Distributed deadlock resolution in store-and-forward networks

We present a simple distributed algorithm that resolves store-and-forward deadlocks in data communication networks. The basic idea of the algorithm is to detect cycles of nodes that may cause store-and-forward deadlocks, and to rotate packets along these cycles. The algorithm uses a fixed amount of storage in each node for its execution, and, under reasonable assumptions upon the routing and packet handling, it ensures that packets that enter the network arrive at their destinations in finite time.