Simultaneous Update of Priority Structures

To avoid centralized bottlenecks, data structures for multiple computer systems must be simultaneously updatable and yet must have some coherent and consistent state. This can be achieved through judicious partitioning of system data structures and control of access to them in such a manner that (1) either different processes operate upon disjoint segments of a given structure; or (2) processes operate upon the same datum object, in a controlled manner. Partitioning of data structures can be by structure or content. It can also occur at multiple times. In this paper we demonstrate content partitioning of a k-ary tree data structure at runtime, to realize a simultaneously updatable priority queue. The tree algorithms are then generalized to banyan data structures and shown to possess attractive properties of simultaneous update and throughput.