Please use this identifier to cite or link to this item: https://ptsldigital.ukm.my/jspui/handle/123456789/513253
Title: Time and cost aware cloud computing service composition using imperialist competitive algorithm
Authors: Amin Jula (P63450)
Supervisor: Zalinda Othman, Prof. Madya Dr.
Keywords: Cloud computing
Competitive algorithm
Quality of service
Dissertations, Academic -- Malaysia
Issue Date: 10-Mar-2016
Description: The considerable growth in the number of provided services in cloud computing, as well as the supply and demand rules that have entered cloud services, has led to perpetual competition for satisfying customers in enhancing quality of service (QoS). The widespread nature of the search space, the abundance of possible choices for each simple service, and several QoS parameters have created the NP-hard optimization problem called cloud computing service composition (CCSC). Although evolutionary algorithms such as the Imperialist Competitive Algorithm (ICA) can significantly contribute to addressing the CCSC, they suffer two major challenges that severely limit their performance. First, in the case of facing a vast search space and large number of dimensions in the CCSC, the number of space points that are actually potential solutions increases to the extent that very few of can be investigated by the algorithm. Accordingly, an innovative, simple but high performance operator called the Color Revolution Operator (CRO) is proposed to reach more satisfactory solutions. The CRO is inspired by the color revolution of the former Soviet Union and the Balkans. While preserving the original structure of a solution, it tries to make ever-enhancing changes only in the weakest part of the solution. The experimental results of applying CRO only to the imperialists (ICACRO-I), and also to all countries (ICACRO-C), have demonstrated that an average performance improvement of 30% can be expected in the two approaches. Second, the abundant number of service providers, each of which present huge numbers of simple services, significantly decreases the quality of the first generated solutions of the algorithm. To make more informed decisions in generating first-generation solutions and avoiding blind selections during algorithm execution, PROCLUS is applied to categorize service providers based on the service-time parameter. PROCLUS, which is one of the best high-dimensional clustering methods, divides available service providers into three recommendation-level groups on the basis of the quality of their provided services, once for all requests. Subsequently, a probability rate will be calculated to select a service provider from each group. This rate is based on the service time average of each group. The categorization resulted in 2% efficiency enhancement for the Categorized Search Space Imperialist Competitive Algorithm (CSSICA). Extending this approach to consider more than one QoS parameter is also examined by independent categorization of service providers for service time and service cost. To reach this goal, aggregation of obtained independent groups is utilized to achieve a comprehensive categorization of service providers in Multi-parameter CSSICA (MCSSICA). Ultimately, as a final achievement, MCSSICACRO is proposed by combining ICACRO and MCSSICA. The obtained results have also shown higher efficiency enhancement than a single parameter investigation. Comparison of the results of ICACRO, CSSICA and MCSSICACRO with ICA, some comparative algorithms, and applied statistical tests demonstrate that the proposed methods not only achieve statistically significant different results but also could be considered an optimal solution that is reliable and scalable.,Certification of Master's/Doctoral Thesis" is not available
Pages: 197
Publisher: UKM, Bangi
Appears in Collections:Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Files in This Item:
File Description SizeFormat 
ukmvital_85401+SOURCE1+SOURCE1.0.PDF
  Restricted Access
639.08 kBAdobe PDFThumbnail
View/Open


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