Performance Analysis of the Adaptive Available Bit Rate (ABR) Voice over ATM Networks Systems

Expert Analysis

There are different levels of service in Asynchronous Transfer Mode (ATM) networks where each service type is suitable for a specific application. Assignment of applications to ATM service types depends on the nature of the application to be served. If the application is more stringent to loss and delay, the network will provide more guarantees about the quality delivered. This means higher cost, in terms of bandwidth from the operator's point of view, and in terms of charge from the customer's point of view.

Voice is a real-time application that is sensitive to loss, delay and delay variation with different degrees. Hence, the network has to provide strict guarantees regarding the quality delivered. Long delays affect the interactivity of voice, and this is where crosstalk starts to occur.

Due to these reasons, ATM supports voice as a constant bit rate (CBR) traffic, where the network delivers the traffic at a constant rate, with no losses or delays. However, it is expensive to use CBR service. Since voice in its inherent nature is variable (it is known that the human voice has a speech activity of about 42%), the CBR service does not take advantage of the silence periods, and hence does not achieve any multiplexing gain.

Since CBR is not the most efficient service for voice, other service categories may be more suitable for voice. The Available Bit Rate (ABR) service has many attractive features that motivated this study. Unlike the best-effort service used in data networks (e.g. the Internet); ABR can guarantee a minimum bandwidth to individual connections using admission control procedures. This means you can provide different levels of service for the same traffic type based on the user's requirements. The higher the minimum cell rate, MCR is, the better the quality. Moreover, ABR service provides continuous feedback to the user in order to control the transmission rate based on the network status. When there is a congestion, it asks the user either to slow down or to stop sending at all, depending on the severity of the congestion. By doing this, the ABR service provides control over the number of losses and queues build-up resulting in a better quality than the best-effort model.

Due to the fact that the ABR service provides no guarantees on delays and losses, it is expected that the quality achieved will not be as good as the one achieved by the CBR service. The question that is important to answer is how much degradation in the quality will occur, and up to what level of degradation the quality offered is still acceptable and for which application.

This work investigated the degradation in voice quality when transferred over the ABR service. It also finds out how much multiplexing gain can be achieved. The efficiency of the scheme under different scheduling/drop policies and other operating conditions and environments has been evaluated using simulation modeling.

This consultant evaluated the performance of ABR for the transfer of voice over ATM networks. The average rate, degradation in voice quality, and different parameters that affect the voice quality are analyzed using simulation modeling. Under high loads, round robin (RR) scheduling policy seems to provide better voice quality than the earliest deadline first (EDF) and longest queue first (LQF). The more frequently RM cells are sent the better the voice quality. Furthermore, even though large queues result in less CLR, they introduce higher delays. It is found that the smaller the per-VC queue size, the better the voice quality, due to the fact that bandwidth is being utilized for useful cells. Using ABR for the transfer of voice does not promise a voice quality as good as that of CBR. However, it provides a cheaper alternative, and allows for providing different levels of service. It is observed that the scheduling algorithm affects the quality achieved. Furthermore, sensitivity analysis is applied to different parameters, such as queue size, and averaging interval length, to investigate their impact on the performance metrics. Sample results are presented below.

To see the resume of the expert associated with this case study, see the link below.