Mobility Aware Computation Offloading Model for Edge Computing

Abstract

The technological advancement of mobile devices such as laptops, smartphones, wearable devices, and other handheld devices has resulted in the emergence of various user applications in the entertainment, learning, social networking, and community computing sectors. However, these devices have limited capacity and battery charge to process computation intensive tasks. As a result, offloading is one of the most important approaches for connecting mobile devices and powerful systems. It minimizes complexity and improves mobile computing capacity. Computation on cloud computing is a powerful solution for computation of tasks on devices with limited computing capacity. Still, due to the distance issue the energy usage for transmission and the network delay to send and receive computing requirement degrades the performance. As the result, edge computing is the promising enabler for latency sensitive and energy efficient computation in proximity. However, user mobility and the restricted coverage of Edge Computing (EC) server service pose new challenge for computation offloading. Delivering task offloading requests to servers and results to users is difficult in networks where user movement is frequent, resulting in increased latency, higher energy consumption, and inefficient resource utilization. The key problem in offloading computing to edge server is, determining how to efficiently decide to assign computation tasks to edge servers in such a way that the decision captures the mobility inherent in mobile devices and results in minimal latency, energy consumption and execution time during application running. In this paper, to tackle the constraints related to intermittent connectivity due to mobility, network changes, device heterogeneity, and resource load mobility aware computation offloading model is proposed. Dependency graph-based mobility prediction is adopted to trace next mobility locations and edge servers. Dependency graph with fuzzy logic algorithm is proposed which considers available computation and communication resources both on device and server. This fuzzy logic decision considers edge device battery level, data size, bandwidth, network coverage, delay sensitivity, edge server Virtual Machine (VM) utilization and load of the server. The proposed model is implemented using PureEdgeSim simulator with mobility traces. The simulation result was analyzed with respect to task failure rate, failure rate due to mobility, energy utilization and average task processing delay. The analysis of the simulation result is done using a comparative analysis with the state of the art works fuzzy decision-based cloud-MEC collaborative task offloading management system (FTOM) and a fuzzy decision tree-based task orchestration (FDT) considering task failure rate, energy utilization, task processing latency, and task completion delay. The proposed model performs better with minimum task failure rate, energy consumption and task processing delay compared to FTOM and FDT with an increase of mobile devices. For example, with an increase of mobile devices from 100 to 700, task failure rate increase from 2% to 70% in FTOM, 1% to 45% in FDT and from 0.9% to 40% in the proposed model. The simulation result also affirm that the proposed mobility aware computation offloading provided improved task failure rate with mobility when compared to FTOM, which is the worst one. The energy utilization with an increase of mobile devices and the task processing delay time (0.45 s for 700 mobile devices, 0.73 s for FTOM) for the proposed model is better than the other task offloading schemes.