The main idea on which Software-Defined Networking (SDN) is based, is the separation of the control plane from the data plane. This new structure has inspired many programmers to find new solutions to improve network performance. SDN controller simplifies network configuration via high-level languages and abstraction. SDN-IP, an example of SDN application developed by ONF (Open Networking Foundation) inside ONOS (Open Network Operating System), allows a SDN network to connect to external networks on the Internet using the standard Border Gateway Protocol (BGP). ONOS builds these connections through the use of high-level policies, ONOS Intents, that, with the Intent Framework, are translated in OpenFlow rules. This allows the programmers not to think about how to write the low-level rules device by device. The NOS (Network Operating System) hides all the details of the process compilation and path computation, which without further constraints, just translate each Intent individually to one of the shortest paths. In this work, the possibility of jointly re-optmizing the paths of the demands has been explored, through the monitoring of flow statistics into the network. The extension of the application SDN-IP would allow to monitor, periodically, the AS-to-AS traffic and re-route it to minimize the network congestion and so the Maximum Link Utilization. At the same time, we don't break the simplicity of the mapping between BGP announcements and the corresponding Intents. To avoid too frequent network reconfigurations, that would create network instability, we use a set of robust routing solution with a minimum time constraint in which a routing should be kept stable. At the beginning, the traffic is forward according to the rules of the shortest path (default behaviour of Intents) and measured for an interval time (e.g. a day). Afterwards, through interfacing with an off-platform module, we compute a set of robust configurations to be applied for the next period. The module, also, via REST-API, gathers the measurements, defines two optimization models to compute the configurations and the activation times, and, finally, schedules the activation into the system. To cope with traffic deviations, with respect to the corresponding traffic profile of the training period, routings are computed to be robust over subsets of traffic matrix space. Our goal is modify SDN-IP application to adapt the Intents, created according to BGP announcements, to PointToPoint Intents and periodically re-optimize the paths according to their statistics.
L'idea principale su cui SDN è basato, è la separazione del piano di controllo da quello di inoltro. Questa nuova struttura ha inspirato molti programmatori a trovare nuove soluzioni per migliorare le performance della rete. Il controller SDN semplifica la configurazione della rete attraverso l'astrazione e l'utilizzo di linguaggi ad alto livello. SDN-IP, un esempio di applicazione SDN sviluppata da ONF (Open Networking Foundations) all'interno di ONOS (Open Network Operating System), permette ad una rete SDN di connettersi a reti esterne attraverso Internet usando lo standard BGP (Border Gateway Protocol). ONOS costruisce queste connessioni attraverso l'uso di policy ad alto livello, chiamati Intenti ONOS, che, con l'Intent Framework, vengono tradotte in regole OpenFlow. Questo aiuta i programmatori a non pensare più a come scrivere le regole a basso livello su ogni device. Il NOS (Sistema Operativo di Rete) nasconde tutti i dettagli del processo di compilazione e del calcolo dei path, e, senza ulteriori vincoli, traduce individualmente ogni Intento in un possibile shortest path. In questo lavoro, viene esplorata la possibilità di ri-ottimizzare contemporaneamente i percorsi delle domande attraverso il monitoraggio delle statistiche dei flussi all'interno della rete. L'estensione dell'applicazione SDN-IP permetterebbe il monitoraggio periodico del traffico AS-to-AS, reinstradandolo con lo scopo di minimizzare la congestione della rete e quindi il massimo utilizzo dei link. Allo stesso tempo, non vogliamo rompere la semplicità con cui vengono mappati gli annunci BGP con i corrispondenti Intenti. Per evitare frequenti cambi di configurazione di rete, che creerebbero instabilità, utilizziamo un set di soluzioni di routing robusto imponendo un tempo minimo in cui il routing deve essere mantenuto stabile. All'inizio, il traffico viene instradato secondo le regole dello shortest path (il comportamento predefinito degli Intenti) e misurato per un intervallo di tempo (per esempio un giorno). Successivamente, attraverso l'interfacciamento con un modulo esterno, calcoliamo un set di configurazioni di routing robusto da applicare nell'intervallo successivo. Il modulo, inoltre, via REST API, raccoglie le misure, definisce due modelli di ottimizzazione per calcolare le configurazioni e i tempi di attivazione, e, infine, schedula le loro attivazioni all'interno del sistema. Per far fronte alle possibili deviazioni di traffico, rispetto al profilo studiato durante il periodo di monitoraggio, le configurazioni sono calcolate per essere robuste ad un sottoinsieme di matrici di traffico. Il nostro obiettivo è quello di modificare l'applicazione SDN-IP per adattare gli Intenti, creati in accordo agli annunci BGP, in Intenti PointToPoint e periodicamente riottimizzare i percorsi secondo le loro statistiche.
Towards smart intents with robust and flexible routing
GULLÌ, MATTIA
2016/2017
Abstract
The main idea on which Software-Defined Networking (SDN) is based, is the separation of the control plane from the data plane. This new structure has inspired many programmers to find new solutions to improve network performance. SDN controller simplifies network configuration via high-level languages and abstraction. SDN-IP, an example of SDN application developed by ONF (Open Networking Foundation) inside ONOS (Open Network Operating System), allows a SDN network to connect to external networks on the Internet using the standard Border Gateway Protocol (BGP). ONOS builds these connections through the use of high-level policies, ONOS Intents, that, with the Intent Framework, are translated in OpenFlow rules. This allows the programmers not to think about how to write the low-level rules device by device. The NOS (Network Operating System) hides all the details of the process compilation and path computation, which without further constraints, just translate each Intent individually to one of the shortest paths. In this work, the possibility of jointly re-optmizing the paths of the demands has been explored, through the monitoring of flow statistics into the network. The extension of the application SDN-IP would allow to monitor, periodically, the AS-to-AS traffic and re-route it to minimize the network congestion and so the Maximum Link Utilization. At the same time, we don't break the simplicity of the mapping between BGP announcements and the corresponding Intents. To avoid too frequent network reconfigurations, that would create network instability, we use a set of robust routing solution with a minimum time constraint in which a routing should be kept stable. At the beginning, the traffic is forward according to the rules of the shortest path (default behaviour of Intents) and measured for an interval time (e.g. a day). Afterwards, through interfacing with an off-platform module, we compute a set of robust configurations to be applied for the next period. The module, also, via REST-API, gathers the measurements, defines two optimization models to compute the configurations and the activation times, and, finally, schedules the activation into the system. To cope with traffic deviations, with respect to the corresponding traffic profile of the training period, routings are computed to be robust over subsets of traffic matrix space. Our goal is modify SDN-IP application to adapt the Intents, created according to BGP announcements, to PointToPoint Intents and periodically re-optimize the paths according to their statistics.| File | Dimensione | Formato | |
|---|---|---|---|
|
Master Thesis Gullì.pdf
accessibile in internet per tutti
Descrizione: Testo della tesi
Dimensione
4.02 MB
Formato
Adobe PDF
|
4.02 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/10589/137392