Resource Management And Optimization Under The Cloud-Edge Collaboration

With the development of wireless communication,big data and other technologies,IoT devices and their processing requirements have increased dramatically in recent years.However,the unbearable propagation latency and jitter make triditional cloud computing can not handle them efficiently.To mitigate this problem,edge computing is proposed,which is considered to have great potential in many application scenarios such as Smart Cities and Vehicle ad hoc Networks.It is to deploy resource-rich servers near the vicinity of the mobile users so that devices can offload their job requests to edge servers with low latency.Compared with cloud computing,edge computing has many advantages.For example,it has a shorter system response time,powerful localized computing capability,less data transmission load,and more secure and reliable distributed architecture.In order to take advantage of edge computing,it is imperative to establish a cloud-edge collaboration architecture with low-power and low latency.However,many problems still need attention to establishing the architecture due to user movement,server heterogeneity and other practical challenges.In addition,the reasonable allocation and arrangement of limited resources are crucial to establishing the architecture.In this thesis,we start from the first step of the actual implementation of edge computing and study the rational allocation of computing,storage,and network resources.The main contributions are summarized as follows.·Low-Cost Edge Server Placement Mechanism.In edge computing,edge server placement and online task dispatching are two fundamental problems that are critical to the rational allocation of limited edge resources.In this work,we first study how to place a limited number of edge servers.Inspired by divide and conquer,we innovatively propose "scenario division," which reduces the scale and complexity of the problem by dividing the original problem in time and space.The proposed algorithm can effectively reduce the split sub-problem to the classical facilicaty location problem for solving.Further,we pour the machine learning technology into the traditional method to propose an online task dispatching algorithm that considers historical requests and possible future changes jointly.Extensive simulations demonstrate that our algorithms outperform state-of-the-art works in the literature,and our online solution is comparable with previous offline solutions.·Efficient Application Configuration Mechanism.In this work,we found that the edge server placement dramatically impacts the performance of the initial application configuration.Therefore,we jointly study the edge server placement and the application configuration to minimize the weighted sum of the service cost and edge server opening cost.We propose a local-search based algorithm,named SPAC,to solve the problem efficiently.Then,we pave the way for application reconfiguration.Fortunately,our algorithm SPAC can be adapted for online application reconfiguration with only minor adjustments.Finally,we verify the proposed algorithm theoretically and experimentally.Extensive simulations demonstrate SPAC reduces the total cost by up to 60%compared with state-ofthe-art methods,and consistently outperforms the baselines in different parameter settings.·Online Task Dispatching and Scheduling.In this work,we study online deadline-aware task dispatching and scheduling in edge computing.We jointly consider the management of the networking and computing resources to meet the maximum number of deadlines.We propose an online algorithm Dedas,which greedily schedules newly arriving tasks and considers whether to replace some existing tasks in order to make the new deadlines satisfied.We derive a nontrivial competitive ratio of Dedas theoretically,and our analysis is asymptotically tight.Besides,we implement a distributed approximation D-Dedas with a better scalability and less than 10%performance loss compared with the centralized algorithm Dedas.We then build DeEdge,an edge computing testbed installed with typical latency-sensitive applications such as IoT sensor monitoring and face matching.We adopt a real-world data trace from the Google cluster for large-scale emulations.Extensive testbed experiments and simulations demonstrate that the deadline miss ratio of Dedas is stable for online tasks,which is reduced by up to 60%compared with state-of-the-art methods.Moreover,Dedas performs well in minimizing the average task completion time.