Distributed Deadlock Detection for Distributed Process Networks

The distributed process network (DPN) model allows for greater scalability and performance over a non-distributed process network model. This paper presents a distributed deadlock detection algorithm applicable to process networks. The algorithm efficiently detects both local and global deadlocks. In addition, the deadlock detection algorithm has minimal bandwidth and memory requirements. Lastly, the implementation of a high-performance DPN framework that detects deadlocks is presented. Furthermore, this implementation supports the future addition of dynamic process creation/migration (ability to create/relocate processes at run-time).