Extending Virtual Synchrony with Persistency

The virtually synchronous execution model provides an appropriate support for developing distributed applications, assuming a crash failure model. Synchronization points are only set when a view change arises, guaranteeing an efficient execution of such reliable applications. Its programming model is similar to that of a centralized application, but not identical. However, a crash model is not always appropriate for all applications. Indeed, those needing a long recovery time either due to their large state (like replicated databases) or because they use a slow network (like collaborative applications being executed on PDAs/smartphones) might need a recoverable model. In such cases, virtual synchrony could be extended for supporting simple recovery protocols. Persistent logical synchrony is one of such variations, that adds other synchronization actions when a view change arises, thus allowing partial recovery when the primary component membership is being assumed.