Abstract

Multiprotocol Label Switching (MPLS) is an emerging Internet technology that facilitates traffic engineering in service provider networks. This paper considers a network performance optimization problem related to traffic engineering over MPLS. We investigate the issue of dynamic partitioning of MPLS ingress traffic into several parallel Label Switched Paths (LSP). Specifically, we present a stochastic framework for the traffic partitioning problem. Within this framework, a set of parallel edge disjoint LSPs is modeled by parallel queues and a partitioning algorithm is devised for different service classes that is adaptive to the prevailing state of the network. The performance of this approach is illustrated by numerical examples. Document type: Part of book or chapter of book

Abstract

Forecasting Internet traffic is receiving an increasing attention from the computer networks domain. Indeed, by improving this task efficient traffic engineering and anomaly detection tools can be developed, leading to economic gains due to better resource management. This paper presents a Neural Network (NN) approach to predict TCP/IP traffic for all links of a backbone network, using both univariate and multivariate strategies. The former uses only past values of the forecasted link, while the latter is based on the neighbor links of the backbone topology. Several experiments were held by considering real-world data from the UK education and research network. Also, different time scales (e.g. every ten minutes and hourly) were analyzed. Overall, the proposed NN approach outperformed other forecasting methods (e.g. Holt-Winters). R&D Algoritmi centre Document type: Part of book or chapter of book

Abstract

The transformation of the Internet into an important and ubiquitous commercial infrastructure has not only created rapidly rising bandwidth demands but also significantly changed consumer expectations in terms of performance, security and services. Consequentially as service providers attempt to encourage business and leisure applications on to the Internet, there has been a requirement for them to develop an improved IP network infrastructure in terms of reliability and performance [1]. Interest in congestion control through traffic engineering has arisen from the knowledge that although sensible provisioning of the network infrastructure is needed together with sufficient underlying capacity, these are not sufficient to deliver the QoS required [2]. This is due to dynamic variations in load. In operational IP networks, it has been difficult to incorporate effective traffic engineering due to the limited capabilities of the IP technology. In principle, Multiprotocol Label Switching (MPLS), a connection-oriented label swapping technology, offers new possibilities in addressing the limitations by allowing the operator to use sophisticated traffic control mechanisms.\r \r However, as yet, the traffic engineering capabilities offered by MPLS have not been fully exploited. Once label switched paths (LSPs) have been provisioned through the service providers' network, there are currently no management facilities for dynamic re-optimisation of traffic flows. The service level agreements (SLAs) between the network operator and the customer are agreed in advance of the commencement of traffic flow, and these are mapped to particular paths throughout the provider's domain and may be maintained for the duration of the contract. During transient periods, the efficiency of resource allocation could be increased by routing traffic away from congested resources to relatively under-utilised links. Some means of restoring the LSPs to their original routes once the transient congestion has subsided is also desirable. Document type: Part of book or chapter of book