Static vs. dynamic recovery models for survivable distributed systems

Abstract

The need for survivability is more pressing for mission-critical systems, especially when they are integrated with other COTS products or services. As information systems became more complex and the interdependence of these systems became higher, the survivability picture became more complicated. In this paper, we introduce two basic models with respect to static and dynamic models, where each model has restart and continue modes. We describe each approach and compare the trade-offs in terms of simplicity, resource efficiency, adaptation, service downtime, immunization, and robustness. Later, we introduce a hybrid model by combining these two basic models. Our approaches are transparent to the client side and can be applied to existing distributed systems by extending the server side with the components necessary for the schemes.