Optimization Approach for Green Cloud Computing

With the increased use of computers and computing power, implementing cloud computing (CC) has become imperative in the present-day global scenario. While CC techniques give the user community and data centres many advantages, they also provide drawbacks regarding energy consumption and quality of service (QoS). These issues have paved the way for developing a new optimisation approach for green cloud computing (GCC). CC with its extensive infrastructure and a vast array of services, inherently necessitates a large amount of energy to operate effectively. The cumulative energy utilisation in CC can be immense, contributing to operational costs. Since cloud services grow to accommodate increasing data and computing requirements, finding methods to optimise energy consumption becomes vital for ensuring sustainability and cost-effectiveness. Research into energy-efficient practices and technologies is vital for addressing these challenges, aiming to maintain high performance and reliability. The contributions of the thesis mainly focus on optimising the energy efficiency of the cloud without compromising the QoS requirements. This research work is divided into two phases. The initial phase focuses on the analysis of energy models for VM consolidation on the practical servers using various load categories. Based on the observations of the initial phase, a novel hybrid heuristic algorithm-based energy-efficient cloud computing service (HH-ECO) is proposed that offers an effective and energy-efficient solution for optimising scientific workflows under various server load categories. Understanding the impact of different load categories on energy consumption helps to reduce the energy wastage significantly. Initially, analysis of energy models is performed in CC using a real time practical server with various load categories. By examining energy models using high-end and standard servers, each hosting multiple VMs for investigating load balancing and energy consumption. By monitoring server behaviour under various loads, key parameters affecting performance and energy usage are identified. The experimental setup comprises VMware ESXi hypervisors handling multiple VMs across one high-end server and three standard servers. Each standard server hosts 7 VMs, whereas the high-end server hosts 17 VMs. The practical test bed is implemented by consolidating VMs on physical servers. By emphasising CPU and memory utilisation as primary factors in energy consumption, the power consumption and performance are measured and verified using the testbed based on these load categories. Results in terms of CPU, memory, and power consumption for each server category are presented. An idle server with all VMs on proved low CPU and memory utilisation and consumed 60-75% of its overall power capacity. Likewise, an underloaded server had maximum utilisation (up to 40%) and slightly more power consumption. A balanced server showed an optimal performance with CPU utilisation of up to 63% and memory utilisation of about 58%, indicating that it could manage additional VMs. An overloaded server showed a high CPU of up to 80%, memory utilisation of 75% and maximum power consumption, demonstrating the requirement for VM migration to prevent performance degradation. By categorising server categories, this study aims to develop a novel approach for optimising resource utilisation, reducing power consumption, and improving QoS. Results proved that maintaining a balanced state through strategic VM migrations is an effective method for achieving these objectives. Considering the bottlenecks, cost, and risk associated with the practical environments, the results observed in the practical approach are further investigated by extending the research using the cloudsim tool to utilise the potential advantages such as cost-effectiveness, versatility and risk-free environment for validating and testing the energy efficiency. Thus, the next phase proposes a novel approach that uses a Hybrid Heuristic algorithm-based Energy-efficient Cloud computing service (HH-ECO) and offers an effective solution for resource allocation, task scheduling, and optimisation of scientific workflows. The complexity of scientific workflow increases energy consumption because of the huge computational and data processing demands they impose. Moreover, since scientific workflows often involve large-scale simulation and iterative processing tasks, substantial computational power and memory resources are required. Hence, achieving energy efficiency is significant in scientific workflow within a cloud environment. A Metaheuristic algorithm called chaotic-based particle swarm optimisation (C-PSO) is used in CC to optimise energy management by dynamically adjusting resource allocation and load balancing. It finds near-optimal solutions for efficient energy utilisation by generating global best plans without local convergence. C-PSO with adaptive mutation prevents the decline of global optima by recognising the best host for VM placement and ensuring an effective resource allocation approach. Considering the workflow task precedence relationships during C-PSO-based task scheduling, the novel hybrid heuristic technique efficiently solves the multi-objective combinatorial optimisation problem without dominance among the workflow tasks. HH-ECO focuses on executing non-dominant workflow tasks through adaptive mutation and an energy-aware migration strategy. C-PSO with adaptive mutation avoids the deterioration of global optima while finding the best host to place the virtual machine and ensures an appropriate resource allocation plan. During the execution of workflow tasks, the load category-based resource migration maintains the estimated QoS even in a dynamic environment. A Cloudsim-based simulation study delivers superior results compared to the existing methods, such as the Hybrid heuristic workflow scheduling algorithm (HHWS) and Distributed dynamic vm management (DDVM). The proposed approach significantly improves the optimal makespan and reduces energy consumption when compared to the existing methods. It meets the criteria for minimum energy consumption and provides an improved QoS to cloud data centres.