Energy Efficient Resource Scheduling Algorithms for Cloud Computing

Abstract

Cloud Computing presents an exciting new horizon for the Information Technology (IT) industry. It provides a cost-effective solution, as it allows hosting of storage, computational and supported network services on a shared infrastructure of physical servers. It offers utility oriented IT services to the users worldwide. Cloud computing empowers companies to host engineering, scientific, and business applications, and makes them accessible across the world. However, to accommodate the increasing trend of online computing applications and the ever growing massive amount of data, the data centers are also continually expanding in size. This means a huge consumption of electrical energy that ultimately results in high operational costs and emission of green house gases into the environment. Therefore, to curb this unsustainable increase in energy consumption, research on "energy efficient computing" becomes a critical need and a roadblock of great magnitude with respect to energy efficient utilization of resources while preserving the desired users' Quality of Service (QoS) standards. This thesis focuses on cloud resource management with special attention to energy efficient resource scheduling and Green Service Level Agreements (GSLAs). Initially, it compares, analyses and reports on the existing energy aware resource scheduling frameworks and heuristics on the basis of various aspects. From the literature survey, it is apparent that there is a great energy saving potential with respect to the system operations and workload specificities. This in particular holds for small and medium sized datacenters which can’t afford expensive hardware and renewable energy sources to save energy. To address this challenge, this thesis presents novel techniques, models, and algorithms for the cloud environment. To achieve the goal of improving resource utilization and reducing energy consumption, the Energy efficient Cloud Resource Scheduling (ECRS) algorithm has been designed. It gathers information such as host utilization level, power consumption, number of VMs and their state etc. regarding available resources. Using this information along with an estimate of resources required for future requests, the energy efficiency of cloud compute cluster is assessed. This assessment is used to manage resources energy efficiently. Further, to involve the users in eco-system cloud services, a Green Service Level Agreement (GSLA) aware Cloud Resource Reservation (GSLACRR) algorithm has been proposed. It is an endeavor to incline the users towards sustainable computing through user negotiation strategies. It ultimately results in cost benefits for users as well as service providers and helps to minimize the energy consumption. To address the various cloud resource management challenges such as performance, energy efficiency etc., an energy efficient cloud framework, named ACACloud has been proposed, designed, developed and tested, and further used to demonstrate the applicability of the proposed algorithms. The proposed energy efficient cloud framework, ACA-Cloud, has four layer architecture, with Host Controller (HC) as the base layer, providing actual resources to a user. The second layer is Cloud Cluster Controller (CCC) which is responsible for monitoring the status of all the physical machines/hosts and making appropriate resource management decisions in response to the current workload and incoming user requests. The next layer is Account Manager, which renders user authentication management to the framework to handle different account types and authentication. The top most layer is the Web Portal Interface, through which users can submit their requests for cloud services. This proposed framework has been installed at Thapar University, Patiala. The experimental results reveal the competitive performance and usage of the proposed algorithms implemented on this framework. Finally, the proposed algorithms are compared with the existing one to validate the outcome. The results show that the scheduling algorithms successfully and collectively address the issues of energy efficiency and performance to establish an efficient Cloud.