Modelling Mobile-X Architecture for Offloading in Mobile Edge Computing

Abstract

Mobile Edge Computing (MEC) assists clouds to handle enormous tasks from mobile devices in close proximity. The edge servers are not allocated efficiently according to the dynamic nature of the network. It leads to processing delay, and the tasks are dropped due to time limitations. The researchers find it difficult and complex to determine the offloading decision because of uncertain load dynamic condition over the edge nodes. The challenge relies on the offloading decision on selection of edge nodes for offloading in a centralized manner. This study focuses on minimizing task-processing time while simultaneously increasing the success rate of service provided by edge servers. Initially, a task-offloading problem needs to be formulated based on the communication and processing. Then offloading decision problem is solved by deep analysis on task flow in the network and feedback from the devices on edge services. The significance of the model is improved with the modelling of Deep Mobile-X architecture and bi-directional Long Short Term Memory (b-LSTM). The simulation is done in the Edgecloudsim environment, and the outcomes show the significance of the proposed idea. The processing time of the anticipated model is 6.6 s. The following performance metrics, improved server utilization, the ratio of the dropped task, and number of offloading tasks are evaluated and compared with existing learning approaches. The proposed model shows a better trade-off compared to existing approaches.