e., whose backoff
evolution is regulated by a Markov chain. It suffices to substitute Eq. (29) with the steady-state distribution of the
considered Markov Chain. A few proposals of backoff models with memory (for example, the slow CW decrease
approach considered in [25]) has been reported.
9 Since we are conditioning on the backoff stage i, we can envision the event of transmitting into a slot as the
recurrence of transmission events separated by the time spent while in backoff stage i, assumed independent
among transmission events. Hence, this computation can be interpreted as an application of the Long-Run
Renewal rate theorem (see, e.g., William Feller, An Introduction to Probability Theory and Its Applications, Vol.
II, Wiley, Cap. XI - pp. 368-380) and is shown to depend only on the average time spent while in the backoff
stage i, and not on its distribution. As a side comment, there are some proposals that draw backoff counters from
distributions different from the usual uniform one. As should be clear now, such a generalization does not appear
to have practical significance, as the performance depend only on the mean value, and thus there is no reason in
using backoff distributions more complex than the uniform one.
Pages:
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248