Boosting application performance using heterogeneous virtual channels

Interactive networked applications require high throughput, low latency, and high reliability from the network to provide a seamless user experience. While meeting these three requirements simultaneously is challenging, there has been an emergence of heterogeneous virtual channels (HVCs), which support some subset of these requirements at the expense of others. For instance, Ultra-Reliable and Low-Latency Communication (URLLC) sacrifices throughput to achieve low latency and high reliability in 5G New Radio (NR), and Wi-Fi 7, along with novel Internet architectures, provides similarly disparate types of service. Prior work either focuses on aggregating the bandwidth of these channels while neglecting their unique properties or fails to generalize in terms of achieving high performance across different applications and channels. To fully utilize HVCs, we argue that there are challenges and opportunities across the network, transport, and application layers of the network stack, and we explore the trade-offs of these architectural choices. We propose a transport layer solution that performs fine-grained segment steering between the HVCs, incorporates an HVC-aware congestion controller, and integrates application inputs such as message priority and sizes via a new application-transport interface. We believe this approach is the right choice for leveraging HVCs to boost application performance. In this thesis, we focused on real-time video streaming, utilizing Scalable Video Coding (SVC) to steer different video layers according to their priority through the high bandwidth enhanced Mobile Broadband (eMBB) channel and the low latency URLLC channel. We initially implemented our priority-based steering as a proof of concept using User Datagram Protocol (UDP), and then advanced to a transport layer solution using Picoquic, a minimalist implementation of the Quick UDP Internet Connections (QUIC) protocol. Our evaluations show that that our UDP-based implementation, we reduced the 95th percentile latency by 1982.05 ms (26.41x improvement) and 98.00 ms (2.26x improvement) over the eMBB only scheme and network layer packet steering scheme (DChannel [1]) respectively. On the other hand, our Picoquic-based implementation decreased the 95th percentile latency by 931.00 ms (12.49x improvement) and 1445.00 ms (18.84x improvement) over the eMBB only scheme and Picoquic’s unmodified Multipath respectively.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2024

Author dc:creator

• Touseef, Talal

Contributors dc:contributor

• Godfrey, Philip Brighten

• Heterogeneous Virtual Channels
• 5g
• Transport Layer
• Real-time Video Streaming
• Scalable Video Coding

Statement dc:rights

• Copyright 2024 Talal Touseef

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/124493

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F124493