Please use this identifier to cite or link to this item:
Title: Renewable Energy Based Efficient Framework for Sustainability of Data Centres
Authors: Aujla, Gagangeet Singh
Supervisor: Kumar, Neeraj
Keywords: cloud computing;data centers;renewable energy;sustainability;Energy efficiency
Issue Date: 15-Jun-2018
Abstract: Cloud computing (CC) has emerged as one of the most powerful technologies from past few years in which end users can access various services as per their demands on pay per basis concept. The need for on-demand state-of-art services (smart sensing, e-healthcare and smart transportation) and computing infrastructure have paved way to the powerful paradigm of CC. These enhanced exible and reliable attributes o ered by the CC platform have led to its widespread popularity amongst the academia and industry. Using virtualization technology, cloud service providers (CSPs) create multiple copies of virtual resources deployed over a physical server to provide various services to the end users. Such a virtualized environment and resources are hosted on large geo-distributed, service-oriented and critical computing infrastructure known as data centers (DCs). Ever since its inception in 2000, CC paradigm has witnessed signi cant transitions in its overall usage, size, computational ability and underlying technology used for accessing various services. However, the huge amount of data generated by various smart devices such as-smart phones, tablets, smart meter, body sensors and wearable devices has escalated the load on DCs to a great extent. Moreover, with the emergence of Internet of things (IoT), the demand of real-time data storage, access and processing at cloud has increased manifold. In recent years, data-intensive applications such as{e-health, e-commerce and e-banking generate a huge volume of heterogeneous data which varies with respect to time and its location. Such a huge amount of data needs to be collected, stored, analyzed and processed e ectively using DC infrastructure. To handle such massive data streams generated from all these applications, existiii iv ing DCs infrastructure needs to be expanded both horizontally and vertically. The dependence of smart communities also make it critical to expand the DCs with millions of servers operating at geo-located sites. The expansion of DCs has a strong impact on economy, environment, and performance of services o ered by the service providers. Therefore, this may lead to many associated challenges such as-high energy consumption, high operational cost, potential source of carbon footprints, grid instability and ine cient utilization of DC resources. However, the growing popularity of renewable energy sources (RES) has opened a new frontier for tackling the increasing level of carbon footprints globally. For the survival of mankind and other organisms, healthy and sustainable ecosystems are necessary. For this purpose, considerable e orts are being put forth globally for the energy transition from fossil fuels to ecologically sustainable systems. Sustainable energy is one of the best ways to serve the present needs without compromising the ability of future generations to cope with their needs. According to a recent survey, carbon emissions associated to energy consumption of DCs are expected to be double as compared to the previous decade. With such an increase in the growth rate of carbon emissions, the environmental sustainability level can also degrade proportionately. Therefore, to cope up with this challenge, the focus of DCs is slowly shifting from the mere usage of renewables towards sustainability of DCs. This is due to the reason that sustainable energy can provide myriad bene ts such as-reduced carbon emissions, grid load, and operational cost along with participation in environmental commitment. For above reasons, the major objective of this work is focused on sustaining the energy consumption of DC using RES. However, due to intermittent nature of RES, it becomes a challenging task. For this reason, four techniques in di erent environments are designed in this thesis to handle the intermittency issues of RES. In the rst technique, an SDN-based DC energy management scheme using RES has been proposed. The purpose of the proposed system is to sustain the energy consumption of DC using RES which it accomplishes successfully. To handle the intermittent nature of RES, EVs are used to support the energy consumption of DC. v Moreover, SDN provides energy e cient ow scheduling in the designed communication sensitive environment. In the second technique, an SDN-based edge-cloud environment has been designed for achieving sustainability of DCs using RES. In this technique, SVM classi es the incoming jobs on the basis of priority and delaysensitivity. The classi ed jobs are then routed using an optimal ow path to di erent cloud DCs or edge devices having su cient amount of renewable energy for execution. For this purpose, a two stage workload scheduling scheme has been designed for sustainability of DCs using RES while ensuring lower SLA violations. In the third technique, a multi-leader multi-follower Stackelberg game has been designed for renewable energy-aware resource allocation. The objective of the scheme is to sustain the energy consumption of DCs using RES alone. For this purpose, an optimal utilization of cloud resources was achieved by aligning them with energy consumption of DCs. The results obtained show that an optimal resource utilization helps to reduce the number of servers provisioned, which in turn minimizes the energy consumption of DCs. In this way, the objective of sustaining the DCs using RES is achieved. Finally, a container-as-a-service (CoaaS) model has been deployed in a geo-distributed cloud environment. To achieve the objective of sustainability, a renewable energy-aware multi-indexed job classi cation and scheduling approach has been designed. Using the global controller, the proposed scheme allocates the incoming workload to those DCs which have su cient amount of renewable energy to handle the incoming jobs. All the proposed schemes are evaluated using realistic workload and weather traces and the results obtained show that the proposed schemes successfully achieve their objectives in contrast to the existing schemes.
Appears in Collections:Doctoral Theses@CSED

Files in This Item:
File Description SizeFormat 
thesis.pdf12.65 MBAdobe PDFThumbnail

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.