Adaptive job allocation scheduler based on usage pattern for computing offloading of IoT

Abstract

Recently, with the rapid growth of information technology (IT), diverse studies have been carried out for the grafting of devices based on the Internet of Things (IoT) for use in real life. With certain sensor functions and downsized mobile devices, IoT devices have improved users' work efficiency, ease of mobility, and convenience in terms of not being restricted by location. In the case of IoT devices as such, computing offloading is regarded to be very important to overcome issues of limited computing power and storage capacity and the limitations of built-in batteries. For the computing offloading of IoT devices, diverse job allocation techniques considering performance resources have been studied. However, since only the static performance, dynamic performance, or performance and battery size of IoT devices are considered in job allocation, job reallocation problems are caused by battery consumption due to the use of patterns in which users execute certain applications. In this paper, an adaptive job allocation scheduler (AJAS) that adaptively redistributes the jobs allocated to IoT devices based on user behavior patterns is proposed. The AJAS allocates jobs using the dynamic performance resources and battery consumption rates of diverse IoT devices. In addition, the AJAS measures the battery consumption rate of user applications executed in the IoT device to assess whether the allocated jobs can be processed. The AJAS identifies IoT devices that cannot process jobs and minimizes states in which allocated jobs cannot be processed due to battery exhaustion and delay time due to job reallocation. For verification, an AJAS is designed and implemented to show that the AJAS improves device availability for job processing. (C) 2019 Elsevier B.V. All rights reserved.