An improved voice—data integration protocol for fiber optic bus networks

Abstract

In a recent paper [15], a new voice-data integration scheme for high-speed fiber optic bus networks (LANs/MANs) was proposed. The method was based on the p i -persistent protocol [14] and a movable-boundary framing mechanism that employed speech detection, and the method could achieve a nearly-unity channel capacity independent of the bus length and data rate. The model in [15] provided the proper protocol parameters (data station transmission probabilities) p i as functions of both the voice subframe size and the data slot position. While this scheme could be considered “optimal” in the sense that is resulted in fair sharing of each data slot by all data stations, it also resulted in enormous storage and/or computational complexity for the p i at each data station. Hence, the goal of our present work is to simplify the previous protocol and its associated complexity by studying a modified, but efficient and possibly “sub-optimal” scheme which specifies a simpler assignment of the p i (so that the protocol can be easily implemented). Now, a data station i operates with a single parameter p i whose new value is independent of the voice subframe size and the data slot position; also, these p i are chosen to approximately provide fairness of access among the data stations over the entire frame duration. Additional practical issues addressed in this paper include: (1) removing the restriction in [15] of having the number of voice stations upper bounded by the number of slots per frame (so that, even in the absence of data load, the channel utilization—due to voice loading only—can approach 100 percent, instead of the 40-percent limit in [15]), and (2) allowing voice packet buffering at the voice stations. For storing voice packets, we examine preemptive buffering [3], which is very suitable for speech and other real-time applications.