Priority based mobile edge cloud offloading (PBMECO) algorithm for latency sensitive applications

Abstract

In the last decades, technological improvements have made it possible to increase the complexity in mobile applications and to offer them to a large number of users. Therefore, the concept of task offloading has become an attractive solution for mobile applications to overcome the problems of limited processing capabilities and limited battery life because tasks are sent for execution to a remote infrastructure. Additionally, the concept of Edge Computing has gained considerable attention in recent years to achieve near real-time end-to-end communication, which is possible due to the close proximity of edge servers to mobile devices. In this work, we are tackling the challenge that different tasks of a mobile application need to be treated with different importance to reduce the latency of latency sensitive tasks. Therefore, we have implemented a Priority Based Mobile Edge Cloud Offloading (PBMECO) algorithm, which is operating within an Edge Cloud Computing environment. This means, that the tasks of the mobile application can be either sent to edge nodes or the public cloud. Our approach makes a joint offloading decision based on the pre-defined priorities HIGH, MEDIUM and LOW and the task’s resource requirements. These three priorities can be assigned to each task for both the communication and computation latency to enable more granular differentiation for finding offloading targets. Hereby, the communication latency specifies the importance with respect to network transfer whereas computation latency is used to set the priority for the execution time. We have additionally proposed a simulation framework, which is also used to evaluate our PBMECO implementation. The simulation framework is using Monte Carlo Simulation and we have applied three different real-world mobile applications to get insights about the performance of the algorithm compared to randomly selecting target nodes and ignoring task priorities. Firstly, the simulation results have illustrated that the performance of our solution remains constant when the number of HIGH prioritized tasks increases. Furthermore, the numerical results have shown that the communication latency of HIGH priority tasks can be reduced by up to 65% and it has been demonstrated that our offloading approach can achieve a latency reduction by up to 30% for tasks with computation latency HIGH.