WSDM-Enabled Autonomic Augmentation of Classical Multi-version Software Fault-Tolerance Mechanisms

Abstract

Web services are increasingly deployed in many enterprise applications. For this type of applications, dependability issues are usually resolved by introducing some form of redundancy in the system. Whereas hardware redundancy schemes have traditionally been defined through static configurations based on worst-case analysis, the enhanced flexibility and interoperability of web services allows for dynamic (self-) management of redundancy at the application layer. Combining this advantage with service-oriented platforms such as OSGi facilitates the replication of software components and their integration within redundancy schemes. The application of such redundancy schemes inevitably comes at a price though — primarily due to the allocation of additional system resources. It is often unknown to the service provider how much redundancy and management complexity is required. Furthermore, the degree of redundancy and the dependability strategy to be employed may be restricted by the budget and requirements of the client, both of which may vary. In this paper, we propose a solution to allow the client to express a trade-off between its dependability requirements and its available budget at request level. A dedicated service provider will then attempt to honour these objectives — failing to do so would obviously result in failure from the client point of view. Furthermore, we show how classical multi-version software fault-tolerance techniques can be augmented with advanced redundancy management leveraging the Web Services Distributed Management standard.