In this project, a method for deciding when to upgrade links in an IP network was developed. The decision is made once a threshold in the utilization of the link is crossed, making the decision method easy to implement once the threshold has been calculated. The threshold is calculated using a Markov decision process which includes the traffic model and the costs and rewards involved in running the link. The threshold for Poisson traffic and bursty ON/OFF traffic is compared, and it is shown that having bursty traffic significantly lowers the upgrade threshold; the specific forecasting model for the long-term traffic fluctuations on the other hand has only marginal impact on the upgrade thresholds.
|Effective start/end date||01/01/2007 → 01/11/2008|