1 / 6

Discrete Event Simulation - 3

Discrete Event Simulation - 3. What other subsystems do we need to simulate? Although Packets are responsible for the largest amount of events, they are NOT the only important items in the system. Discrete Event Simulation - 3. What are these subsystems?

Download Presentation

Discrete Event Simulation - 3

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Discrete Event Simulation - 3 What other subsystems do we need to simulate? Although Packets are responsible for the largest amount of events, they are NOT the only important items in the system. June 10, 1999

  2. Discrete Event Simulation - 3 • What are these subsystems? Sources make up one class. We "solved" the problem up to this point by simply saying that the source "generates packets according to some probability distribution". A major problem is the determination of which probability distribution corresponds to which kind of traffic. There are several types of traffic that are of critical interest in networks: • Voice Traffic - this is a carry-over from the old telephone traffic and has been studied for about 100 years. • File transfer traffic - this has been around for 30+ years. • Video Traffic - this is also relatively recent and is probably the hardest to characterize. June 10, 1999

  3. Discrete Event Simulation - 3 Both Voice traffic and Video traffic have two subcategories: • Real-Time 2-way voice traffic. • 1-way voice traffic. • Real-Time 2-way video traffic. • 1-way video traffic. The first question would be: does it make any difference to a source whether the traffic it generates is part of a 2-way stream or just 1-way? Probably only in the quality-of-service it must negotiate with the network: 2-way traffic has stringent delay requirements, since it assumes an interaction between two interlocutors who wish to believe they are "face-to-face". 1-way traffic requires that the timing of the source be reproduced at the destination, but (what is reasonable?) delays can be tolerated. June 10, 1999

  4. Discrete Event Simulation - 3 The main problem is the generation of traffic with given statistical properties. This involves two distinct issues: 1) Finding out the statistical characteristics of "real" traffic. 2) Generating synthetic traffic with the desired characteristics. The first problem is the harder of the two, since "real" traffic will behave in ways that may not be accurately representable by any of the probability density functions we would like to use: Poisson arrivals represent some portions of voice traffic "reasonably well" - whatever that might mean - while they represent other portions (e.g., the silence periods) substantially less well. Furthermore, the parameters that can be obtained from one speaker (or from a set of speakers from the same linguistic group) may not provide a match for other speakers (especially if they are from different linguistic groups). June 10, 1999

  5. Discrete Event Simulation - 3 The second problem is (easily) solvable only if the first problem has been solved AND one of the "convenient" probability distributions can be used. A problem in this case is the need to keep computation to an acceptable level: one might have been able to obtain a probability distribution from an empirical study of "real" data, but generating synthetic data via the empirically derived distribution is too expensive. Another problem is that only relatively few probability distributions allow us to "predict" behavior via simple mathematical manipulations: thus the predictions would have to be obtained via extensive numerical (or simulation) activity, with further possibilities for introducing errors. June 10, 1999

  6. Discrete Event Simulation - 3 We need to develop techniques for deriving probability distributions from samples and for determining that the behaviors we are generating do indeed satisfy the desired probabilistic characteristics. How? June 10, 1999

More Related