Article Preview
TopIntroduction
Service Oriented Architecture (SOA) paradigm allows the integration of several service components to develop complex business applications. In advanced service-oriented computing, complex applications with QoS requirements are usually specified as abstract business processes with global QoS constraints. The execution of these applications requires the selection of a set of services to implement abstract business tasks while fulfilling the QoS constraints. With the growing number of potential candidate services of each business task that offer the same functionality but differ in their QoS attributes (e.g., response time, availability), the selection of the best combination of services that satisfies business process constraints and end-to-end user’s requirements is a challenging task.
The selection of the best service combination with end-to-end QoS requirements has been widely treated in the literature (Hwang, Lim et al., 2008; Ma, Bastani et al., 2013; Ardagna and Pernici, 2007; Zeng, Benatallah et al., 2004; Yu, Zhang et al., 2007). Two strategies are generally adopted: global optimization strategy which consists in searching candidate services that optimize the overall QoS of the business process and fulfill global QoS constraints while considering all possible combinations of services (Zeng, Benatallah et al., 2004; Ardagna and Pernici, 2007) and local optimization strategy which aims to select the best service from each class of candidate services (i.e., candidate services for each task) independently from other classes. This strategy is especially used in distributed environments where a global QoS management is not required (Benatallah, Sheng et al., 2002). In recent years, some works proposed new service selection strategy adopting a top-down method. These approaches assume that global constraints can be considered as the aggregation of a set of local ones (Alrifai and Risse, 2009; Qi, Tan et al., 2010; Sun and Zhao, 2012; Mardukhi, Nematbakhsh et al., 2013). Based on these latter, local optimization is applied to select the best service for each abstract task such that all local constraints are fulfilled. Despite active research in the context of service selection for abstract business processes, some issues still remain unsettled so far.
First, most selection approaches assumes that services are always available and that QoS values cannot change over time. However, within different time periods, QoS attributes of candidate services can have different values (Chen, Yang et al., 2011; Wagner, Klein et al., 2012). For instance, the invocation of services during business hours can be more expensive than invoking them outside these hours. Thus, considering permanent availability of services and assuming static QoS values is very restrictive to effectively represent services and reflect the impact of time on the QoS attributes.
Second, in real world scenarios, several constraints can be specified at the business process level (e.g., structural, QoS and temporal constraints). Current selection approaches consider only structural constraints. Usually, temporal constraints are considered when modeling and verifying business processes (Lanz, Kolb et al., 2013; Cheikhrouhou, Kallel et al., 2014) and neglected during the selection of the best combination of services. For example, a partner of electronics manufacturing organization can require in its business process that the manufacturing of peripheral parts has to finish no later than 20 time units after the starting of the process. Moreover, some QoS constraints can also be specified in the business process (Ardagna & Pernici, 2007). Considering QoS and temporal constraints when selecting the best service combination is a vital task since the violation of one or more constraints may affect the successful execution of the business process.