AUTONOMIC RESOURCE ALLOCATION FOR SERVICEBASED AND PARALLEL CLOUD APPLICATIONS

Abstract

Wireless Body Area Networks (WBANs) is an emerging platform for most of the humancentered applications, ranging from sports to medical performance monitoring. The ubiquitous and scalable nature of cloud makes it the most suited architecture for integrating it with WBAN to overcome its limitations. There are various research challenges in cloud-enabled WBAN for Quality of Service (QoS) improvement such as (i) architectural design for Cloud-enabled WBANs, (ii) Reliable and energy-efficient routing protocols for WBANs, (iii) Cloud resource allocation mechanisms, (iv) Semantics interactions, and (v) Data security. Among all the aforementioned research challenges, the architectural design and cloud resource allocation have been addressed in this research work. Therefore, to fill the void, we have designed developed Cloud-WBAN, a simulation toolkit, that not only brings the cloud computing system closer to the WBANs user (edge-of-things computing) but also automatically adjust the computing resources (at cloudlet) so as to maintain the service level agreements (SLAs) of the WBANs users on the basis of it's sensory data volume and application's type. The thesis is organized in six chapters as discussed below: Chapter 1: The introductory chapter briefs the details about the autonomic resource provisioning frameworks for WBAN services in cloud computing. It then highlights the research problems and objectives that leads to the proposed framework and resource allocation mechanisms. Subsequently, it explains the methodology used in the research and the contributions. In conclusion, it covers the thesis organization. Chapter 2: This chapter discusses the definition, constraints, and challenges of WBANs. It also highlights the type and characteristics of WBANs applications. It then details the definition, deployment and service models, and elasticity concepts of cloud computing. It investigate the role, need and use of cloud in empowering WBANs. Finally, it presents architectural design, research directions for quality of service (QoS) improvements for cloudenabled WBANs ecosystem. Chapter 3: It has been observed that there is a lack of an experimental toolkit for addressing the Cloud-enabled WBANs environments. This chapter presents the design and development of Cloud-WBAN, a simulation toolkit that brings the cloud computing system closer to the I WBANs user and automatically adjust the computing resources (at cloudlet) so as to maintain the SLA of the WBANs users on the basis of it's sensory data volume and application's type. The WBANs module also acts as the client entity, which is by in large missing in most of the cloud computing research till date. Some of the parameters configured in Cloud-WBAN are : experimentation time, number of WBAN users, WBAN modalities, data transmission rate, data sampling rate, mobility model of WBAN users, communication mediums/channels, number of virtual machines. This chapter also presents a hybrid resource provisioning framework which is the combination of autonomic computing and queuing model. The effectiveness of the resource provisioning framework is evaluated and the experimental results shows that it improves the resource utilization by at least 26% and response time by at least 49% as compared with other approaches. Chapter 4: This chapter argues the research gap in terms of optimal resource provisioning that predicts and automatically adjust the computing resources on the basis of sensory data volume and application's type. This chapter presents a hybrid autonomic resource provisioning framework, which is the combination of autonomic computing, fuzzy logic control and linear regression model. The effectiveness of this approach is evaluated under a real workload trace. The experimental results shows that it minimizes the cost by at least 27% and SLA violations by at least 78% as compared with other approaches. Chapter 5: This chapter presents a novel autonomic resource provisioning framework for parallel scientific application. We have developed an elasticity controller which is the combination of fuzzy logic control and autonomic computing. This controller computes the required amount of CPU core(s) considering the information about application's internal workload and resource utilization of the virtual machine. This is the first study that uses CloudSim toolkit to address : (i) execution of parallel application, and (ii) fine-grained resource provisioning. The experimental results show that this approach minimizes the finish time by up to 64% and increases the resource utilization by up to 36% compared with other approaches. Chapter 6: This chapter concludes the thesis with the discussion on the results and contributions. The future scopes of work also have been presented in this chapter.