Distributed Computation Offloading of an application from mobile/IoT device to cloud

TL;DR

This paper proposes a distributed computation offloading approach for mobile/IoT devices to cloud, exploiting parallelism and VM redundancy to reduce execution time and data resend, thereby saving energy and improving performance.

Contribution

It introduces a novel method of scheduling independent application call paths on multiple VMs with failure recovery, optimizing execution time and data transfer in cloud offloading.

Findings

Reduced execution time by up to 43.43% in tested applications.

Minimized data resend during VM failures, saving energy.

Demonstrated benefits of parallelism and redundancy in cloud offloading.

Abstract

In Covid-19 pandemic, the number of users connecting to the Internet using mobile devices increased. People are doing there every task using mobile phones [16]. These devices are battery-powered and have limited computation capabilities. Their computational capabilities can be enhanced by computation offloading in which required computation is to be done on a third-party server on a cloud instead of the device itself. The cloud offers virtually infinite computation and storage. We proposed that by exploiting parallelism within an application call hierarchy we can decrease the execution time of off-loadable parts and minimize data resend in case of VM crash. We determine function call paths that are independent of each other within an application and schedule each of them on separate VMs in a distributed way. Wherever such independent paths merge, we collapse to a single VM and whenever the paths diverge again, we schedule multiple VMs. If any single VM fails another copy will be created. However, only the code and data associated with the crashed VM needs to be re-transmitted from the client device. In the case of face reorganization application and montage application we decrease execution time to 27.5% and 43.43% respectively. Whereas the data resend in case if any of both VMs crash will be the portion of the application that had been offloaded to respective VM at depending upon the level of parallelism they have which save mobile battery in case of Resend. We will also discuss the energy consumption effect of using multiple Vms for a job VS single Vm for the same job.