Non-Terrestrial Networks (NTN) based on Low Earth Orbit (LEO) satellites are a key component of 5G New Radio (NR) Rel-17 standardization, enabling global broadband con- nectivity. However, the extended propagation delays intrinsic to satellite links—combined with NTN-specific scheduling offsets—create significant challenges for standard TCP con- gestion control algorithms that were designed for terrestrial networks with sub-millisecond round-trip times. This thesis presents an empirical evaluation of TCP congestion control performance over a real 5G NTN LEO testbed built using OpenAirInterface5G (OAI) software and USRP N300 software-defined radios, connected through a Keysight F64 channel emulator. The work first compares TCP BBR, Cubic, and Reno throughput at 106 PRB (20 MHz) and 160 PRB (30 MHz) configurations, and then investigates the impact of three protocol- layer timing parameters—PDCP t-Reordering, RLC t-Reassembly, and PDCP discard- Timer—on TCP BBR throughput at 160 PRB. A key contribution is the implementation of PDCP discardTimer functionality in OAI, which was previously declared but never executed in the codebase. The experimental results demonstrate that BBR outperforms Cubic and Reno by a factor of 2.9× and 3.7× respectively at 160 PRB, and that PDCP t-Reordering set to 0 ms yields a 72% through- put improvement over the OAI default. RLC t-Reassembly shows an optimal value at the 3GPP default of 35 ms, revealing that timer optimization must be performed indepen- dently at each protocol layer.
Le reti non terrestri (NTN) basate su satelliti in orbita bassa (LEO) rappresentano una componente fondamentale della standardizzazione 5G New Radio (NR) Rel-17, abilitando la connettività a banda larga globale. Tuttavia, i ritardi di propagazione estesi intrinseci ai collegamenti satellitari—uniti agli offset di scheduling specifici delle NTN—creano sfide significative per gli algoritmi standard di controllo della congestione TCP, progettati per reti terrestri con tempi di andata e ritorno inferiori al millisecondo. Questa tesi presenta una valutazione empirica delle prestazioni del controllo della con- gestione TCP su un testbed reale 5G NTN LEO costruito con il software OpenAirInter- face5G (OAI) e radio software-defined USRP N300, collegati attraverso un emulatore di canale Keysight F64. Il lavoro confronta dapprima il throughput di TCP BBR, Cubic e Reno nelle configurazioni a 106 PRB (20 MHz) e 160 PRB (30 MHz), per poi analizzare l’impatto di tre parametri di temporizzazione a livello protocollare—PDCP t-Reordering, RLC t-Reassembly e PDCP discardTimer—sul throughput di TCP BBR a 160 PRB. Un contributo chiave è l’implementazione in OAI della funzionalità PDCP discardTimer, in precedenza dichiarata ma mai eseguita nel codice. I risultati sperimentali dimostrano che BBR supera Cubic e Reno rispettivamente di un fattore 2.9× e 3.7× a 160 PRB, e che il parametro PDCP t-Reordering impostato a 0 ms produce un incremento del through- put del 72% rispetto al valore predefinito di OAI. Il timer RLC t-Reassembly presenta un valore ottimale in corrispondenza del valore predefinito 3GPP di 35 ms, rivelando che l’ottimizzazione dei timer deve essere effettuata in modo indipendente a ciascun livello protocollare.
Empirical evaluation of TCP congestion control over 5G NTN LEO using OpenAirInterface: impact of protocol layer timing parameters
JAVAHERI NEYESTANAK, IMAN
2025/2026
Abstract
Non-Terrestrial Networks (NTN) based on Low Earth Orbit (LEO) satellites are a key component of 5G New Radio (NR) Rel-17 standardization, enabling global broadband con- nectivity. However, the extended propagation delays intrinsic to satellite links—combined with NTN-specific scheduling offsets—create significant challenges for standard TCP con- gestion control algorithms that were designed for terrestrial networks with sub-millisecond round-trip times. This thesis presents an empirical evaluation of TCP congestion control performance over a real 5G NTN LEO testbed built using OpenAirInterface5G (OAI) software and USRP N300 software-defined radios, connected through a Keysight F64 channel emulator. The work first compares TCP BBR, Cubic, and Reno throughput at 106 PRB (20 MHz) and 160 PRB (30 MHz) configurations, and then investigates the impact of three protocol- layer timing parameters—PDCP t-Reordering, RLC t-Reassembly, and PDCP discard- Timer—on TCP BBR throughput at 160 PRB. A key contribution is the implementation of PDCP discardTimer functionality in OAI, which was previously declared but never executed in the codebase. The experimental results demonstrate that BBR outperforms Cubic and Reno by a factor of 2.9× and 3.7× respectively at 160 PRB, and that PDCP t-Reordering set to 0 ms yields a 72% through- put improvement over the OAI default. RLC t-Reassembly shows an optimal value at the 3GPP default of 35 ms, revealing that timer optimization must be performed indepen- dently at each protocol layer.| File | Dimensione | Formato | |
|---|---|---|---|
|
Thesis_Iman _v2.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Thesis Report
Dimensione
6.27 MB
Formato
Adobe PDF
|
6.27 MB | Adobe PDF | Visualizza/Apri |
|
Executive_Summary_Iman_v2.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Executive Summary
Dimensione
2.18 MB
Formato
Adobe PDF
|
2.18 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/10589/261534