A Fault-Tolerant Distributed Legacy-based System and Its Evaluation

In this paper, we present a complete architecture for improving the dependability of complex COTS and legacy-based systems. For long-lived applications, such as most of those being constructed nowadays via integration of legacy subsystems, fault treatment is a very important part of the fault tolerance strategy. The paper advocates the need for careful diagnosis and damage assessment, and for precise and effective recovery actions, specifically tailored to the a®ecting fault and/or to the extent of the damage in the affected component. In our proposal, threshold-based mechanisms are exploited to trigger alternative actions. The design and implementation of the resulting solution is illustrated with respect to a case study. This consists of a distributed architectural framework, handling replicated legacy-based subsystems. Replication and voting are used for error detection and masking. An experimental prototype deployed over a COTS-based LAN is described and has allowed a dependability analysis, via combined use of direct measurements and analytical modeling.