Guidara, Ikbel (2016) Efficient Time and QoS-Aware Selection for Service Composition. École doctorale Mathématiques, Informatique et Télécommunications (Toulouse).

Preview

Text Download (4MB) | Preview

Abstract

Service-Oriented Computing (SOC) paradigm has emerged in last years to support cooperation between loosely coupled services to build complex applications. It involves the description, discovery, selection, and composition of services to support rapid development of complex applications. Usually, theses applications can be specified as abstract business processes and the goal is to select a service for implementing each abstract task. In addition to the functional requirements that must be accomplished, the QoS (Quality of Service) parameters are of paramount importance. Due to the large number of candidate services with same functionalities but offering different QoS values, the selection of the most suitable services for implementing abstract tasks while fulfilling QoS in a timely manner is not trivial. Moreover, in real-world applications, services can have different dependencies between them (i.e., structural and temporal). Considering these dependencies, the selection problem becomes more complex. Additionally, services usually operate in highly uncertain and dynamic environments, which can cause erroneous behaviors during the execution. In this context, it is crucial to tackle the selection problem while considering functional requirements associated with QoS and temporal constraints at design and run time.In this thesis, we contribute towards addressing the aforementioned challenges. Specifically, the main contributions of this thesis are as follows: (1) We propose pre-processing techniques to allow a scalable service selection without affecting the optimality of the selected solution. (2) We develop an efficient QoS-aware service selection approach that allows selecting the suitable service composition while fulfilling QoS and temporal constraints. The proposed approach can handle complex service selection problems while considering the aforementioned dependencies between services. (3) We propose a heuristic service selection approach to select a close-to-optimal solution based on clustering and constraints decomposition techniques. (4) To deal with dynamic and uncertain environments, we propose a proactive service selection approach for enforcing service composition adaptation at run time. The aim is to take early re-selection actions in order to reduce the possibility of execution interruption and increase the likelihood of finding a feasible solution. This approach deals with QoS fluctuations and changes in execution environments during execution (e.g., the availability of a new better service). The different contributions of the proposed approach are implemented and their efficiency is demonstrated and validated analytically and empirically through experimental results.