Resource management and elasticity control in edge networks

Abstract

Edge computing has been recently introduced as an intermediary computing entity between Internet of Things (IoT) deployments and the cloud, providing data or control facilities to participating IoT devices. New modern applications have emerged at the intersection of these domains that require real-time access to sensor data from the environment with low latency. Traditionally, executing such applications happens in resource-rich environments such as the cloud. However, the massive amount of data transfer, heterogeneous devices, and networks involved affect latency, which in turn causes high latency in IoT systems. To overcome these bottlenecks of the current cloud-centric systems, researchers from academia and industry have suggested deploying IoT applications with stringent requirements closer to the edge of the network - thus, better satisfying their various demands such as high availability, performance, or privacy. Nevertheless, centralized resource management — typically in the cloud and evident in today’s IoT-cloud architectures is one solution but requires cloud control structures to be always available and within low latency. Therefore, this calls for novel resource management techniques deployed and executed on edge networks and aids various end-users in deploying and managing an application in the target edge network. Nonetheless, edge environments have been identified as dynamic, resource-constrained, characterized by uncertainty, and heterogeneous computing devices. Specifically, these characteristics of edge networks have profound implications for execution and managing application functionality at runtime.This thesis provides novel methodologies and edge-based resource management frameworks to assist and manage runtime aspects of IoT applications (i.e., edge applications) executed on resource-constrained edge networks. More precisely, we develop a novel edge-based system that provides resource management features based on three perspectives: (1) resource discovery in a decentralized manner, (2) controlling elasticity of edge applications at the edge, and (3) resource coordination at the edge. Our proposed edge-based system enables and supports the execution of edge applications on various edge networks and maintains their correct functionality throughout the execution time. The proposed edge-based system is scalable and expandable in terms of functionalities and designed to work under various circumstances that can appear in edge networks (e.g., dynamicity).