Network Function Virtualization (NFV) has dramatically increased the flexibility in the deployment of network services, but the virtualization of functions on compute nodes can hinder their performance compared to that of the middleboxes they try to replace. The use of programmable network hardware to perform part of the processing at line rate can drastically increase throughput. This thesis proposes a framework for the deployment of heterogeneous Service Function Chains (SFCs), defined as a combination of functions for regular compute hosts as Docker containers and for programmable switches using the P4 language. Programmable data planes are also exploited to perform real time monitoring of the services through In-band Network Telemetry (INT) to guarantee requested levels of performance by redeploying and rerouting sections that are affected by adverse conditions, allowing applications with critical requirements to be deployed as SFCs. The solution has been tested by simulating various topologies and services on the FOP4 platform with bmv2 switches. This analysis showed that the system is capable of detecting faults in the order of hundreds of milliseconds, and the overhead it causes in the process of redeployments is negligible compared to the startup time of functions. Measurements also revealed that the current bottleneck for the runtime relocation of heterogeneous functions is the deployment of P4 programs.
La Network Function Virtualization (NFV) ha notevolmente esteso la flessibilità della messa in campo di servizi di rete, ma la virtualizzazione delle funzioni su server può ridurre le loro prestazioni rispetto a quelle delle middlebox che cercano di rimpiazzare. L'esecuzione di una parte dell'elaborazione a line rate, grazie all'uso di hardware di rete programmabile, può aumentare considerevolmente il throughput. Questa tesi propone un framework per l'installazione di Service Functions Chains (SFCs) eterogenee, definite come combinazione di funzioni sotto forma di container Docker per tradizionali host di calcolo e di funzioni in linguaggio P4 per switch programmabili. I Programmable Data Planes vengono sfruttati anche per eseguire monitoraggio in tempo reale dei servizi grazie alla In-band Network Telemetry (INT), la quale permette di garantire i livelli di prestazione richiesti attraverso la rilocazione ed il reinstradamento delle sezioni influenzate da condizioni avverse. Ciò permette la realizzazione di applicazioni con requisiti critici sotto forma di SFC. La soluzione è stata testata simulando varie topologie e servizi sulla piattaforma FOP4 con switch bmv2. Questa analisi ha dimostrato che il sistema è in grado di rilevare guasti nell'ordine di centinaia di millisecondi e che l'overhead causato durante il processo di rilocazione è trascurabile rispetto al tempo di avvio delle funzioni. Le misurazioni hanno inoltre mostrato che nel processo di trasferimento di funzioni eterogenee durante la loro esecuzione, l'attuale collo di bottiglia è rappresentato dall'installazione di programmi P4.
CHIMA : a framework for network services deployment and performance assurance
Battiston, Elia
2020/2021
Abstract
Network Function Virtualization (NFV) has dramatically increased the flexibility in the deployment of network services, but the virtualization of functions on compute nodes can hinder their performance compared to that of the middleboxes they try to replace. The use of programmable network hardware to perform part of the processing at line rate can drastically increase throughput. This thesis proposes a framework for the deployment of heterogeneous Service Function Chains (SFCs), defined as a combination of functions for regular compute hosts as Docker containers and for programmable switches using the P4 language. Programmable data planes are also exploited to perform real time monitoring of the services through In-band Network Telemetry (INT) to guarantee requested levels of performance by redeploying and rerouting sections that are affected by adverse conditions, allowing applications with critical requirements to be deployed as SFCs. The solution has been tested by simulating various topologies and services on the FOP4 platform with bmv2 switches. This analysis showed that the system is capable of detecting faults in the order of hundreds of milliseconds, and the overhead it causes in the process of redeployments is negligible compared to the startup time of functions. Measurements also revealed that the current bottleneck for the runtime relocation of heterogeneous functions is the deployment of P4 programs.| File | Dimensione | Formato | |
|---|---|---|---|
|
2021_12_Battiston.pdf
accessibile in internet per tutti
Descrizione: Executive summary e testo tesi
Dimensione
6.52 MB
Formato
Adobe PDF
|
6.52 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/10589/181816