Quality-of-Service Optimization In Edge Computing

With the rapid development and aggressive popularization of Internet of things(Io T),the number of inter-networked physical devices(e.g.,mobile phones,tablets,smart watches)has increased dramatically.Meanwhile,various Io T applications with low response latency requirements,such as virtual reality,augmented reality and online mobile games,are constantly emerging.The traditional network takes cloud servers as computing platform,indicating all data needs to be transmitted to the cloud servers for processing,which cannot meet the low response latency requirements of emerging applications in the Io T era.In order to overcome shortcomings of cloud computing,a promising computing paradigm of so-called edge computing has emerged as the times require.The main idea of edge computing is to deploy various services to the edge network closer to end users,so as to realize the effective utilization of computing,storage and communication resources.Quality-of-Service(Qo S)driven resource management is one of the hot research topics in the field of edge computing.It guides the resource allocation of edge computing systems by using assorted Qo S metrics,so that individual components of edge computing systems are able to cooperate with each other to meet specific design requirements.However,existing Qo S-driven resource management techniques neglect both mobility and reusability of end devices,which give rise to a series of problems such as reducing service experience duo to the large response time difference between end users,decaying system lifetime due to the redundant executions of overlapping tasks,and diminishing computation accuracy due to the prohibition against sharing end devices.Considering the mobility and reusability of end devices in edge computing,this thesis focuses on characterizing the system Qo S from three typical aspects of response time,system lifetime,and computing accuracy.To be specific? Considering the mobility of end devices in edge computing,this thesis designs an edge server deployment scheme that not only reduces the average response time of all users but also guarantees the fairness in response time among different users across domains.The proposed scheme consists of offline and online stages.At offline stage,this thesis first uses integer linear programming technology to model the placement of edge servers and the mapping of base stations to edge/cloud servers,and then utilizes integer linear programming solver to deal with the problem of static response time optimization.To accommodate the mobility of end devices,the static mapping from base stations to edge/cloud servers needs to be adjusted at runtime.Therefore,at online stage this thesis first uses cooperative game theory to model the remapping of base stations to edge/cloud servers,and then utilizes Nash bargaining method to solve the problem of dynamic response time optimization.This thesis also develops a mobility-aware response time verification platform.Extensive simulations are carried out on this platform to verify the effectiveness of the proposed scheme in reducing the average response time of all users and ensuring the fairness in response time between different users across domains.? Considering the mobility of end devices in edge computing,this thesis designs an energy-efficient task scheduling mechanism that balances the energy consumption between battery-powered end devices for prolonging system lifetime.The proposed mechanism consists of offline and online stages.At offline stage,this thesis first adopts the mixed integer linear programming technology to model the system lifetime optimization,and then leverages the mixed integer linear programming solver to produce an optimal static task schedule.To accommodate the mobility of end devices,the static task schedule needs to be adjusted at runtime.Therefore,at online stage this thesis proposes a cross-entropy based task scheduling strategy to wisely tune task executions according to the energy state of end devices.This thesis also develops a mobilityaware system lifetime verification platform.Extensive simulations are carried out on this platform to verify the effectiveness of the proposed mechanism in prolonging system lifetime.? Considering the reusability of end devices in edge computing,this thesis designs an energy-efficient application scheduling mechanism that improves the computation accuracy of the whole and individual applications powered by fluctuant renewable energy supply.The proposed mechanism consists of producing local and local-remote application scheduling stages.At producing local application scheduling stage,this thesis devises application-level and component-level energy assignment schemes with the assumption that all applications are locally executed on end devices.At producing local-remote application scheduling stage,this thesis devises a simple yet effective computation offloading scheme that intelligently offloads the computation of select-ed applications to remote edge servers for dealing with the fluctuant renewable energy supply.This thesis also develops a reusability-aware computation accuracy verification platform.Extensive simulations are carried out on this platform to verify the effectiveness of the proposed mechanism in enhancing application computation accuracy.