Computation Offloading In Heterogeneous Wireless Cloudlet Network

With the advent of Internet of things,mobile devices are expected to carry on more and more computation-intensive applications,and mobile cloud computing(MCC)also comes to expand local computing capabilities.Infrastructure-based cloudlet,which has been deployed in public wireless metropolitan area network(WMAN),can provide cloudlet-based computation offloading service for mobile users to improve application performance.However,with the explosive increase of offloading requests,the current single cloud environment still encounter resource depletion,which will lead to offloading failure and user dissatisfaction.Therefore,we have proposed to introduce resource heterogeneity to enhance the reliability and scalability of offloading services.In such a heterogeneous cloud scenario,which integrates mobile cloudlet with infrastructure-based cloudlet,we mainly focus on offloading strategy and task allocation to efficiently utilize multiple resources to accomplish application execution in an offloading service.To realize optimal resource selection,we have proposed a mobile cloudlet assisted offloading strategy.When the infrastructure-based cloudlet becomes unavailable due to insufficient resources,mobile cloudlet can help to provide computation offloading service.In the scheme,we can first realize energy-efficient collaborative task execution with the infrastructure.When the cloudlet server becomes too overloaded to meet user requirement,we can determine whether task transfer in cloudlet network is feasible for current task by a load balance based judgment method.Finally,mobile cloudlet can undertake those failed tasks and accomplish task execution under control of the central scheduler.Simulation results demonstrate the performance improvement of heterogeneous computing platform and effectiveness of resource selection strategy.To realize computation offloading in the mobile cloudlet,we have developed a centralized task scheduling algorithm to tackle task allocation issues.Considering the intermittent connection of cloudlet nodes due to their mobility,first,we need to guarantee the service reliability by means of worker qualification judgment based on properties of mobile cloudlet.Then,we have proposed different task allocation algorithms for different user requirements,such as balance of energy consumption and makespan minimization.Simulation results show that our proposal can effectively prolong network lifetime and reduce task response time.Also,to address the acquisition of global knowledge for central scheduler,we have developed an energy efficient cluster chain based data aggregation mechanism.Under design of the dynamic frame structure and mobility management,we can realize information transfer and aggregation.Simulation results validate 40% improvement of energy efficiency and stable packet delivery of the proposed mechanism.Above all,our proposed offloading strategy and allocation algorithm can realize effective combination of infrastructure-based cloudlet and mobile cloudlet,and obtain improvement of service reliability and scalability.