Analysis of Random Access Protocol under Bursty Traffic

Abstract

Aloha-type random-access protocols have been employed as access control protocols in wireline and wireless, stationary and mobile, multiple-access communications networks. They are frequently employed by the control and signaling subsystem of demand-assigned multiple access protocols for regulating the sharing of a control channel. The latter is used for the transport of reservation packets requesting the allocation of channel resources to an active terminal. Such a random access channel is used, among others, by cellular wireless networks, by two-way CATV networks (such as those based on the DOCSIS protocol recommendation), and by demand-assigned satellite networks. The correct design and sizing of the random access operated control/signaling channel is a critical element in determining the performance of these networks. Excessive delays in the transport of signaling messages (induced by too many collisions and retransmissions) lead to unacceptable session connection set-up times. Consequently, in this paper, we investigate the performance behavior of a random-access protocol when loaded by bursty traffic processes. The latter exhibit long range dependence (LRD). The LRD traffic flows are modeled here as multiplicative multifractal processes. The random access protocol is modeled as an Aloha channel with blocking. Parameters are defined to characterize the protocol behavior. We demonstrate that the burstiness feature of the traffic processes, rather than their LRD character, is the essential element determining the performance behavior of the protocol. When the loading traffic process is not very bursty, we show that the performance of the random-access channel can be better than that exhibited under Poisson traffic loading; otherwise, performance degradation is noted. We demonstrate the impact of the selection of the protocol operational parameters in determining the effective performance behavior of the random-access protocol.