This paper is a very well-written extension of the Banker’s algorithm to a distributed operating system. The high cost of centralized decisions is avoided by imposing a hierarchy on the processors under consideration, and applying the Banker’s algorithm at each level of the hierarchy. If the hierarchy is a shallow tree (maximized fanout), the results approximate those for a centralized allocator, but the scheme tends to be much more conservative for a deep tree. For the readers who may still be unfamiliar with structural induction, which the authors use in a proof, the appropriate reference is Burstall [1].