In this thesis, the problem of estimating from data the Traffic Matrix of a computer network is tackled. Such a matrix, which contains the volume of traffic flowing among each couple of nodes in a network, is a powerful tool exploited in many aspects of network management, such as capacity planning, load balancing and anomaly detection. The proposed technique exploits a new observer based on Kalman Filter, suitably modified to better fit into the specific case study. Throughout the thesis, data extraction and analysis aspects, modeling and validation are discussed in detail. Experimental results show that the proposed estimation method can obtain very low error rates, even in presence of hard technological constraints. All the data used for the analysis are extracted from CAMPUS-3, a Metropolitan Area Network of the Milan municipality, managed by British Telecom.
In questa tesi viene affrontato il problema della stima della matrice di traffico in una rete informatica. Questa matrice, che contiene il volume del traffico scambiato tra ogni coppia di nodi della rete, è un potente strumento utilizzato in varie fasi della gestione della rete stessa, come capacity planning, load balancing e anomaly detection. La tecnica proposta utilizza una versione modificata del Filtro di Kalman, adattata allo specifico caso in analisi. L'elaborato descrive le varie fasi del processo di stima: estrazione e analisi dei dati, modellizzazione e validazione. I risultati sperimentali mostrano che la soluzione proposta riesce a raggiungere valori di errore molto bassi, nonostante la presenza di notevoli difficoltà tecnologiche. Tutti i dati utilizzati provengono dalla rete CAMPUS-3, una rete metropolitana gestita da British Telecom per conto del Comune di Milano.
A Kalman filtering approach for traffic matrix estimation in computer networks
POZZI, GIANMARIO
2016/2017
Abstract
In this thesis, the problem of estimating from data the Traffic Matrix of a computer network is tackled. Such a matrix, which contains the volume of traffic flowing among each couple of nodes in a network, is a powerful tool exploited in many aspects of network management, such as capacity planning, load balancing and anomaly detection. The proposed technique exploits a new observer based on Kalman Filter, suitably modified to better fit into the specific case study. Throughout the thesis, data extraction and analysis aspects, modeling and validation are discussed in detail. Experimental results show that the proposed estimation method can obtain very low error rates, even in presence of hard technological constraints. All the data used for the analysis are extracted from CAMPUS-3, a Metropolitan Area Network of the Milan municipality, managed by British Telecom.File | Dimensione | Formato | |
---|---|---|---|
kalman-filtering-approach.pdf
non accessibile
Descrizione: Thesis text
Dimensione
2.74 MB
Formato
Adobe PDF
|
2.74 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/136035