Prediction based link state update

  • Authors:
  • Q. Wang;J. Vincent;G. King

  • Affiliations:
  • Southampton Solent University, UK;Bournemouth University, UK;Southampton Solent University, UK

  • Venue:
  • International Journal of Computers and Applications
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

The deployment of communication-intensive, real-time multimedia applications on the Internet presents challenges to network routing, as these applications often demand more bandwidth and are less tolerant to delay, delay jitter and loss than traditional data applications. The current best-effort service cannot satisfy such Quality-of-Service (QoS) requirements. Recent studies have shown that QoS routing can provide increased network utilization and balance the network load, with the prerequisite that up-to-date link state information is available for routing decisions. Considering the dynamic nature of networks, this requirement is difficult to fulfil in practice. To improve dynamic routing performance without introducing additional link state update overhead, a link state update policy based on the prediction of link utilization is proposed. The results of statistical analysis and network simulation are presented. The statistical analysis finds that a time-series of mean available bandwidth (mABW) is predictable within certain limits. The prediction approach is shown in simulations of representative networks to offer substantial performance improvements when using periodic updates and also improvements for triggered updates with hold-down timer. A reduction in signalling failures and the effective link state update rate is indicated for a range of circumstances without reduction in routing performance.