الفهرس 
Chapter 1 IntroductionOnly 14 pages are availabe for public view
 |  | from | 158 |  |  |
| | | from | 158 | | |
Abstract
Recently, web service composition is a challenging research issue. The main reason for that is the increasing number of web services deployed over the cloud infrastructure. Many factors are taken into consideration while applying algorithms for web service composition. These factors include execution time and the optimality of the composition process to satisfy the needs of the user requirements. Many candidate services are available to participate into the composition problem. The process of choosing the best candidate service among many large number of candidate services to best participate in composition is a challenging task. The time required for the composition is another issue. MapReduce framework is used to handle large number of tasks by distributing task into multiple smaller tasks. In this thesis, we proposed a framework to address these challenges. The proposed framework consists of five modules. Firstly, Normalizer gives a certain range for all QoS attributes and historical user orders. Using Clustering Composer module the search space is reduced and a set of QoS prediction values for the next module is prepared. QoS predictor is responsible for predicting the QoS values that may be changed or missed because of the dynamic changes of the network. Service reduction reduces the available candidate services to improve the composition process. Service reduction is classified into Stable Service selector and Skyline Service sub-modules. Finally, the composition process is done and a list of candidate composite services is generated through Service composer and evaluator module. To prove the applicability of the proposed framework, two techniques are proposed. Multi-objective genetic based technique and hybrid bio-inspiration technique. Each of the two techniques implement some of the modules to be used in a certain situations. We introduce Modified HAPA (MHAPA) algorithm to implement predictor module. The presented algorithm (MHAPA) enhances the prediction of QoS values of services participating in the composite service. It modifies HAPA algorithm through using fuzzy clustering method to enhance the prediction process.The simulations and results proved the optimality and the less execution time needed for the two techniques. The effect of changing the number of candidate services and abstract services on both fitness value and execution time is a challenge faced.