Microservice Architecture and Efficiency Model for Cloud Computing Services

  • Lamees Mahmoud Mohd Said Al Qassem

Student thesis: Doctoral Thesis

Abstract

The popularity of cloud computing services for delivering and accessing infrastructure on-demand has significantly increased over the last few years. With the exponential workload increase in data centers, many businesses are shifting their workloads to microservices to benefit from the cloud-native paradigm. In addition, major cloud service providers have been progressively adopting the microservice approach as an alternative to the traditional, monolithic, virtualization solution based on virtual machines (VMs). However, their microservice frameworks are inefficient as they use the traditional threshold-based resource allocation strategies, which are ineffective for dynamic workloads due to reaction time delays. The growing adoption of microservice architectures (MSA) has led to significant research and development efforts to address their challenges and improve their performance, reliability, and robustness. Important MSA aspects not sufficiently covered in the open literature include efficient cloud resource allocation and optimal power management. Other MSA aspects remain widely scattered in the literature, including cost analysis, service-level agreements (SLAs), and demand-driven scaling. This thesis attempts to look at several strategies to improve the efficiency of future public cloud microservices. In this thesis, the need for microservice-based efficiency models is first highlighted. Then, the existing MSA frameworks and solutions are discussed with their limitations. To develop an efficient microservice system with performance similar to or close to the best cloud systems in the literature, we propose a microservice framework for service-level management of cloud workloads that is flexible, open-source, and easy to integrate. The proposed framework consists of four main models: microservice architecture, predictive model, reactive resource manager, and hybrid efficiency model. The microservice architecture comprises resource monitoring, performance evaluation, task scheduling, and workload processing services. The predictive model is a proactive autoscaling system that uses a learning-based forecast model to predict the future hardware resource utilization of the supported microservices. The predicted values will be used to adjust the resource pool. The reactive resource manager is an optimization model that reacts to changes in the system workloads to avoid SLA violations. Finally, the hybrid efficiency model is a resource-management framework for microservice architectures under cost-effectiveness and service-level agreement constraints. This is achieved by integrating the predictive and reactive models. The thesis further describes the architectural design, user interface, tool flow, and hardware back-end. The performance of the proposed system is then evaluated under different conditions, and guidelines for optimal system performance are provided.
Date of AwardMay 2022
Original languageAmerican English

Keywords

  • Web Services
  • Microservices
  • Efficiency
  • Orchestration
  • Autoscaling.

Cite this

'