Edge Computing Cooperative Strategy In Metro Optical Networks

With the rapid development of mobile Internet and 5G technology,more and more emerging applications that require unprecedented access speed and low latency have emerged.At the same time,since the metro optical network close to the user side bears more and more traffic and tasks,the rapidly upgrading metro optical network will have more intelligence and capacity.Multi-access Edge Computing(MEC)is an emerging technology that can solve the requirements of computationally intensive and delay sensitive applications.It deploys storage,computing and other resources to the network edge to provide users with low-latency access.However,the computing capacity of edge servers is very limited compared with remote cloud data centers,so it is easy to overload the edge servers by directly offloading tasks to the edge servers closest to users.The offloading of tasks on overloaded edge servers will cause long queuing latency,which may eventually cause the task response time to exceed its latency threshold or bring a poor service experience to users.Therefore,how to cooperatively offload the tasks that arrive at the edge server through the metro optical network to better utilize the resources of the edge server and the metro optical network and provide low-latency services for applications is a problem worth studying.This paper aims at the phenomenon that some edge servers are overloaded due to the limited computing capacity of edge servers and the unbalanced load of each edge server.Taking the edge computing system interconnected by metro optical networks as the background,the problem of cooperative offloading of tasks arriving at edge servers in metro optical networks is studied.The work of this paper mainly focuses on three parts:1)The characteristics of edge computing systems interconnected by metro optical networks and computing tasks are analyzed,and their offloading constraints and characteristics are analyzed;2)Based on the SARSA algorithm,the cooperative offloading and resource allocation algorithm for data partitioned-oriented application tasks is proposed and implemented.The simulation results show that the offloading algorithm based on the SARSA algorithm can achieve the highest offloading success rate within the range of delay constraints.3)A cooperative offloading algorithm for dependency-aware tasks based on genetic algorithm is implemented.It jointly optimizes the allocation of routing and computing resources,while also taking into account task dependencies.The algorithm is divided into three parts:optimization of the task dependency graph;the initial determination of the task offloading strategy and resource allocation scheme;and the further optimization of the allocation of transmission resources.The simulation results show that the cooperative offloading algorithm based on genetic algorithm has better performance in terms of average response time and offloading success rate,and can improve the utilization of system resources.