Research On Multi-user Computation Offloading Strategy In Mobile Cloud Environment

With the rapid adoption and proliferation of mobile devices such as smartphones and tablet PCs,the demand has been steadily on the rise for more sophisticated mobile applications such as face recognition.These applications are typically both resource intensive and time sensitive.However,due to limited hardware resources available on mobile devices,users are experiencing longer delays and reduced battery life when running such applications.To tackle this problem,we explore the feasibility of migrating some tasks to cloud or Cloudlet which offers much more computation power and storage space.Still the challenge is that,even with offloading,the cloud resources will become constrained when handling a large number of mobile users.Therefore we aim at finding solutions which can dynamically migrate components from multiple users to the cloud in the case of limited resources,enabling mobile devices to run such applications in a low latency and energy efficient way.Specifically,this paper addresses the following three topics:(1)This thesis proposes a multi-user offline computation offloading strategy FMOCO for mobile cloud environments to reduce overall application latency and power consumption.Firstly,a genetic algorithm is used to obtain the initial optimal migration scheme for each user without considering the constraints of cloud resources.Then,those time periods which exceed the cloud resource limit is adjusted.Lastly,component replication and DVFS are used to further reduce response time and power consumption based on the implementation of components and their utilization of cloud resources.Simulation results show that the proposed strategy can achieve a good balance between response time of mobile applications and energy consumption of mobile devices.(2)This thesis proposes a multi-user offline computation offloading strategy UMOCO based on utility value in terms of reduction in high transmission time caused by long distance and low bandwidth between mobile devices and the cloud.This strategy takes full account of Cloudlet and cloud resource limitations.It determines the execution positions of components based on their utility values as well as priority to ensure the fairness among multiple users.Simulation results show that this strategy also maintains low energy cost while achieving low response time.(3)This thesis proposes a multi-user online computation offloading strategy MOCPR for hybrid mobile cloud environment where the performance of time slots affect each other.It is divided into two stages.In the first stage,based on the predicted number of requests in the next time slot,a certain number of virtual machines will be reserved in advance.In the second stage,the offline computation offloading strategy UMOCO will be invoked to obtain the current optimal migration scheme for each user.Simulation results show that this strategy can balance the performance of each slot,so as to achieve the best overall performance.In conclusion,three different computation offloading strategies under virtual machine resource constraints are proposed for mobile applications with general topology in both two-tier and three-tier mobile cloud environments.Our simulation results show that all three can effectively maximize battery life and minimize response time of all user applications.