On Efficient Distributed Deadlock Avoidance for Distributed Recursive Processes

César Sánchez, Henny Sipma, and Zohar Manna

We study deadlock avoidance for resource allocation in distributed systems. While a general solution of distributed deadlock avoidance is considered impractical, we propose an efficient solution of the important particular case where the possible sequences of remote calls, modeled as call graphs, are known a-priori. The algorithm presented here generalizes to recursive processes our earlier work on deadlock avoidance for non-recursive call-graphs.

An essential part of this generalization is that when a new process is created, it announces the \emph{recursive depth} corresponding to the number of times a recursive call can be performed in any possible sequence of remote invocations.

Download: BIB PDF