Analysis of the High-Speed Network Performance through a Prediction Feedback Based Model

Analysis of the High-Speed Network Performance through a Prediction Feedback Based Model

Manjunath Ramachandra, Pandit Pattabhirama
Copyright: © 2012 |Pages: 17
DOI: 10.4018/978-1-4666-0203-8.ch008
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Performance modeling of a high speed network is challenging, especially when the size of the network is large. The high speed networks span various applications such as the transportation, wireless sensors, et cetera. The present day transportation system makes uses of Internet for efficient command and control transfers. In such a communication system, reliability and in-time data transfer is critical. In addition to the sensor information, the present day wireless networks target to support streaming of multimedia and entertainment data from mobile to infrastructure network and vice versa. In this chapter, a novel modeling method for the network and its traffic shaping is introduced, and simulation model is provided. The performance with this model is analyzed. The case-study with wireless networks is considered. The chapter is essentially about solving the congestion control of packet loss using a differentially fed neural network controller.
Chapter Preview
Top

Background

Independent of its origin, the internet traffic has a few know characteristics in common. It includes

  • •

    A few Predictable parameters in statistical sense

  • •

    Traffic follows poisson distribution

  • •

    Internet traffic is bursty. This is because packets of various size are involved Burstiness of one flow affects the other adaptive flows. This property is useful for the traffic control

  • •

    Overlapping of the independent on-off data sources leads to the arrival pattern distribution with heavy-tailed autocorrelation function.The long-range dependancy in the traffic leads to no “flattening” towards a mean when zoomed out. The same structures may be found at different time scales; hence the traffic is self similar

  • •

    TCP is known to propagate bottleneck self-similarity to the end system. The workaround is to use a model to predict traffic instead of guessing

The other features of the internet traffic include jitter, packet losses, delay, large buffer requirements, less data for decision, time varying characteristics and congestion.

Complete Chapter List

Search this Book:
Reset