Resource management for large systems: Concepts, algorithms, and an application

Certain resource-allocation problems encountered in large operational systems are characterized as the resource management (RM) problem. Known Lagarange multiplier techniques for decentralization of large problems are then extended to large "real-world" problems, where functions and sets may be ill behaved. Simple conditions are given for existence of suitable multipliers, and provably convergent solution algorithms are also presented. These results are based on novel sets of assumptions; an attempt is made to justify these as an alternative for analyzing large real-world systems. Application of these concepts and algorithms to a large spare-parts warehouse is described.