Analyzing reconfigurable algorithms for managing replicated data with strict consistency requirements: a case study

We address reconfigurable algorithms for managing replicated data with strict consistency requirements, that is, whenever the user performs an update operation, the update is applied to all reachable copies as part of the update protocol. A key issue of designing such algorithms is to determine how often one should detect and react to failure conditions so that reorganization operations can be performed at the appropriate time to improve the availability of replicated data. We use dynamic voting as a case study to illustrate how often such failure detection and reconfiguration activities should be performed so as to maximize the data availability. We show that there exists an optimal period at which the failure detection and reconfiguration activities should be performed to optimize the system availability. Stochastic Petri nets (SPNs) are used as a tool to facilitate our analysis