Font Size: a A A

Bernoulli 'input Given Long Service Priority Queuing System Behavior Analysis And Optimize The Design Of High-end Routers,

Posted on:2004-11-18Degree:MasterType:Thesis
Country:ChinaCandidate:W C HeFull Text:PDF
GTID:2190360092492678Subject:Applied Mathematics
Abstract/Summary:PDF Full Text Request
In the optimization design of advanced router, since it exists HOL Blocking in IQ program, and when the number of ports is large or the line rate is very high, the present memory cannot have so fast frequency in OQ program, so CIOQ program is adopted now. Because of the complexity of this program, we only consider the optimization design about Output Buffer in the program, so as to obtain performance analysis of Output Buffer. Besides, each operation actually need time in the fulfillment of hardware of any algorithms in computer and communication fields. If the decision time is very small relative to service time, it may be ignored; otherwise it must be considered. Two models that need considering decision time and one model that doesn't need considering decision time are constructed in this thesis, and the background of the first two models is originated from communication field.In the first charpter of the thesis, the performance analysis is presented for the queuing system that has foresight and priority about the input of two queues and the service of one server. In the first section, the service rule is detailedly presented. It stipulates that the service of the server is non-preemptive, and beforehand decides which queue will be served the next time, so as to try to avoid idleness of the server. In the second section, the discrete model is constructed that is based on the rule given in the first section. The analytic solutions of the mean lengths of the two queues are obtained by introducing the generating function.In the second charpter of the thesis, the performance analysis is presented for the queuing system that has foresight and priority about the input of three queues and the service of one server. In the first section, the service rule is detailedly presented that is similar to the one given in the first charpter. In the second section, the discrete model is constructed that is based on the rule given in the first section. The analytic solutions of the mean lengths of the three queues are obtained by introducing the generating function. By testing several groups of data, It is well known that the number of the customers in the Output Buffer is not too much even if the input probability of the three queues is very large. So we must try our best to design the Input Buffer when we optimize CIOQ program. In addition, so long as any successive queues except for the first queue are considered as one queue, we may extend to the model of n queues from the models of two or three queues.In the third charpter of the thesis, the performance analysis is presented for the queuing system that has priority but hasn't considered decision time about the input of three queues and the service of one server. In the first section, the service rule is detailedly presented. It stipulates that the service is non-preemptive, and the service has not decision time, that is to say that the server is not just idle only if the queues have customers, moreover the server serves these queues according to the priority. It is obvious that this model is a ideal model. In the second section, the discrete model is constructed that is based on the rule given in the first section. The analytic solutions of the mean lengths of the three queues are obtained byintroducing the generating function. We compare the model that has considered decision time with the model that hasn't considered decision time by testing several groups of data. From the comparison, it is well known that the respective mean lengths of the three queues in the former model are approximate the mean lengths of the corresponding queues in the rear model because the fortsight is considered in the former model. So the two models in the first two charpter have very important real significance.
Keywords/Search Tags:priority queueing system, decision delay, non-absolute priority, predictive decision, generating function, markov process.
PDF Full Text Request
Related items