Why is Flow Control Hard: Optimality, Fairness, Partial and Delayed Information

Flow control has been studied since the early stages of research in telecommunication networks. In these early studies the term ow control has been used to describe almost any type of network control. With the rapid growth of telecommunication networks control and management, ow control has evolved as a distinct class of algorithms and substantial research e ort has been devoted to its study. An extensive survey of the most representative of these e orts is presented in this paper. The motivation for this work is the need for of a thorough, deep understanding of the ow control problem in gigabit, multiclass networks. A broad class of ow control schemes, namely optimal ow control, is considered within a game theoretic framework. The class of adaptive ow control mechanisms is not presented extensively, but valuable insight about the performance of such schemes can be gained from the study of optimal ow control policies. A number of important e orts to de ne and study fairness in the ow control context are discussed. The last part of the survey is devoted to works that address the fundamental problem of propagation delays in ow control for very-high-speed, wide-area networks. Presented in part in the 2nd ORSA Conference on Telecommunications, Boca Raton, FL, March 1992 Submitted to IEEE INFOCOM'93