A Stochastic Model for Message Assembly Buffering with a Comparison of Block Assignment Strategies

Abstract

A stochastic model is developed for the process of dynamic buffering for inbound messages in a computer communications system. For a specific characterization of message traffic, complementary viewpoints—one considering the process as a queueing system, the other considering it as a compound process—lead to the same probability generating function. Two buffer assignment schemes, both dynamic but differing in binding strategy, are compared in terms of optimal buffer size and total buffer pool requirements for a given overflow criterion. A simple model of optimal blocking, earlier proposed by Gaver and Lewis, is extended to cover both buffer strategies and also heterogeneous message sources. Finally, the derived characterization of total storage requirements is compared by numerical example with conservative and nonconservative asymptotic treatments of the process. © 1972, ACM. All rights reserved.