الفهرس 
chapter 1Only 14 pages are availabe for public view
 |  | from | 206 |  |  |
| | | from | 206 | | |
Abstract
Abstract
The Cloud Computing vision is how to implement the large number of parallel computations on large scale distributed resources. The Cloud Computing is more involved in purchasing and consuming manners between providers and customers than other computing paradigms as the Grid and the Cluster. The core of providing the required services in the Cloud is the on-demand manner, where the customer usually asks for a particular service (e.g. Computation, Storage, and/or Memory) from the Cloud paradigm. The cloud provider has to provide the service with fulfilling its Quality of Service (QoS) requirements.
Cloud Computing delivers the user’s required services through what is known as the infrastructure virtualization capabilities. This virtualization layer has the role of providing the service to the user with hiding any other abstracting layers. The process of how the available resources are scheduled and allocated to the required services is called “Resource Provisioning”. Prediction of the suitable allocation resource and consummation of the service providing with fulfilling the customer’s QoS requirements, and also with satisfying the service’s provider, is considered the cornerstone in the resource provisioning process. Accordingly, identifying the convenient resource from the available pool of resources is considered the goal of any resource provisioning algorithm. The provisioning algorithm has to find the way of spreading the applications load on the proper cloud resources to achieve the optimization objective of satisfying both the customer and the provider. Achieving the Quality of Service (QoS) requirements usually depends on two essential factors which are; the Response time (i.e. many required services have a predefined deadline time that makes the responding after this time unavailing), and the Actual paid cost (i.e. some users need their requests to be fulfilled within a defined budget). On the other hand, good resources utilization is one of the main issues that must be considered in the provisioning process. The term “Good Resource Utilization” or “Good Load Balancing” means maximizing the platform utilization
without wasting any resources quantity as much as possible and without any overload in the resource nodes. This good utilization gives the service provider the chance to gain profit after submitting the demanded service and handling any fault occurs in the current provisioning process.
In addition to the above requirements, there are some external influences affecting the performance of provisioning algorithm, such as data communication parameters. Sometimes, the local resources of the data center are not adequate to satisfy the users’ requirements, so the providers need to spread the users’ applications on several data centers at different geographical areas around the world to satisfy the QoS requirements. Accordingly, by considering the expansion of the resources and applications, the transmission cost and time have to be concerned as significant factors in the provisioning process.
There are various directions of researches to solve the resource provisioning challenges. Some researchers try to fulfill the QoS requirements by minimizing makespan (execution time) as Iqbal; W. in [113], and Chen; Y. in [114], or minimizing resource utilization cost with neglecting the makespan as Byun; E. in [108], and Chaisiri; S. in [100], or try to optimize both of the cost and the time as Raj; G., and Nischal; A., in [112]. Other researchers care also about the provider revenue and how to increase his profit by balancing the computation load as Zhang; Z, and Suganya; S., in [119].
According to the work in this thesis, the Resource Provisioning Optimization Algorithm (RPOA) has been proposed. The RPOA is based on Particle Swarm Optimization (PSO) algorithm in order to find the near optimal solution for the resource provisioning problem. The optimization function role in RPOA achieves the availability by allocating the required service at the physical resource within the predefined deadline time and user budget. Also, the RPOA algorithm concerns with satisfying the service’s provider by trying to maximize the load on the utilized resources and freeing up the light loaded resources. This helps in fault tolerance, solving the problem of resources wastage as much as possible, and providing the free resources with much full capacity to be able to serve other coming services.
The RPOA algorithm efficiency in users’ satisfaction is measured by determining how many users’ requests are successfully completed within the predefined deadline time and budget. RPOA algorithm is approximately efficient in cost optimization with 83%. According to compare the deadline time with respect to both the maximum predicted responding time and the average of this responding time, it is found that, the efficiency rates are approximately 81%, and 96% respectively. Also, from the viewpoint of the service provider, the RPOA algorithm achieves the provisioning process with utilization efficiency approximately 84% of the utilized capacity with freeing up the light loaded resources to make the provider be able to handle any failure that can occur during the provisioning process, and also be able to receive other workloads need to be allocated.
The thesis is organized in six chapters as follows:
Chapter 1: gives an introduction to the distributed systems and their types as distributed computing, cluster computing, grid computing, and cloud computing, with handling the environment, characteristics, benefits, and challenges of every computing system.
Chapter 2: represents a survey on some soft computing algorithms, especially evolutionary algorithms like Genetic, Swarm and Ant Colony algorithms. Every algorithm is displayed with its characteristics, paradigm, flow chart and pseudo code. Also this chapter handles the optimization problem and how it can be solved by the displayed algorithms.
Chapter 3: discusses the resource management processes in the Cloud computing systems, and focuses on the resource provisioning and allocating problems as they are considered the main research issue of this thesis. Also, this chapter displays the different types of provisioning policies, and the related researches which handled the resource provisioning problem with their advantages and shortcomings.
Chapter 4: illustrates the problem definition which has be considered in this thesis, and states the supposed Resource Provisioning Optimization Algorithm (RPOA) with its environment and pseudo code. Also, it handles the case study implementation by using the CloudSim simulator.
Chapter 5: displays the evaluation of applying the proposed RPOA algorithm and other provisioning algorithms like Service Proximity Based Routing (SPBR), and Best Response Time Provisioning (BRTP). It represents the comparative study between the efficiency of the RPOA algorithm and the other algorithms.
Chapter 6: states a conclusion of the work in the thesis, and the future work that can be suggested depending on this thesis.