Increasing resiliency through priority scheduling of asynchronous data replication

Distributed systems commonly replicate data to enhance system dependability. In such systems, a logical update on a data item results in a physical update on a number of copies. The synchronization and communication required to keep the copies of replicated data consistent introduces a delay when operations are performed. In time-constrained systems or systems distributed over a bandwidth-constrained area, such operational delays generally prove unacceptable. Asynchronous data replication is commonly used to mitigate these delays. We look to develop a general solution for the introduction of an adaptive data replication scheduler to optimize asynchronous replications based on a user-developed priority model in overloaded situations. The solution uses a multi-layer perceptron neural network to mimic the behavior of a historically optimal scheduler through functional approximation with its evaluation through simulation.