A dynamic bandwidth-allocation-based priority mechanism for the p/sub i/-persistent protocol for MAN's

Miller, G.J.; Paterakis, M.

doi:10.1109/49.245910

Cited by 3 publications

(1 citation statement)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are performance analysis studies of packet-switched slotted-ring networks, such as [4,5] for Fiber Distributed Data Interface, [6] for PLAYTHROUGH networks, [7,8] for Distributed Queue Dual Bus, and [9,10] for probabilistic p i -persistent networks. These studies usually model a ring node as an M/G/1 vacation system with Bernoulli schedule [6,9,10] or with time-limited service discipline [4,5], and obtain approximate values for the mean access delay and throughput at each node. Since the interdependence between nodes makes exact analysis intractable, most studies use the assumption of independent arrival of empty slots at a tagged node.…”

Section: Introductionmentioning

confidence: 99%

A Conditional Probability Approach to Performance Analysis of Optical Unslotted Bus-Based Networks

Brandwajn¹,

Nguyễn²,

Atmaca³

2009

Current Research Progress of Optical Networks

View full text Add to dashboard Cite

This paper presents a novel approach to the performance analysis of optical packet switching bus-based networks with unslotted Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) protocol. Because of the interdependence among bus nodes, an accurate performance analysis of such networks has been an open question for a long time. We model the bus as a multiple-priority M/G/1 queuing system with preemptive-repeat-identical (PRI) service discipline. To solve this model, we use a recurrent level-by-level analysis technique where the interference from higher levels (upstream nodes) is taken into account in terms of reappearance and disappearance rates for the server. The key features of this method are as follows. First, it specifically accounts for the distribution of voids seen by a node (via the number of attempts before a successful transmission) as a function of the node's position and offered network load. Second, it approximately computes the queue-length distribution at each node, and thus provides a tool for buffer size selection so as to meet a loss rate criterion. A comparison of our approximate model solution with network simulations indicates that our model generally offers good accuracy in assessing the performance of the network, including in terms of the queue-length distribution. Occasionally, the results of our model may deviate from simulation results. A discussion of the causes of such deviations, most likely to occur when nodes are close to saturation, offers additional insight into the properties of the bus-based network and its approximate solution.

show abstract