Font Size: a A A

Approaching reality with a finite network node model analysis of DQRAP (Distributed Queueing Random Access Protocol)

Posted on:1994-06-22Degree:Ph.DType:Dissertation
University:Illinois Institute of TechnologyCandidate:McPheters, Michael JohnFull Text:PDF
GTID:1478390014492492Subject:Computer Science
Abstract/Summary:
The Distributed Queueing Random Access Protocol (DQRAP) is a revolutionary approach to managing channel access which could bring about a unified method of implementing communications. DQRAP is best described as a random multiple access protocol that operates in a slotted channel with minislots and ternary feedback. The system employs multiple distributed queues, two at each station, a request queue and a transmission queue, in order for stations to manage access to the channel and the transmission of data, respectively. It achieves the best performance of all known protocols in this class, approaching the performance of an idealized protocol. However, all known research to date of DQRAP has used an infinite node network model. We extended this work by examining DQRAP in a finite node network, i.e., a more realistic network employing DQRAP. We focus on a set of critical issues for modelling and analyzing the performance of a DQRAP network.; The model is also extensible to other techniques related to the implementation of DQRAP in MANs and WANs. For example, a priority mechanism is desirable in order to control the delay for high priority traffic, e.g., time dependent voice data. And a reservation mechanism is important for applications that may require the transmission of packets in sequential slots. Both mechanisms were studied as a part of this research.; Performance of the distributed queueing system was analyzed under various load conditions. The throughput was found to approach one while the delay remained bounded for input rates less than one. The results of simulating a 10Mbps DQLAN supporting a mix of high and normal priority traffic showed that the lowest average delay occurred for a mix of 30% high priority and 70% normal priority traffic. We found using the reservation mechanism that the average frame delay is less than making requests for each packet separately and sending them in non-consecutive slots.
Keywords/Search Tags:DQRAP, Access protocol, Distributed queueing, Random, Network, Node, Model, Delay
Related items