A Self Organised Workload Classification and Scheduling Approach in IoT-Edge-Cloud Ecosystem

Internet of Things (IoT) has brought major changes in the way the workload is processed closer to the location of the data source. The need for near-to-real time provisioning of IoT workload has necessitated the emergence of Edge Computing. However, it is not entirely possible to shit the entire workload on to the edge layer due to the computational limitations of the edge devices. Hence, this challenge ended up with the amalgamation of IoT-Edge-Cloud ecosystem. But, one of the major challenges in this ecosystem is workload management in a self-organized manner (or according to the nature of the workload). This article tries to overcome this challenge by utilizing the benefits of Self Organized Map (SOM). This article comprises of three strands, 1) a SOM-based workload classification approach to handle the IoT workloads in a flexible manner, 2) an energy-efficient workload scheduling scheme using container-based virtualization, and 3) a workload migration mechanism based on secure caching technique. The proposed strands are evaluated using a simulated environment and the outcomes seem promising in contrast to generalized container-based workload scheduling.

I. Introduction

Internet of Things (IoT) has paved its way into every technological sector and even the entire way of living across the globe. The multi-fold benefits of IoT paradigm has increased the global dependence and presence altogether in the form of a transformative force [1], [2]. Although IoT has brought different benefits but it also gave birth to a large set of challenges. The huge amount of data (these days known as big data) has out-powered every technical solution. According to a recent report by Juniper [3], almost 38.5 billion IoT devices are connect in 2020 with an increase of 285% in comparison to 2015. However, the major problem raised in this report says that as the number of interconnected devices or things increase, the number of malfunctions or unrealized issues are bound to increase. If one has to believe with the figure, even 2-3% connected things comprises of nearly 500-750 million devices tangled with these unrealized issues (basically performance).