Design And Simulation Of High-Performance Multi-Level Scheduling Model For SDN Data Forwarding Plane

The switching devices are at the core of the network and the performance of switching devices need to be improved to meet the stormy growth of network traffic.With the increased new types of network service flows and the expanding scale of network users,it requires that switching equipments should not only be able to refine the network flows with different attributes,but also provide differentiated services for different transmission objects.These demands are difficult to achieve for general QoS scheduling policy,therefore,we need to study and improve the hiberarchy QoS scheduling policy.The appearance of a new network architecture named SDN has changed the structure of traditional switching equipment and realized the separation of its control plane from its data plane,which make the switching equipments more flexible.Recently,research on how to apply multi-level scheduling techniques to the SDN data forwarding plane has attracted much attention.This paper proposes a multi-level queue scheduling model suitable for SDN data forwarding processing units based on a general multi-stage scheduling model defined by the DSL Forum in accordance with the TR-059 protocol.Then,through software design and simulation to verify the correctness and validity of the multi-level scheduling model,and to provide guidance for the design and revision of multi-level scheduling schemes before it puts into hardware implementation.First of all,this paper analyzes the system structure and module function of SDN data forwarding processing unit,and extracts some key modules for multi-stage scheduling model simulation.Then,it introduces the design idea of each module of the hierarchical scheduling model.The hierarchical scheduling model is developed by C++ programming language.In this simulation model,the scheduling process of input data is mainly divided into five levels,which are enqueue-level scheduling,service-level scheduling,user-level scheduling,destination subnet-level scheduling,and output-port-level scheduling.The key algorithms implemented and improved in this paper include WRR,DWRR,and WFQ.Finally,the performance of the multi-level scheduling model and the scheduling algorithms is tested under various input loads and different types of service flows.Compared with the test results of single-level scheduling model,it is proved that the multi-level scheduling scheme designed in this paper can provide a hierarchical QoS,and can control the output datas when the network traffic is congested.What's more,it can guarantee the output network traffic of service flow level,user level,and subnet level at a predetermined ratio.