Mobility and Execution Time Aware Task Offloading in Mobile Cloud Computing

Abstract

Nowadays, mobile devices perform almost all tasks that can be performed by a computer but empties the battery and consumes memory. It is not necessary to execute the tasks on mobile devices; instead, it is executed in the far-away cloud. To save battery energy, the tasks are offloaded and hopped through several access points to reach the cloud and executed which increased the execution time of the task. Therefore, to save execution time and energy, the tasks are offloaded to a nearby cloudlet and as the device moves, the cloudlet and mobile device are disconnected. The mobile device is connected to the next cloudlet; while the offloaded tasks are partially executed in the previous cloudlet VM migrates to the new cloudlet. The previous cloudlet examined the remaining execution time of the task. If it is less than the connection time, the task is finished and the result is transferred to the new cloudlet; otherwise, the task is offloaded to the new cloudlet. It is seen that the mobility and execution time aware task offloading model reduces the execution time and power consumption by 21-40% and 26-34% approximately to the existing mobility-aware offloading approach.

Task Scheduling with Altered Grey Wolf Optimization (AGWO) in Mobile Cloud Computing using Cloudlet

“…The best task is assigned to the cloudlet's last group of VMs after the maximum iteration. The execution time and power consumption of the allotted tasks are found using equations 10 and 11 from our previous research article Mobility and Execution Time Aware Task Offloading method (METATO) [22]. This way, the tasks within one second are assigned to the cloudlet and continue with the next second of tasks.…”

A Novel Approach to Improving Distributed Deep Neural Networks over Cloud Computing

“…In recent years, DL has seen significant success, notably in computer vision; DNNs have improved their accuracy of the state-of-the-art in many types of visual identification tasks. Hence, also there are a lot of acts that have developed classical DNNs from augmentation of data, a function of loss, the structure of the network, an algorithm of optimization, the function of activation, neutral vector variables decorrelation of aspects [19].…”