Retour au site du LAAS-CNRS

Mots-Clés / Keywords

Abstract

Ubiquitous communicating systems have particular characteristics, such as their dynamic nature and the great number of users and heterogeneous devices involved. In our work, we focus on collaborative activities. For such activities, users are organized into groups and communicate in order to achieve a common goal. Therefore, the problem of designing and implementing collaborative applications on top of ubiquitous communicating systems is a complex task that requires adequate modeling. In our view, addressing this problem needs a model-driven approach in order to ensure the coherence and correctness of the built systems. In this paper, we propose a multi-level modeling approach for collaborative ubiquitous systems. Moreover, relevant abstraction levels are identified and means are provided for inter-level model transformation (refinement) and for adaptation to context changes (selection). This adaptation is guided by both high level requirements and low level constraints.

Mots-Clés / Keywords

Abstract

Web service composition allows a fast and modular creation of applications by orchestrating several Web services. Such applications are frequently faced to performance and availability problems which may affect the partner Web services or the orchestration process itself. This requires mechanisms for adapting the architecture and the behaviour to this variable context. In this paper, we deal with strong mobility of orchestration processes as a mechanism for adaptation. We provide a solution that relies on checkpoint/rollback mechanisms. It is also based on source code transformation of the orchestration process. We apply our approach on WS-BPEL based orchestration processes. Hence, we establish a set of rules which transform WS-BEPL processes to equivalent mobile ones. When an adaptation is to be performed, the execution of some or all instances of a mobile process will be interrupted, and then they will be migrated to another node. After migration, the interrupted instances will resume starting from the last checkpoint. Experimentation results show the the efficiency of our approach and the low overhead it introduces.

Abstract

Ubiquitous computing environments provide a wide range of new perspectives in Computer Science. Among these, the dynamic collaborative activities carried out by groups of users present particularly complex issues. The highly dynamic nature of such activities, along with the heterogeneity of devices and communication systems, requires adaptive software solutions. In this paper, we propose a multi-layer modeling of system architectures in order to address this problem. This is necessary for taking into account both high level requirements and low level constraints. Moreover, this approach allows adapting the system architecture to handle context changes. We consistently handle inter-layer adaptation actions by considering a lower layer architecture as a refinement of an upper layer architecture. First, modeling layers for collaborative ubiquitous systems are provided. Then, we detail an implementation using OWL and graphs for models and rule oriented techniques, such as SWRL and graph grammar productions, for refinement. Finally, we consider a scenario on board of urban bus fleets as a case study from the more general group collaborative systems to which our models and procedures apply.

Mots-Clés / Keywords