The research presented in this thesis aims to demonstrate that fuzzy logic is a useful tool for developing mechanisms for controlling traffc flow in ATM based multimedia networks to maintain quality of service (QoS) requirements and maximize resource utilization. The study first proposes a hierarchical, multilevel control
structure for ATM networks to exploit the reported strengths of fuzzy logic at various
control levels. Then, an extensive development and evaluation is presented for a subset of the proposed control architecture at the congestion control level.
An ATM based multimedia network must have quite sophisticated traffc control capabilities to effectively handle the requirements of a dynamically varying mixture of voice, video and data services while meeting the required levels of performance.
Feedback control techniques have an essential role for the effective and efficient management of the resources of ATM networks. However, development of conventional feedback control techniques relies on the availability of analytical system models. The characteristics of ATM networks and the complexity of service
requirements cause the analytical modeling to be very difficult, if not impossible.
The lack of realistic dynamic explicit models leads to substantial problems in developing
control solutions for B-ISDN networks. This limits the ability of conventional techniques to directly address the control objectives for ATM networks. In the literature, several connection admission and congestion control methods for B-ISDN networks have been reported, and these have achieved mixed success. Usually they either assume heavily simplified models, or they are too complicated to implement, mainly derived using probabilistic (steady-state) models.
Fuzzy logic controllers, on the other hand, have been applied successfully to the task of controlling systems for which analytical models are not easily obtainable.
Fuzzy logic control is a knowledge-based control strategy that can be utilized when an explicit model of a system is not available or, the model itself, if available, is highly complex and nonlinear. In this case, the problem of control system design is based on qualitative and/or empirically acquired knowledge regarding the operation of the system. Representation of qualitative or empirically acquired knowledge in a fuzzy logic controller is achieved by linguistic expressions in the form of fuzzy relational equations. By using fuzzy relational equations, classifications related to system parameters can be derived without explicit description.
The thesis presents a new predictive congestion control scheme, Fuzzy Explicit Rate Marking (FERM), which aims to avoid congestion, and by doing so minimize the cell losses, attain high server utilization, and maintain the fair use of links.
The performance of the FERM scheme is extremely competitive with that of control schemes developed using traditional methods over a considerable period of time.
The results of the study demonstrate that fuzzy logic control is a highly effective design tool for this type of problems, relative to the traditional methods. When controlled systems are highly nonlinear and complex, it keeps the human insight alive and accessible at the lower levels of the control hierarchy, and so higher levels
can be built on this understanding.
Additionally, the FERM scheme has been extended
to adaptively tune (A-FERM) so that continuous automatic tuning of the parameters can be achieved, and thus be more adaptive to system changes leading to better utilization of network bandwidth. This achieves a level of robustness that is not exhibited by other congestion control schemes reported in the literature.
In this work, the focus is on ATM networks rather than IP based networks. For historical reasons, and due to fundamental philosophical differences in the (earlier) approach to congestion control, the research for control of TCP/IP and ATM based networks proceeded separately. However, some convergence between them has recently become evident. In the TCP/IP literature proposals have appeared on active queue management in routers, and Explicit Congestion Notication (ECN) for IP. It is reasonably expected that, the algorithms developed in this study will be applicable to IP based multimedia networks as well.
| Identifer | oai:union.ndltd.org:ADTP/216481 |
| Date | January 1999 |
| Creators | Sekercioglu, Ahmet, ahmet@hyperion.ctie.monash.edu.au |
| Publisher | Swinburne University of Technology. |
| Source Sets | Australiasian Digital Theses Program |
| Language | English |
| Detected Language | English |
| Rights | http://www.swin.edu.au/), Copyright Ahmet Sekercioglu |
Page generated in 0.002 seconds