QoS-aware Resource Utilization and Allocation in Cloud Computing

Abstract

Cloud computing provides resources on-demand on a rent basis through the internet. The cloud users request various services like computing power, storage, networking, etc, from the Cloud Service Providers (CSPs) on a rent basis. The demands of cloud users are increasing day by day because using the resources on rent is much easier and economical

than purchasing a system with their own storage media, computing power, and network- ing etc. It is very challenging for CSPs to handle these service requests and manage

the resources efficiently. Cloud computing has transformed the delivery of computational

services to users as on-demand, customizable services, making them resource- and cost- effective. However, several obstacles prevent the widespread application of this technol- ogy, especially in educational institutions, central banks, and Cloud- Enterprize Resource

Planning (C-ERP) etc. Other characteristics, such as on-demand service, resource pool- ing, pay-per-use, flexibility, etc., have enticed scientists to put scientific applications on

the cloud. For successful exploitation of virtualized resources in the cloud, efficient re- source allocation based on task resource utilization is required to maximize performance and reduce execution time. Scientific Computing leverages cutting-edge, high-performance computing capabilities to

handle complex problems in various scientific fields, such as weather forecasting, earth- quakes, subatomic particle behavior, turbulent flows, industrial processes, etc. As the

resource requirements for resolving scientific problems are dynamic, there is a need for a platform capable of managing the data; as mentioned earlier, storage and processing limits in scientific applications. Further different scientific applications are categorized on

the basis of their basic shape, size and structure which can be deployed on the cloud envi- ronment. These applications are further classified based on their computational runtime and task dependencies. To achieve the set objectives, an extensive literature survey of existing resource utilization

iii

and QoS-aware resource allocation techniques in cloud computing has been done. Besides, the survey of existing resource utilization and allocation techniques, these techniques are compared on the basis of different QoS parameters such as time, cost, power, SLA violation, scalabilty, CPU load, memory usage etc. From the literature, it can be inferred that resource utilization based resource allocation is a challenging issue which needs to be handled carefully. To address these problems, a resource utilization based resource allocation technique ”P2C” has been proposed. Further, various case studies have been taken into account to find out the obstacles that are hindering the adoption of cloud computing in higher learning institutions in Kenya, central banks, and consumer goods industry in United States’. Concerns regarding the dependability of CSPs and a lack of cloud computing expertise have been recognized as the primary obstacles to cloud computing’s widespread adoption. The proposed dependency and time-based scheduling technique considers the Parent to Child (P2C) node dependencies, Child to Parent node dependencies, and the time of different tasks in the workflows. The proposed P2C technique emphasizes on proper utilization of the resources and overcomes the limitations of the time-based schedulers for scientific applications. Furthermore these scientific applications such as CyberShake, Montage, Epigenomics, Inspiral, and SIPHT are represented in terms of the workflow. The tasks can be represented as the nodes and relationships between the tasks can be represented as the dependencies in the workflows. All the results have been validated by using the simulation-based environment created with the help of the WorkflowSim simulator for the cloud environment. It has been observed that the proposed approach outperforms the mentioned time and dependency-based scheduling algorithms in terms of the total execution time by efficiently utilizing the resources.

Concerns regarding the dependability of cloud service providers and a lack of cloud com- puting expertise have been recognized as the primary obstacles to cloud computing’s widespread adoption. Inadequate support and training from cloud service providers and government policies on cloud computing, data security, and confidentiality further raise issues. Influential factors preventing cloud computing adoption in central banks are data

protection, privacy, and risks. Moreover, the design methodology, implementation and re- sults of the case studies in higher learning institutions in Kenya, central banks, and cloud

ERP for consumer goods industry is also explained in detail. Furthermore, an overview of the different factors considered for the cloud computing adoption and utilization are highlighted for the different case studies of cloud computing utilization. Hypothesis analysis of proposed P2C techniques and case studies of cloud computing utilization in different sectors such as higher learning institutions in Kenya and cloud computing utilization for ERP in consumer goods indusrty has been done. A three-stage analysis process have been used culminating in the identification of the key factors that hinder the adoption of cloud computing and evaluation of the research hypotheses. The first stage was the analysis of the demographic factors and the state of cloud computing adoption and the type of cloud model and cloud services adopted. The second stage was the regression analysis and analysis of the variations of the factors that influence the adoption of cloud computing. The analysis enabled the testing of the hypotheses and

identification of the key influential factors in the study and formulation of recommenda- tions to address those factors. The factors in each of the research contexts were tested

and evaluated to determine their significance and evaluate the research hypothesis by comparing variables significance against the p-value to gauge their level of influence on the dependent variable. The study applied at a confidence level at 95 percent to test the dependent and the independent variables. Further for reliability analysis of questionnaire survey the Cronbach’s alpha test is used and the factors having value greater than 0.70 implies that the scale item has a high reliability. The experimental results are compared with the existing scheduling approaches on the basis of total execution time.