Modeling the Gradual Degradation of Eventually-Consistent Distributed Graph Databases


Under the ‘eventual consistency’ approach to updates in a distributed graph database, it is possible that edge information may be corrupted. Errors may then be propagated to other parts of the database by subsequent queries. The process by which this occurs is modeled, with the aim of estimating the time that it takes for a clean database to become degraded to the point of being unusable. A fluid approximation is developed and two solution methods are proposed. The accuracy of those solutions is examined thoroughly, for databases with different sizes, structures and parameter settings, using simulations as a basis of comparison.

Queueing Models and Service Management