Files in this item



application/pdfEnabling Theore ... g Large Scale Networks.pdf (3MB)
(no description provided)PDF


Title:Enabling Theoretical Model Based Techniques for Simulating Large Scale Networks
Author(s):Kim, Hwangnam
Abstract:Modern data communication networks are extremely complex and do not lend well to theoretical analysis. It is not unusual that network analysis can be rigorously made after leaving out several subtle details that cannot be easily captured in the analysis. As a result, packet mode, event driven simulation studies are usually resorted to better study the performance of network components, protocols, and their interaction. The major obstacle in packet mode simulation is, however, the vast number of packets that have to be simulated in order to produce accurate results, especially in large scale networks. What seems to be a reasonable solution is really to incorporate theoretical modeling into packet mode simulation. The notion of fluid model based simulation is recently proposed to alleviate the computational overhead in packet mode simulation. Conceptually, a fluid model is developed and incorporated into the simulation engine. In the course of simulation, a sequence of closely-spaced packets are abstracted into a fluid information, and the fluid model is used to determine its behavior. As the first theme, we investigate whether or not the fluid model based simulation is effective in simulating IEEE 802.11-operated wireless LANs, and to develop a fast simulation framework to expedite simulation, while not compromising the fidelity of simulation results. In spite of its effectiveness in terms of reducing the execution time, fluid model based simulation is not well-suited for studying the network behavior under light and/or sporadic traffic, as it is built upon the assumption of a large number of active flows in the network. To address the issue, we contrive network calculus based simulation as another main theme in the thesis. We firstly characterize how TCP congestion control interacts with AQM strategies with network calculus theory, and then determine a set of scheduling rules to regulate TCP traffic, and finally incorporate the rules into a network simulation engine to improve simulation performance. Although both fluid model based simulation and network calculus based simulation indeed give encouraging results, they cannot provide the packet level dynamics, such as the instantaneous queue length and packet dropping probability, due to the use of abstract simulation units, i.e., fluid rate and traffic amount. In order to address this issue, we propose hybrid simulation techniques, called mixed mode simulation, to discover packet level details of one packet mode, foreground flow, approximating all the other flows with theoretical model based, background flows. In the mixed mode framework, packet mode simulation co-exists with theoretical model based simulation within one simulation framework, and therefore analytical models of specifying the interactions should be devised. Lastly, we also contrive a new rescaling simulation methodology (RSM) to simulate large scale IP networks with TCP and/or UDP traffic. Even though mixed mode simulation can produce packet level details, there exist the cases that all the network behaviors, inclusive of all the flows and all the networking points, should be inspected. The underlying idea of RSM based simulation is to reduce the computational cost by scaling down the network to tractable one that can be simulated for a short time interval to produce sufficient results at packet mode, and then to extrapolate the results expected from the original network with the obtained those. In order to give a guideline for both down- and up-scaling or to explain how to preserve the network properties unique in the original network within the down-scaled network, a rescaling model is contrived and presented.
Issue Date:2004-12
Genre:Technical Report
Other Identifier(s):UIUCDCS-R-2004-2456
Rights Information:You are granted permission for the non-commercial reproduction, distribution, display, and performance of this technical report in any format, BUT this permission is only for a period of 45 (forty-five) days from the most recent time that you verified that this technical report is still available from the University of Illinois at Urbana-Champaign Computer Science Department under terms that include this permission. All other rights are reserved by the author(s).
Date Available in IDEALS:2009-04-17

This item appears in the following Collection(s)

Item Statistics