10.07.2015 Views

DOWNLOAD MY Ph.D Thesis - UNAM

DOWNLOAD MY Ph.D Thesis - UNAM

DOWNLOAD MY Ph.D Thesis - UNAM

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Chapter 1Introduction pThese are the most important performance considerations according to [114]. Resultsfor such parameters are generally plotted either as a function of the offered load, whichis the actual load or traffic demand presented in the network, or as a function of thenumber of (active) stations transmitting traffic to the network. In addition, there areseveral other properties that can affect performance. These are:1. Channel capacity 4. Local network protocol2. Propagation delay 5. Offered load3. umber of bits per frame 6. umber of stationsThe first three factors listed above, can be seen as the parameters that characterise thenetwork and are generally treated as constants. The local network protocol is the focalpoint of the design effort and consists mainly of the medium access and physical layers.The physical layer is not likely to be much of a factor. Generally, at this layer datainformation is transmitted with little delay. The medium access layer however does havea significant effect on network performance and is discussed in length in this thesis. Thelast two factors are concerned with determining the performance as a function of offeredload or number of active stations.One factor that was not listed above is the error rate of the channel. With errorcorrection techniques used by communications protocols, such as CATV protocols [52],[22] and [34], link errors are not a significant factor in performance [114]. Therefore,the error rate of the channel will not be mentioned again.1.2.2 Modelling techniquesInitial techniques used to estimate the performance analysis of communicationsprotocols relied on mathematical models using stochastic processes based on probabilityand queuing theory [120]. The high complexity involved in the solutions for queuingnetworks led to the formalisation of approximation methods such as Mean ValueAnalysis (MVA) [89], [90], convolution [91] and linear programming [65] and [66].Such models have made several assumptions, examples of which are random rates ofpacket arrival and fixed number of stations. According to [71], the random arrival1-5

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!