Architecture And Optimizing Application Placement For Mobile Edge Cloud System

Mobile edge cloud(MEC)is a model for enabling on-demand elastic access to,or an interaction with a shared pool of reconfigurable computing resources such as servers,storage,peer devices,applications,and services,at the edge of the wireless network in close proximity to mobile users.It overcomes some obstacles of traditional central clouds by offering wireless network information and local context awareness as well as low latency and bandwidth conservation.However,such cloud environments often contain unreliable nodes and links that are failure prone.Therefore,the deployment of applications requiring availability guarantees is a current research challenge.If such a placement algorithm is CPU-,memory-,network and availability-aware,the applications can use the resources as optimal as possible with a small failure probability.The optimal deployment of applications in the network infrastructure is a NPhard problem and,as a consequence,exact algorithms to solve it are not scalable.This thesis presents a comprehensive analysis of MEC systems,including the concept,architectures,and technical enablers.First,the MEC applications are explored and classified based on different criteria,the service models and deployment scenarios are reviewed and categorized,and the factors influencing the MEC system design are discussed.Then,the architectures and designs of MEC systems,the technical issues,existing solutions,and approaches are presented.To solve the availability-aware application placement problem,a distributed genetic algorithm was proposed to place service-oriented applications on a MEC,by defining a representation of an application placement in a biased-random-key chromosome and using a fault-tolerance distributed pool model.When compared with an existing Integer Linear Programming approach,simulations show that the distributed genetic algorithm is scalable and obtain near optimal performance results.Finally,the thesis has been summarized.In addition,the open challenges and future research directions of MEC and the distributed genetic algorithm are further discussed.