Neural-Based Quality of Service Estimation in MPLS Routers

Overview

The MultiProtocol Label Switching (MPLS) is being considered as one possible approach for scalable Quality of Service (QoS) provisioning in the Internet. As MPLS is standardized by Internet Engineering Task Force (IETF) and is expected to be implemented in the near future, this paper addresses the issue of QoS estimation in a MLPS router fed with bursty traffic by means of a neural network. The latter is trained to anticipate the average delay and the Packet Loss Rate (PLR) within the MPLS node for each class of service, given the mean rates, the peak rates and the coefficients of variation of the packet inter-arrival time of the incoming traffic flows.