|
DSpace at ICSY >
ICSY - Integrated Communication Systems Lab >
04: Diploma Theses / Diplomarbeiten >
Please use this identifier to cite or link to this item:
http://hdl.handle.net/123456789/35
|
| Title: | An Approach to Intelligent Network Service Parameterization |
| Authors: | Frey, Jan |
| Keywords: | Network Service |
| Issue Date: | Mar-2003 |
| Publisher: | University of Kaiserslautern |
| Abstract: | The main goal of this thesis is to provide an automatic Network Service Parameterization mechanism. This mechanism intelligently selects the best Network Service available and sets Service Parameters appropriately with respect to a certain application. The parameterization manually parametrization of protocol stacks is unfavourable, since the decisions must be made at compile time of the application and cannot be changed without modifying the application. The solution of using a rule-based system configured by networking experts occurs to be more reasonable, but also fails in practice: many of the parameters mandatory are unknown or cannot be determined accurately. In contrast the method presented in this thesis does not depend on general knowledge. The basic idea is to learn relationships between Network Service parameters and resulting networking performance of the application observed. If such a learning process completed successfully the learned relationships can be used to make intelligent decisions. In order to assess the quality of a Network Service with respect to a specific networking application the service must actually be used. Of course neither real-world applications nor real-world networks can be used, as existing applications do not allow parameterization of their network utilization and the networks used should have absolutely stable characteristics in order to obtain reproducible results. Instead, a network emulation environment and a network performance measurement tool are implemented that allow realistic validation of the mechanism. |
| URI: | http://dspace.icsy.de/handle/123456789/35 |
| Appears in Collections: | 04: Diploma Theses / Diplomarbeiten
|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
|
