Skip to Main content Skip to Navigation

Architecting Resilient Computing Systems: a Component-Based Approach

Miruna Stoicescu 1
1 LAAS-TSF - Équipe Tolérance aux fautes et Sûreté de Fonctionnement informatique
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : Evolution during service life is mandatory, particularly for long-lived systems. Dependable systems, which continuously deliver trustworthy services, must evolve to accommodate changes e.g., new fault tolerance requirements or variations in available resources. The addition of this evolutionary dimension to dependability leads to the notion of resilient computing. Among the various aspects of resilience, we focus on adaptivity. Dependability relies on fault-tolerant computing at runtime, applications being augmented with fault tolerance mechanisms (FTMs). As such, on-line adaptation of FTMs is a key challenge towards resilience. In related work, on-line adaption of FTMs is most often performed in a preprogrammed manner or consists in tuning some parameters. Besides, FTMs are replaced monolithically. All the envisaged FTMs must be known at design time and deployed from the beginning. However, dynamics occurs along multiple dimensions and developing a system for the worst-case scenario is impossible. According to runtime observations, new FTMs can be developed off-line but integrated on-line. We denote this ability as agile adaption, as opposed to the preprogrammed one. In this thesis, we present an approach for developing flexible fault-tolerant systems in which FTMs can be adapted at runtime in an agile manner through fine-grained modifications for minimizing impact on the initial architecture. We first propose a classification of a set of existing FTMs based on criteria such as fault model, application characteristics and necessary resources. Next, we analyze these FTMs and extract a generic execution scheme which pinpoints the common parts and the variable features between them. Then, we demonstrate the use of state-of-the-art tools and concepts from the field of software engineering, such as component-based software engineering and reflective component-based middleware, for developing a library of fine-grained adaptive FTMs. We evaluate the agility of the approach and illustrate its usability throughout two examples of integration of the library: first, in a design-driven development process for applications in pervasive computing and, second, in a toolkit for developing applications for WSNs.
Document type :
Complete list of metadata

Cited literature [64 references]  Display  Hide  Download
Contributor : Arlette Evrard <>
Submitted on : Friday, July 4, 2014 - 3:27:01 PM
Last modification on : Thursday, June 10, 2021 - 3:05:42 AM
Long-term archiving on: : Saturday, October 4, 2014 - 12:41:28 PM


  • HAL Id : tel-01018621, version 1


Miruna Stoicescu. Architecting Resilient Computing Systems: a Component-Based Approach. Ubiquitous Computing. Institut National Polytechnique de Toulouse - INPT, 2013. English. ⟨tel-01018621⟩



Record views


Files downloads