Design And Implementation Of An Execution Optimization System For Real-time Data Processing Application In Edge Computing Environment

Real-time data processing applications on mobile terminals are emerging in an endless stream.These applications induce a significant surge in demand for data and computing resources that are indispensable.However,terminals cannot meet the needs of these applications due to their limited and weak capacity.The current solution that tasks will be offloaded to the cloud data center to obtain enough computing and data resources results in high processing delay due to the long propagation distance between the cloud and terminals,which is unacceptable for realtime data processing applications.In recent years,Edge Computing has been proposed as a promising computing paradigm,which aims at extending computing and storage resources to the edge of the radio access network,hence providing real-time,high-bandwidth and lowlatency access to radio network resources for mobile terminals.This paper aims at constructing an edge computing system for real-time data processing applications,which is faced with two research challenges: how to reasonably schedule tasks among terminals and edge servers,and how to keep necessary data for task processing in the edge layer as much as possible.The traditional research works ignore the impact of the limited and diversity capacity of edge servers on edge processing cost and data storage.These works are also not suitable for real-time data processing applications because they ignore the tight coupling of computing and data requirements.This paper takes the limited and diversity capacity of edge servers and the tight coupling of computing and data requirements into consideration.First,this paper proposes a task collaborative scheduling mechanism of terminals and edge layer,which includes two sub phases: terminal task scheduling mechanism based on the “best effort” principle and edge layer scheduling mechanism based on a priority scheduling policy.Second,this paper proposes a processing migration-cost-sensitive data cooperative storage mechanism for the edge layer,which aims at keeping necessary data for task processing in edge servers as much as possible and avoiding the data missing problem when processing tasks.Third,this paper designs and constructs an edge computing prototype system and an object detection application demo.Eventually,diversified and comprehensive experiments,including numerical simulations and experiments in practical environment,are carried out to verify the effectiveness and superiority of the mechanisms proposed by this paper.