Article Preview
TopIntroduction
WSs became the main source for constructing software systems over the Internet. Because of the increasing number of published WSs, several WSs may share similar functionalities. However, if many functionally-equivalent WSs exist, it becomes a fastidious task to discover and thereafter select the appropriate one. In fact, this issue has gained a great research attention in literature, and motivated researchers to propose several mechanisms that will select the appropriate WS for the requester based on his needs.
To distinguish between similar WSs, it is mandatory to consider non functional parameters, i.e. Quality of Service (QoS), to find out the efficient WS (Chen, Zheng, Liu, Huang, & Sun, 2013). However, using QoS parameters raises other challenges such as how to specify, store and rank WSs?
A number of research efforts have studied this open issue by suggesting some solutions (Li et al., 2012; Wang et al., 2010; Zheng et al., 2011; Manikrao, & Prabhakar, 2005). However, most of these solutions did not address the problem seamlessly without extending WSs registry or WSs languages such as WSDL for instance, did not provide a WS ranking algorithm, and/or not integrate user feedback into WS selection process, etc.
Quality of Service (QoS) determines the service availability, usability, pertinence and utility, which influence the popularity of the service. This ascertainment motivates us to look, in this paper, at various WS QoS metrics and requirements, and how these metrics can may influence the published WSs consumption. As a result, we will propose a new flexible approach using an efficient algorithm based on WS Popularity Score (WSPS). This score consists of ranking WS, including simple, composite and semantic WSs, by using metrics such as usability, age, pertinence, etc. In addition we opt to assign weight to each metric as a filter to select the more adequate one since one metric may have high importance over the other according to user importance.
Recall that the popularity notion is reminiscent of the PageRank (Page, Brin, Motwani, & Winograd, 1999; Tyagi, & Sharma, 2012) popularity score for pages of the World Wide Web or, more generally, for nodes in a graph. Popularity has gained more attention in several domains such as web page search, Information Extraction, web mining, etc. In the case of web search domain, popularity focuses on external and internal link popularity (Pandia Search Central, 1998). Internal link popularity refers to the number of links to a website from web pages that belong to it. While external link popularity represents the number of links from outside sources that lead back to the particular website. Therefore, a website with high link popularity has what is called “link cardinality” or “link superiority”. The website then is considered an informative one, and has a high rank on web search engines.
Based on the best results attained in web popularity, and in order to propose an efficient approach for WS popularity, we should efficiently study and experiment this potential solution. More specifically, we must start by defining quality criteria or quality metrics representing parameters to compute such a score. Consequently our main contribution in this paper is to define fundamental QoS metrics related to WSs, elaborate a refined QoS WS taxonomy, and propose the required logical formula using these metrics to calculate WS popularity score of each discovered WS.
An interesting research effort presented in Maamar, Santos, Wives, Badr Faci, and de Oliveira (2011) consists of facilitating WS discovery and composition by integrating social networking at WS level. In their paper, the authors tried to model WS relations by acquiring inter-actions between services in a social network. This model provided a grouping WSs mechanism that establishes the degree of similarity between WSs. Consequently this will be helpful to organize and to extract sequences of anterior successful interactions for future needs of the user requester.