Advanced Data Plane Techniques for 5G: Hardware-Accelerated Hybrid User Plane Functions and Intelligent Traffic Identification and Prioritization

Fuente: Zenodo
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Esteve Rothenberg, Christian, Singh, Suneet Kumar
Format: Recurso digital
Veröffentlicht: Zenodo 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866901506755657728
author Esteve Rothenberg, Christian
Singh, Suneet Kumar
author_facet Esteve Rothenberg, Christian
Singh, Suneet Kumar
contents <p>Network functions are typically executed on network switches and general-purpose processors, depending on the requirements of network applications and aiming for particular flexibility and programmability. Fixed functionality network switches process packets faster, but because of vendor lock-in, even minor functional modifications take years to implement. On the other hand, general-purpose CPUs give more flexibility and programmability to carry out network activities. Still, they process packets more slowly due to handling interrupts from network interfaces to process incoming and outgoing packets and cache misses. Because of these fundamental limitations, as noted above, the network’s performance can be significantly impacted by: 1) using x86 servers for network activities causes an increase in latency and jitter. 2) classifying traffic on x86 servers to give priority to particular flows. This kind of x86 classification may affect the application’s flow detection time, especially for applications with strict latency requirements. This directly affects the Quality of Service (QoS). To overcome these issues, this dissertation leverages Programmable Data Plane (PDP) switches. PDP switches enable operators to design and implement line-rate packet processing methods. This thesis builds on PDP features to improve the 5G core user plane function (UPF), the most critical part of the mobile network, and enables critical applications requiring strict performance requirements. The contributions consist of: 1) 5G UPF acceleration and scalability by (a) offloading UPF to programmable hardware and (b) dividing UPF functionalities among programmable devices, called hybrid-UPF, based on target-specific features and application needs. 2) traffic classification in the switch hardware, especially for (a) cloud gaming and (b) heavy hitter flows.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_14932486
institution Zenodo
language
publishDate 2024
publisher Zenodo
record_format zenodo
spellingShingle Advanced Data Plane Techniques for 5G: Hardware-Accelerated Hybrid User Plane Functions and Intelligent Traffic Identification and Prioritization
Esteve Rothenberg, Christian
Singh, Suneet Kumar
WP.NW1
5G
P4
heavy-hitters
cloud gaming
machine learning
<p>Network functions are typically executed on network switches and general-purpose processors, depending on the requirements of network applications and aiming for particular flexibility and programmability. Fixed functionality network switches process packets faster, but because of vendor lock-in, even minor functional modifications take years to implement. On the other hand, general-purpose CPUs give more flexibility and programmability to carry out network activities. Still, they process packets more slowly due to handling interrupts from network interfaces to process incoming and outgoing packets and cache misses. Because of these fundamental limitations, as noted above, the network’s performance can be significantly impacted by: 1) using x86 servers for network activities causes an increase in latency and jitter. 2) classifying traffic on x86 servers to give priority to particular flows. This kind of x86 classification may affect the application’s flow detection time, especially for applications with strict latency requirements. This directly affects the Quality of Service (QoS). To overcome these issues, this dissertation leverages Programmable Data Plane (PDP) switches. PDP switches enable operators to design and implement line-rate packet processing methods. This thesis builds on PDP features to improve the 5G core user plane function (UPF), the most critical part of the mobile network, and enables critical applications requiring strict performance requirements. The contributions consist of: 1) 5G UPF acceleration and scalability by (a) offloading UPF to programmable hardware and (b) dividing UPF functionalities among programmable devices, called hybrid-UPF, based on target-specific features and application needs. 2) traffic classification in the switch hardware, especially for (a) cloud gaming and (b) heavy hitter flows.</p>
title Advanced Data Plane Techniques for 5G: Hardware-Accelerated Hybrid User Plane Functions and Intelligent Traffic Identification and Prioritization
topic WP.NW1
5G
P4
heavy-hitters
cloud gaming
machine learning
url https://doi.org/10.5281/zenodo.14932486