Please note: This PhD seminar will take place in DC 1304.
Mohamed Rouili, PhD candidate
David R. Cheriton School of Computer Science
Supervisor: Professor Raouf Boutaba
The stringent quality of service (QoS) requirements in 5G networks, particularly for ultra-reliable low-latency communication (urLLC) use cases, pose significant challenges for the centralized unit user plane (CU-UP) within the Radio Access Network (RAN). Current general-purpose processor (GPP)-based CU-UP deployments struggle to meet stringent QoS requirements, such as sub-millisecond latency and high reliability, especially under high traffic loads.
We propose Blink, a P4-based system that offloads CU-UP functions to programmable switch ASICs, enabling high-speed packet processing and improved QoS. Blink employs a novel trigger-based offloading mechanism that can be dynamically adapted to network conditions, ensuring seamless transitions between GPP-based and P4-based processing. In our evaluations, Blink achieves a 1,941× lower median packet processing latency compared to GPP-based CU-UP deployments, while delivering 12% higher throughput and maintaining near-zero packet loss of 0.001%, crucial for urLLC applications. These improvements are sustained across multiple distributed units (DUs) and user equipment (UEs), making Blink a scalable and flexible solution for enhancing CU-UP performance in 5G networks.