Research On Computing And Unloading Algorithm Based On Mobility Prediction In Mobile Ad Hoc Cloud Environment

The rapid popularity of mobile devices, represented by smartphones and tablets, has brought great convenience to people’s daily life. Due to limited CPU performance, battery capacity, storage capacity and other factors, mobile devices show poor performance on computation-intensive tasks. For example, the slow speed of operation, power down quickly, etc. The emergence of mobile cloud computing gives out a good idea to solve this problem. By offloading computation-intensive tasks from the mobile client to the target agents to execute, not only the processing time of the tasks can be greatly reduced, but also the energy consumption of the mobile devices can be cut down to a great extent.In mobile Ad Hoc cloud environment, the location of the mobile client node and the target agent node are changing dynamically all the time. Affected by the effective coverage of the wireless networks and the mobility of nodes, the network connection between client node and agent node is intermittent. As a consequence, the computation offloading failure problem arises when offloading compute tasks from the mobile client node to the agent node to execute. Although traditional computation offloading algorithms perform well in static offloading environment, they cannot overcome the computation offloading failure problem in mobile Ad Hoc cloud.To solve the computation offloading failure problem in mobile Ad Hoc cloud environment, this paper puts forward to five new algorithms that can be applied to dynamic offloading environment, based on the existing static offloading algorithms (i.e. MET, MCT, MinMin, MaxMin, Sufferage), including DynMETComm, DynMCTComm, DynMinMinComm, DynMaxMinComm and DynSufferageComm. Compared with the traditional static offloading algorithms, these algorithms not only take into account the communication overhead of task offloading, but also integrate with the processing mechanism as task offloading failed. That is, when the task offloaded to the agent node has been carried out failed, update the arrival time of the task to its failure time, so as to participate in the follow-up scheduling. For the sake of avoiding the overhead produced by task offloading failure to the greatest extent, a new algorithm based on mobility prediction (i.e. DynPredict) is proposed. This algorithm chooses a second-best agent from the agent nodes which have the ability to perform the task successfully to re-offload, when predicting the task of failed-execution, thus ensures that tasks can be completed smoothly. Simulation results show that DynPredict performs best on most of the performance indicators while DynMETComm performs worst. However, the performance of the on line algorithm, DynMCTComm, is usually close to or even greater than DynMinMinComm, DynMaxMinComm and DynSufferageComm, which are more complex than DynMCTComm.Our achievements can not only be applied to computation offloading in mobile Ad Hoc cloud environment well, but also lay a solid foundation for the future research on task migration. The research methods and ideas used in this paper can also enlighten the further studies of computation offloading in mobile cloud computing.