Traditional computer networks made of switches, routers and specialized hardware use a multiplicity of proprietary configuration interfaces that vary over different vendors, and even among different devices from the same vendor. Software-Defined Networking (SDN) is a new network paradigm based on the decoupling of the Control Plane from the Data Plane, moving the control logic to a remote centralized control platform. The goal is to make networks more programmable in order to increase the interoperability between vendors and speed up the development of new solutions. Traffic engineering benefits of the SDN flexibility include adaptive and more efficient forwarding decisions and faster network reconfiguration. Operators have to deal with uncertainty in the demands and, moreover, with traffic scenarios, i.e. traffic matrices, varying among time. Thus, the faster network reconfigurability becomes an important tool avoiding oversized networks, but at the same time guaranteeing good network performance without congestion situations. The traditional approach computes a robust routing solution working for a multiplicity of traffic matrices, which, tailored on the most critical scenarios, may lead the network to underperform during normal operations. Exploiting the global view of the controller, instead, it is possible to monitor the status of the network and find the most appropriate routing solution. However, finding an optimal solution requires unrealistic computational times, therefore, we propose three heuristic procedures. They can discover similarities in network scenarios and group the traffic matrices in clusters according to the routing solution in each of them. They have different features, trying to cover different operators' requirements, but with the common target of reducing the network congestion. Finally, the resulting clusters are used for finding robust routing solutions within each cluster and, then, the network is reconfigured according to these solutions. We have carried out a complete performance evaluation of our algorithms, comparing the solutions found by the traditional approach with those proposed by the clusterization procedures. We shows how our algorithms can reduce the network congestion exploiting SDN flexibility.
Le reti di calcolatori composte da switch, router e dispositivi specializzati, utilizzano una molteplicità di interfacce proprietarie che variano tra i diversi produttori e anche tra i diversi dispositivi dello stesso produttore. Il Software-Defined Networking (SDN) è un nuovo paradigma di rete basato sul disaccoppiamento tra control plane e data plane, spostando la logica di controllo in una piattaforma di controllo remota e centralizzata. L’obiettivo è rendere le reti più programmabili in modo da aumentare l’interoperabilità tra produttori e velocizzare lo sviluppo di nuove soluzioni. L’ingegneria del traffico trae beneficio dalla flessibilità del SDN introducendo decisioni di instradamento adattative e più efficienti, oltre a più veloci riconfigurazioni di rete. Gli operatori devono far fronte all'incertezza nelle domande e, inoltre, a scenari di traffico, cioè matrici di traffico, che variano nel tempo. Quindi la più veloce riconfigurazione della rete diventa uno strumento importante per non sovradimensionarla, ma al contempo garantire buone prestazioni, evitando situazioni di congestione. L’approccio tradizionale calcola una soluzione di instradamento robusta per una molteplicità di matrici di traffico, la quale, essendo adattata per gli scenari più critici, può portare la rete a funzionare con prestazioni peggiori durante il normale utilizzo. Sfruttando la visibilità globale del controllore, è invece possibile monitorare lo stato della rete e trovare la soluzione di instradamento più accurata. Tuttavia, trovare la soluzione ottima richiede tempi computazionali proibitivi, quindi proponiamo tre procedure euristiche. Questi metodi possono individuare somiglianze negli scenari di rete e riunire le matrici di traffico in gruppi, calcolando la soluzione di instradamento per ogni gruppo. Le procedure hanno caratteristiche differenti cercando di assecondare le diverse esigenze degli operatori, ma con l’obiettivo comune di voler ridurre la congestione di rete. Infine, i gruppi risultanti vengono utilizzati per trovare soluzioni robuste di instradamento all'interno di ciascun gruppo e successivamente la rete viene riconfigurata in accordo a queste soluzioni. Faremo alcune valutazioni riguardo le prestazioni, comparando la soluzione trovata dall'approccio tradizionale con il metodo di raggruppamento proposto, mostrando come sia possibile ridurre la congestione di rete sfruttando la flessibilità del SDN.
Clustered robust routing solutions in software defined networks
ROSADA, NICOLA
2015/2016
Abstract
Traditional computer networks made of switches, routers and specialized hardware use a multiplicity of proprietary configuration interfaces that vary over different vendors, and even among different devices from the same vendor. Software-Defined Networking (SDN) is a new network paradigm based on the decoupling of the Control Plane from the Data Plane, moving the control logic to a remote centralized control platform. The goal is to make networks more programmable in order to increase the interoperability between vendors and speed up the development of new solutions. Traffic engineering benefits of the SDN flexibility include adaptive and more efficient forwarding decisions and faster network reconfiguration. Operators have to deal with uncertainty in the demands and, moreover, with traffic scenarios, i.e. traffic matrices, varying among time. Thus, the faster network reconfigurability becomes an important tool avoiding oversized networks, but at the same time guaranteeing good network performance without congestion situations. The traditional approach computes a robust routing solution working for a multiplicity of traffic matrices, which, tailored on the most critical scenarios, may lead the network to underperform during normal operations. Exploiting the global view of the controller, instead, it is possible to monitor the status of the network and find the most appropriate routing solution. However, finding an optimal solution requires unrealistic computational times, therefore, we propose three heuristic procedures. They can discover similarities in network scenarios and group the traffic matrices in clusters according to the routing solution in each of them. They have different features, trying to cover different operators' requirements, but with the common target of reducing the network congestion. Finally, the resulting clusters are used for finding robust routing solutions within each cluster and, then, the network is reconfigured according to these solutions. We have carried out a complete performance evaluation of our algorithms, comparing the solutions found by the traditional approach with those proposed by the clusterization procedures. We shows how our algorithms can reduce the network congestion exploiting SDN flexibility.| File | Dimensione | Formato | |
|---|---|---|---|
|
2016_12_Rosada.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
3.03 MB
Formato
Adobe PDF
|
3.03 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/10589/132220