Resolving Classical Concurrency Problems Using Outlier Detection

Abstract

In this paper outlier detection is used to determine anomaly between tasks to prevent occurrence of resource conflicts in prepared schedule. Determined conflictless schedule bases on controlling access of tasks to groups of shared resources. Proposed approach allows to prepare conflictless schedule of efficient parallel task processing without resource conflicts and is dedicated to environments of task processing with high contention of shared resources. In this paper the outlier detection is used to resolve two classical concurrency problems: readers and writers and dining philosophers. In opposition to other known solutions of concurrency problems, proposed approach can be applied to solve different problems and do not require to use additional mechanisms of task synchronization. The universality of proposed approach allows to prepare conflictless schedule even in environments, where classical concurrency problems will be significantly expanded and complicated.