Traffic Controller for Handling Service Quality in Multimedia Network

Introduction

The multimedia traffic over the Internet has characteristic requirements and features such as large buffers, real time data transfers, user interaction and monitoring, bursty traffic etc. The traffic has a few Predictable parameters in statistical sense that should be made use for effective control. The traffic pattern obeys poisson distribution. The multimedia traffic is inherently bursty and time varying due to different degree of compression for the data of a unit time. The burstiness of one flow affects other adaptive flows. This property is used in traffic control.

The multimedia data sources are often modeled as on-off sources. The overlapping of independent on-off sources leads to arrival pattern distribution with heavy-tailed autocorrelation function. This heavy tail distribution imparts Long range dependency to the traffic. As a result, there is no flattening of the distribution towards a mean when zoomed out in time scale.

As a result of burstiness or Long-range dependency, the resources would get flooded resulting in packet losses for a long duration. The other problems of long range dependency include jitter, Delay, large buffer requirements, congestion etc. As a fall out of these issues, it would be nearly impossible to meet the committed quality of service.

In a home network scenario, there will be limited support to overcome congestion. The best way to fix this issue is to avoid the congestion to happen. So, a proactive control algorithm is required than an active one (Hollot,C., Misra,V., Towsley,D., & Gong,W, 2001). A good controller has to foresee the trends in the network traffic variations and provide inputs to the traffic source well in advance as shown in Figure 1.

Figure 1.

Traffic controller

The source would get sufficient time to adjust the traffic rate or provide sufficient redundancies with the appropriate channel coding schemes so that it would not flood the channel when it is disturbed. In this chapter, the proactive queue management model GREEN (Wu-chun Feng; Kapadia, A., & Thulasidasan, S) together with a neural network is considered.

Background

Meeting stringent constraints on the delay is very important in the networks supporting multimedia traffic for seamless user experience. In addition to the delay and packet losses, the variable delay suffered by the packets i.e. jitter is to be given due weightage. Analysis shows that the jitter degrades the perceptual quality as much as the packet loss (Mark Claypool, Jonathan Tanner,1999).

To support the QoS in the Internet, the IETF has defined two architectures:

• • The Integrated Services (Intsev) and

• • Differentiated Services (Diffserv).

They have important differences in both the service definition and the implementation architectures. At the service definition level, the Intserv provides end-to-end guarantees or controlled load service (El-Haddadeh, R., Taylor, G.A., & Watts, S.J, 2004) on a per flow basis, while the Diffserv provides a coarser level of service differentiation among a small number of traffic classes.