Service Function Chain Scheduling With High Reliability And Low Latency

In order to adapt to future personalized application scenarios and extreme performance requirements,5G network makes full use of emerging technologies such as edge cloud,software-defined network and network virtualization to provide high reliability and low latency network services for augmented reality,remote driving and industrial Internet applications.Using software-defined network and network virtualization technology,multiple logical networks can be created and run on the same infrastructure according to demand,decoupling network functions from hardware,and realizing flexible and extensible networks to meet different differentiated needs.Service function chain,as a way of task execution in virtual network,allows various network service functions to be linked together to form an end-to-end complete service.A key problem is how to schedule each network function in the service chain in the virtual network environment with limited resources,so as to ensure the reliability and minimize the delay.In the existing work,researchers mostly consider the deployment optimization problem of the service chain in the network when the types and quantities of all service chain requests are known.However,we propose the online scenario where the service chain requests arrive dynamically,which makes the problem more challenging.Second,we considered the time overhead introduced by redeploying on the virtual machine,so our solution required a tradeoff between redeploying the virtual machine and maintaining a service chain waiting state until a suitable node was available to run.Finally,in order to achieve the goal of high reliability and low latency of the network,how to decide the number of redundant instances of the virtual network function and reduce the time cost are also the issues concerned in this paper.In this paper,the scheduling problem of service chain with high reliability and low delay in virtual network is studied.The main work is as follows:(1)Firstly,it defines the scheduling problem of the service function chain in a dynamic network environment.Considering the scenario where the service function chain arrives randomly,how to schedule the network function in the service chain to the appropriate computing node in the heterogeneous network for execution,so that the final completed time of the request is within the required deadline.In addition,the computing nodes in the network may fail.This paper also needs to determine the amount of redundant tasks to achieve the reliability constraints given by the request.(2)RDSSA(Reliability-Aware Dynamic SFC Scheduling Approach)based on reinforcement learning is proposed to solve the problem of service chain scheduling.Experimental results show that RDSSA algorithm has a higher success rate than other rulebased scheduling algorithms and classical reinforcement learning DQN(Deep Q Network)algorithm.(3)For the reliability scheduling problem of service function chain oriented to directed acyclic graph,in order to solve the reliability dependence problem between tasks,we use the recursive reliability model and establish the optimization model of the problem.A DAG-SFC reliability scheduling algorithm considering task fairness(FDSRSA)is proposed to maximize the success rate while satisfying high reliability constraints.