy.These methods are lack of global perspective so as to lead to low utilization and congestion problem of network links.In 2012 SDN technology has been successfully introduced into IP backbone network by Google,which realizes centralized traffic scheduling mechanism between cross datacenters,and improve more than 95% of link utilization.It has provide the new ideas for the operators of its telecommunications WAN traffic scheduling.This thesis explores the feasibility problem of applying SDN technology into IP backbone network.Basing on the specialties of the IP backbone network,it designed a traffic scheduling architecture based on SDN and traffic scheduling algorithm with variable granularity.The aim is to effectively solve the low utilization rate of traditional IP backbone link,and link load imbalance and other issues.In this thesis,the main research results are as follows:(1)Analyze the five challenges of apply the SDN into IP backbone network: interactive pressure between controller and switches become the bottleneck of the performance,the inconsistence of the controller issued routing table entries and the difficulty to obtain the network information by controller,table entries capacity limits of basement equipment,high performance requirements of traffic scheduling algorithm.As to these challenges,the design is based on centralized IP backbone network traffic scheduling framework of route identification.The main features of the architecture: 1)preallocate the flow table,pre-distribute the global routing information to the underlying device in order to reduce the interaction between the controller and switches.2)path identifier,using a globally unique identifier to represent a globally unique path in order to resolve an inconsistency.3)classification of non-overlapping entries,the matching range of flow items does not overlap,and the flow table classification,combined with SDN south interface protocol to achieve network convergence flow monitoring mechanism.4)aggregate flow granularity and traffic scheduling are executed interactively,the convergence flow is adjusted to flow rate scheduling for convergence flow.It will be controlled in a reasonable range,to avoid excessive number of entries.It improves the efficiency of traffic scheduling,and the success rate of convergence flow.Based on the Floodlight + Mininet platform,it has realized the centralized IP backbone network traffic scheduling framework based on route identification and verify the feasibility of the architecture.(2)To solve the problem of link congestion,it designs the traffic scheduling algorithm with variable granularity.The algorithm is designed from the aspects of scheduling object and scheduling method,and proposes a flow size adjustment algorithm and traffic scheduling algorithm.The role of the flow size adjustment algorithm is to control the convergence flow in a reasonable range.In order to prevent the convergence flow from being too thin,a convergence flow aggregation algorithm and a convergence flow exchange algorithm are proposed.In order to prevent the convergence flow from being coarse,a convergence flow split algorithm is proposed.The role of the traffic scheduling algorithm is to reduce the load of congestion links by convergence flow migration.The three kinds design of traffic scheduling algorithm: 1)the MILP traffic scheduling algorithm,it is based on minimizing reroute built by business number as the goal to build mixed integer linear programming model.2)layered link dynamic upper limit scheduling algorithm,it is to minimize the reroute business number as the target to build MILP model,and to realize the load balancing by changing the link capacity limit.3)combination congestion link Algorithm,it is preferential migration flows through several congestion link converging flow,which will be several funnel flow equilibrium on the link to the article number of the alternate path.Verified by simulation,these three algorithms in different performance have different advantages.