En-OsCo

Abstract

The proliferation of the Internet of Things (IoT) has paved the way for many cloud based applications such as smart grid, healthcare, traffic management, finance, etc. In this vein, the need of transferring large data-streams to remote data centers is a key concern for modern Cloud-based IoT paradigms. This disrupts the remote Cloud Computing model, moving applications, data and computing resources to the logical extremes of the network. Thus, to handle streaming data in IoT environments, an efficient IoT-based computing model that can dynamically handle the interplay between Cloud and Edge data centers is required. In this direction, a recent paradigm, popularly known as Osmotic Computing, has emerged to ensure the acceptable performance of widely dispersed services. However, the burden of data-offloading across multiple data centers usually leads to a consequent increase in their energy consumption which in-turn will affect the overall Quality of Service (QoS) of the IoT-based applications. Keeping focus on all these issues, a consolidated decision making framework for Osmotic Computing, i.e., En-OsCo, is designed to ensure the energy-aware dynamic management of resources. The proposed framework incorporates four significant contributions: i) Resource monitoring of Edge data centers using Extended Kalman Filter, ii) Optimal dispatch of incoming services to the Edge/Cloud setup using Hyper-heuristics, iii) Minimizing the energy consumption of underlying data centers and reducing the service latency, and iv) Reducing the search space of Hyper-heuristics by keeping track of previously made decisions using Universal Streaming Monitoring. Further, in order to validate the efficacy of the proposed En-OsCo framework, ContainerCloudSim has been used in combination of HyFlex on PlanetLab datasets. The obtained results validate the purpose of the proposed scheme in minimizing the overall energy consumption of the computing setup while considerably reducing the latency.