Virtual switching is a key component of modern network infrastructures, particularly in cloud computing, data centers, and virtualization. This thesis examines the design, performance, and scalability of various virtual switching solutions, focusing on software-based and hardware-accelerated architectures. It provides a comprehensive evaluation of key virtual switch technologies, including Open vSwitch (OVS), SnabbSwitch, Hyper-Switch, and the Virtual Filtering Platform (VFP). The primary goal is to understand the trade-offs between flexibility, performance, and resource efficiency in virtualized environments. An analysis of existing benchmarking studies compares these solutions in terms of throughput, latency, CPU utilization, and scalability. The findings reveal that while hardware-assisted solutions like SR-IOV and VFIO offer near-native performance, software-based approaches provide greater flexibility and cost-effectiveness. This research provides a comparative analysis of existing studies, drawing from controlled experiments and assessments in industry-standard white papers to highlight each switch’s strengths and limitations. Results indicate that no single virtual switch outperforms all others in every scenario; the optimal choice depends on specific use case requirements. OVS-DPDK and Snabb-Switch excel in high-performance packet processing, Hyper-Switch demonstrates superior scalability in multi-VM environments, and VFP offers robust SDN integration for cloud-native workloads. This thesis contributes by (1) presenting a structured review of existing performance analysis of leading virtual switching solutions and (2) offering recommendations for selecting the most appropriate virtual switch for specific applications. As virtualized networking evolves, these findings serve as a valuable reference for researchers, cloud architects, and network engineers optimizing network virtualization frameworks.
Il virtual switching è un componente chiave delle moderne infrastrutture di rete, in particolare nel cloud computing, nei data center e nella virtualizzazione. Questa tesi esamina il design, le prestazioni e la scalabilità di diverse soluzioni di virtual switching, concentrandosi su architetture basate su software e su accelerazione hardware. Fornisce una valutazione completa delle principali tecnologie di virtual switches, tra cui Open vSwitch (OVS), SnabbSwitch, Hyper-Switch e Virtual Filtering Platform (VFP). L’obiettivo principale è comprendere i compromessi tra flessibilità, prestazioni ed efficienza delle risorse negli ambienti virtualizzati. Un’ analisi di benchmarking esistenti confronta queste soluzioni in termini di throughput, latenza, utilizzo della CPU e scalabilità. I risultati rivelano che, mentre le soluzioni hardware come SR-IOV e VFIO offrono prestazioni quasi native, approcci basati su software sono più flessibili ed economici. Questa ricerca fornisce un’analisi comparativa degli studi esistenti, basandosi su esperimenti controllati e valutazioni presenti in white paper di riferimento nel settore per evidenziare i punti di forza e le limitazioni di ciascun switch. I risultati indicano che nessun virtual switch supera tutti gli altri in ogni scenario; la scelta ottimale dipende dai requisiti specifici del caso d’uso. OVS-DPDK e SnabbSwitch eccellono nell’elaborazione di pacchetti ad alte prestazioni, Hyper-Switch dimostra una scalabilità superiore negli ambienti multi-VM e VFP offre un’integrazione SDN solida per i carichi di lavoro cloud-native. Questa tesi contribuisce (1) presentando una revisione strutturata delle analisi delle prestazioni esistenti delle principali soluzioni di virtual switching e (2) offrendo raccomandazioni per la selezione del virtual switch più appropriato per applicazioni specifiche. Con l’evoluzione della rete virtualizzata, questi risultati rappresentano un riferimento prezioso per ricercatori, cloud architects e ingegneri di rete che desiderano ottimizzare i framework di virtualizzazione di rete.
A comprehensive view of virtual switching: trends and future steps
VERGONA, GIORGIO
2023/2024
Abstract
Virtual switching is a key component of modern network infrastructures, particularly in cloud computing, data centers, and virtualization. This thesis examines the design, performance, and scalability of various virtual switching solutions, focusing on software-based and hardware-accelerated architectures. It provides a comprehensive evaluation of key virtual switch technologies, including Open vSwitch (OVS), SnabbSwitch, Hyper-Switch, and the Virtual Filtering Platform (VFP). The primary goal is to understand the trade-offs between flexibility, performance, and resource efficiency in virtualized environments. An analysis of existing benchmarking studies compares these solutions in terms of throughput, latency, CPU utilization, and scalability. The findings reveal that while hardware-assisted solutions like SR-IOV and VFIO offer near-native performance, software-based approaches provide greater flexibility and cost-effectiveness. This research provides a comparative analysis of existing studies, drawing from controlled experiments and assessments in industry-standard white papers to highlight each switch’s strengths and limitations. Results indicate that no single virtual switch outperforms all others in every scenario; the optimal choice depends on specific use case requirements. OVS-DPDK and Snabb-Switch excel in high-performance packet processing, Hyper-Switch demonstrates superior scalability in multi-VM environments, and VFP offers robust SDN integration for cloud-native workloads. This thesis contributes by (1) presenting a structured review of existing performance analysis of leading virtual switching solutions and (2) offering recommendations for selecting the most appropriate virtual switch for specific applications. As virtualized networking evolves, these findings serve as a valuable reference for researchers, cloud architects, and network engineers optimizing network virtualization frameworks.| File | Dimensione | Formato | |
|---|---|---|---|
|
Master_Thesis _Vergona_Giorgio.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
2.53 MB
Formato
Adobe PDF
|
2.53 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/10589/234707