Retour au site du LAAS-CNRS

Abstract

Defining clear Quality of Service agreements between service providers and consumers is particularly important for the successful deployment of service-oriented architectures. The related challenges include correctly elaborating and monitoring QoS-aware contracts (called SLA: Service Level Agreement) to detect and handle their violations. In this paper, first, we study and compare existing SLA-related models. To address the insufficiencies of these models, we propose a complete, generic and semantically richer ontology-based model of Service Level Agreements. In this model, we use the SWRL language (Semantic Web Rule Language) to express SLA obligations. This language facilitates the SLA monitoring process and the eventual action triggering in case of violations. In a second step, we use our SLA model to automatically generate semantic-enabled QoS obligations monitors. The main algorithms that perform the monitoring process are presented in this article. We implement these algorithms in an automatically generated service-oriented architecture. Finally, we believe that this work is a step ahead to the complete automation of SLA management process.

Mots-Clés / Keywords

Abstract

Technology advances continue to make computing environments ever changing and more complex. In the presence of such environments software systems are increasingly expected to continue operating at run-time. As human intervention becomes costly, time-consuming and error-prone, these systems should be equipped with self-adaptation capabilities in order to adapt themselves in response to environmental changes. While most of the research in this area focuses on individual parts of an adaptive system, our work leverages on this research but tackles the problem where interdependent and distributed adaptations are concurrently performed. In this paper, we approach behavioural changes of component-based systems in two stages. First, we propose a process to individually adapt one component at a time. Second, we elaborate a coordination protocol to maintain globally consistent state when implementing distributed adaptations. To achieve correct coordination, rather than only considering dependency relations between multiple adaptations, our approach further focuses on dependency relations between components at run-time. Motivated by the potential benefits of using formalisms, we construct a formal model of our protocol using Coloured Petri Nets in order for an adaptive system to be trusted after adaptation. In the model, we make sufficient abstraction of details, but still deal with the core of the protocol. This makes the model simpler and the analysis easier due to restricted state space size. We verify key behavioural properties and conduct CTL model checking to assess the correctness of the model and thereby the correctness of the protocol.

Mots-Clés / Keywords

Mots-Clés / Keywords

Abstract

Architectural adaptation is important for handling self- configuring properties of autonomic distributed systems. It can be achieved by model-based management of dynamic architectures. Describing dynamic architectures includes defining rules for reconfiguration management. Within this research context, several works have been conducted using formal specification to handle this complexity. Graph and graph rewriting-based approaches showed, through many studies, their appropriateness to tackle architectural adaptation problems. However, scalability of such approaches remains an open issue and has been rarely explored. In this paper, we investigate this issue. We introduce a graph-based general approach for handling of dynamic architectures, and we illustrate it within a scenario of collaboration support in Crisis Management Systems. We elaborate the formal models for dynamic architecture management. Using the French Grid GRID5000, we conducted an experimental study to assess the scalability of the elaborated models.

Mots-Clés / Keywords

Abstract

Failures during web service execution may depend on a wide variety of causes, such as network faults, server crashes, or application-related errors, such asunavailability of a requested web service, errors in the orchestration of choreography of applications, missing data or parameters in an execution °ow, or low Quality of Service (QoS). In this paper, we propose a healing architecture able to handle web service faults in a self-healing way, discussing infrastructural faults and web service and Web application faults. The self-healing architecture manages repair actions, such as substitution of a faulty service or duplication of overloaded services. Implemented prototypes involving QoS in coordinated web services are illustrated and discussed.

Mots-Clés / Keywords

Abstract

Maintaining the Quality of Service (QoS) is important for self-healing web service-based distributed interactive applications. It requires the ability to deal with permanently changing constraints both at the communication and the execution levels. Preventing or repairing QoS degradation also requires the capacity of identifying its possible or actual sources and the capacity of reconfiguration decision and enforcement. Dealing with these issues is especially challenging for web services since the self-healing solution has to preserve the dynamic composition property and to be seamless for the service requesters, while being always usable under the different deployment constraints. In this paper, we present a self-healing middleware framework able to provide the self-healing properties for QoS management in web service-based distributed interactive applications. The framework implementation has been achieved in the context of the WS-DIAMOND project. It covers the whole cycle of adaptation management including monitoring and analysis of QoS values, and substitution-based reconfiguration.