PGAQK: An Adaptive QoS-aware Web Service Composition Approach

Abstract

Web Service Composition (WSC) has attracted considerable attention and research to support Service Oriented Architecture (SOA). WSC constructs complex applications by combing atomic Web services together to achieve users' requirements. The selection of the best Web service that fulfils the Functional Requirements (FRs) and optimizes the Quality of Service (QoS) requirements, such as response time, cost, reliability, etc., is a critical part of WSC, especially in a dynamic environment. Since requirements frequently change, this demands that WSC must be adapted to provide the most suitable composite services and fulfil the users' requirements that emerge. This paper presents a new hybrid approach for dynamic optimization of WSC using Parallel Genetic Algorithm (PGA) based on Q-learning, which we integrate with K-means clustering. We call it PGAQK. Q-learning is utilized to generate an initial population to enhance the effectiveness of PGA. K-means clustering is used to cluster Web services based on a fitness function. It is applied in the mutation operator, and used for Web service substitution to prune the Web services in the search space to find the best Web services for the environment changes that might occur at runtime. To the best of our knowledge such a hybrid approach has not been used for WSC before. We implemented PGAQK over the .NET Framework using C# programming language. A series of comparable experiments were carried out showed that PGAQK outperforms traditional PGA and Q-learning approaches in terms of fitness values. PGAQK allows WSC to modify itself dynamically and achieve a better fitness value to fit the changing environment, where the characteristics of composite web services continue to change.

Automatic Web Service Composition for SaaS Business Intelligence

“…Several workflows can be combined to more complex workflows. Web service composition can be done using existing tools such as Business Process Execution Language (BPEL) Designer Project [3], Business Process Model and Notation (BPMN) [4], and several other methods that have been proposed by previous authors [5][6][7][8][9][10][11][12][13][14][15][16][17].…”

“…For example, the user requests the TotalSales and ProductName outputs, but at the stage of searching Ultimately, these methods cannot find the ProductName output, because they cannot utilize the ProductId keyword to search for the ProductName output. D. Elsayed, E. Nasr, A. El Din Ghazali, and M. Gheith also proposed a Web service composition method they called PGAQK [10]. This method uses a Parallel Genetic Algorithm (PGA) based on Qlearning, which they integrate with K-means clustering.…”