Lookup NU author(s): Professor Alexander Romanovsky
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
The possibility of effects like an unpredictable growth of resource and computation time losses and of expenditures on recovery (e.g. the domino effect) makes backward recovery schemes unpracticable for systems with high fault-tolerance requirements, although they have a number of advantages. The main purpose of this paper is to propose a formal model of the execution of concurrent programs, in which the domino effect can happen, and of the way this effect can be detected statically by analysing the programs. In particular, this can be done by reducing the domino effect problem to that of detecting a deadlock condition. This would allow the existing techniques for analysing the correctness of concurrent systems and detecting deadlocks statically to be used to analyse the properties of recovering systems and statically obtain additional information on the behaviour of concurrent systems in the event of a fault and rollback. By performing a static analysis of all system paths (which relies on looking through the entire reachability tree of process joint behaviours), several behavioural system properties concerning recovery can be checked prior to system use. Detecting bottle-necks at the system design stage will help to avoid unpredictable resource expenditures, and their elimination will help to ensure certain properties of system behaviour when recovering from faults.
Author(s): Romanovsky A
Publication type: Article
Publication status: Published
Journal: Computer Systems Science and Engineering
Print publication date: 01/05/2000
ISSN (print): 0267-6192
Publisher: CRL Publishing Ltd.