الفهرس 
AcknowledgmentOnly 14 pages are availabe for public view
 |  | from | 115 |  |  |
| | | from | 115 | | |
Abstract
Cloud Computing is getting popular by offering everything as a service. Cloud Services Users (CSU) face challenges in finding the needed services effectively. One of the critical limitations of a newly CSU is the lack of available efficient search engines. Cloud Services Discovery Framework (CSDF) is presented to highlight the discovering and utilizing of Cloud Services (CS).CSDF combines the functionality of software agents and Web Services. In CSDF, Software Agents are Crawler, Analyzer, Evaluator, Tracker, Recommender, and Ranker agents. CSDF Web Services are Cloud Service Manager (CSM), Query Manager (QM), Query Processor Manager (QPM), and User Profile Manager (UPM). The most effective vital part in CSDF is to understand the functionality of CS. The case that CS contain a description by Web Service Description Language (WSDL). Thesis approach is to study a special part of WSDL which is port type. Port type is extracted since it contains the main operations performed by Web Services. Operations provide main input and output data that will help in recommending such effective CS to CSU. While CS have no standard way in understanding the main functionality provided to CSU. Thesis illustrates tree taxonomy for CS in order to detect the CS’s main characteristics. Regarding the expansion of Cloud Computing, CS are categorized according to three layers IaaS, PaaS, and SaaS. Each layer has more sub layers. Based on the previous idea, thesis proposed a tree taxonomy that will be parsed into a specific path. The path of the Cloud Service regarded as the description of the CS. Cloud Service’s description path will be effective in recommending appropriate CS to CSU. Recommender provides a list of recommended CS showing the degree of similarity between user query and the recommended Services. Ranking methods uses Tree Structure Taxonomy Reasoning System (TSTRS), performance measures, and user rates. Reasoning algorithm detects the similarity between user query and CS. By applying the similarity reasoning and equivalent reasoning, TSTRS will be able to recommend a list of appropriate CS with similarity measures. Ranker agent reorders the recommended list according to TSTRA. In addition to reorder the list with respect to performance measures and user rate. Evaluation of CSDF used to estimate the performance of system. The resources of CS have been collected and classified to be experimented. The reasoning is applied on the classified data. Prediction Accuracy evaluation used to predict the resources preferred by users with two broad classes of prediction which are measuring Ratings Prediction Accuracy and measuring the accuracy of usage predictions. By Measuring Ratings Prediction Accuracy RMSE, MAE, and NRMSE, the average mean of NRMSE is 0.27836. Next step is to measure usage prediction, results will be 0.87 for precision, 0.65 for True Positive Rate mean, and 0.25 for False Positive Rate mean.