Algorithms and Techniques for Dynamic Resource Management across Cloud-Edge Resource Spectrum

Abstract

An increasing number of Internet of Things (IoT) and other latency-sensitive applications are cloud-hosted. However, limitations in performance assurances from the cloud, and the longer and often unpredictable end-to-end network latencies between the end user and the cloud can be detrimental to the response time requirements of the applications, specifically those that have stringent Quality of Service (QoS) requirements. Although fog/edge resources, such as cloudlets, may alleviate some of the latency concerns, there is a general lack of mechanisms that can dynamically manage resources across the cloud-edge spectrum. The problem becomes even more challenging when performance interference on multi-tenant fog servers along with workload variations, and user mobility are considered. To address these concerns, this dissertation presents the design and implementation of the Dynamic Data Driven Cloud and Edge Systems (D3CES) framework. It defines approaches to utilize the performance metrics collected from adaptively instrumenting the cloud and edge resources to learn and enhance performance interference-aware models of the distributed resource pool. In turn, the framework optimizes resource provision in a way that satisfies service level objectives (SLOs) while minimizing cost to the service providers. This dissertation evaluates the approach on a variety of real world scenarios.