Time-Dependent QoS Aware Best Service Combination Selection

Abstract

Service Oriented Architecture allows developing complex business applications from existing services. Given that many services are available with the same functionality and with different Quality of Service (QoS) attributes, one common challenge is to select the best service combination regarding user's requirements. Existing solutions often consider static QoS values for candidate services. Nevertheless, in real world applications, QoS values can change during time. In addition, besides structural constraints, several QoS and temporal constraints can also be specified at the business level. Considering time-dependent QoS values associated with business level constraints makes the selection process a very complex and time consuming decision problem given the large number of service combinations to be compared. To deal with this issue, in this paper, the authors propose a novel service selection approach based on QoS and temporal pruning techniques to reduce the number of candidate services. The proposed approach allows pruning uninteresting services based on a set of local thresholds. These latter are measured using constraint optimization models while dealing with general flow structures including sequential, parallel, choice and loop patterns and different types of QoS and temporal constraints. Experimental studies show the benefits of the proposed approach in particular in terms of computational time.

Heuristic Based Time-Aware Service Selection Approach

“…Hence, a near-tooptimal solution can be found more quickly. In our previous work [6], we demonstrated that the computation time of the pruning process is equal to 23 ms in the worst case. This is an expected result since the computation of the different QoS and temporal local thresholds can be applied in parallel.…”

“…1) Constraints Specification: A composed service is usually specified as an abstract business process. Each business process is composed by a set of activities or abstract tasks A = {A 1 , ..., A n } which depend on each other by a set of structural dependencies that can follow different patterns including sequential, parallel, choice and loop patterns [6]. In addition to structural constraints, business designers can also require supplementary intra and inter-task temporal constraints at the business level in order to increase their market share and profitability.…”

“…This is not trivial since some constraints may be overlapped or included in each others. To overcome these issues, in [6] and [7], we have proposed a set of constraint optimization models to compute local thresholds while dealing with complex business structures and several local and inter-task constraints. To avoid discarding any service that might be part of a feasible solution, local thresholds of each task are relaxed as much as possible.…”

“…To avoid discarding any service that might be part of a feasible solution, local thresholds of each task are relaxed as much as possible. For more details we refer reader to [6] and [7].…”

“…In this paper, we do not give any details about the aggregation function. More details about this function while considering several QoS categories and complex business structures can be found in our previous work [6].…”