IBC2024 Accelerator project: Scalable Ultra-Low Latency Streaming for Premium Sports

Proposed by Comcast and supported by champions Comcast, BT Media & Broadcast, AP, and Globo, and participants, Videoclarity, Castlabs, Ateme, AMD and HP, the Scalable Ultra-Low Latency Streaming for Premium Sports project seeks to achieve Twitter-equivalent latency and near-instant playback start using standard HTTP streaming technical stack and infrastructure. The areas explored will be the use of low-latency encoding and segment-based ingest, the latest low-delay extensions to MPEG DASH, and possible uses of MV-HEVC video codec and the QUIC protocol.

The goal of the proof of concept (PoC) is to explore an end-to-end system comprising an encoder, a latency-optimised origin, and a mix of open source and proprietary DASH players. In addition, the innovation would be the achievement of sub-2s glass-to-glass delay with standard adaptive streaming, which will allow premium sports experiences.

Watch now IBC Accelerator project achieves low-latency streaming for premium sports

POC Objectives

The technical solution for the project aims to achieve (at high level):

• Social media equivalent end-to-end delay

• 1 sec inspirational

• 2 sec acceptable

• Betting-grade synchronization

• -45ms … 125 msec inspirational

• ± 250 msec acceptable

• Good performance in lossy WiFi networks

• Does not fall apart at and above 70dBm

• Compatibility with current HAS infrastructure

• Full compatibility: DASH / LL-DASH

• Segment-level compatibility: HLS / LL-HLS

• Support for non-conditioned advert insertion

The POC focuses on several key areas:

• Encoding Efficiency: Low-latency encoding techniques reduce the delay in processing and delivering the video feed to the viewers.

• Segment-Based Ingest: This approach enables faster video delivery by breaking the video stream into smaller, manageable segments that can be transmitted quickly.

• QUIC Protocol: A transport protocol designed for low-latency internet connections, QUIC optimises video delivery by reducing the overhead of traditional protocols like TCP.

By addressing these areas, the project seeks to provide a scalable, cost-effective solution for low latency streaming that is compatible with existing HTTP Adaptive Streaming (HAS) infrastructure, including DASH and HLS (HTTP Live Streaming).

Project goals

“The main goal was reducing the latency and improving the responsiveness of streaming live sports events at scale,” says project leader Alex Giladi, Fellow in Advanced Technologies at Comcast. “There is a desire to provide as good or better latency and responsiveness experience than current terrestrial broadcast and traditional cable systems in a streaming ecosystem to improve fan experiences.”

Low-latency streaming is not without its challenges. Achieving synchronization between multiple streams, ensuring good performance on lossy Wi-Fi networks, and maintaining compatibility with existing infrastructures are just a few of the hurdles broadcasters must overcome. The project addresses these challenges head-on by:

• Ensuring synchronization: For applications like live sports betting, synchronization between streams is crucial. The project aims for synchronization within 250ms, ensuring that viewers in the same room or venue receive the same feed, without any noticeable lag between different screens.

• Handling wi-fi performance: Many viewers stream sports content over wi-fi, which can be unreliable, especially in crowded areas. By optimising the stream for networks with signal strength as low as -70dBm, the project ensures that the viewing experience remains smooth and uninterrupted, even in less-than-ideal conditions.

• Cost-effective implementation: One of the project’s primary goals is to develop a solution that is not only effective but also economical. While other low-latency solutions like WebRTC and MoQ (Media over QUIC) require a complete infrastructure overhaul, this project aims to deliver comparable performance using standard HTTP-based technologies. This makes it an attractive option for broadcasters looking to upgrade their streaming capabilities without significant investment in new infrastructure.

The project seeks to deliver a latency experience comparable to that of social media, aiming for sub-two-second glass-to-glass delay. The goal is to ensure that “you don’t get tweets telling you the goal was scored before you actually see it,” says Giladi. This focus on synchronization, particularly for premium sports content, ensures fans can enjoy real-time engagement without lag, even in scenarios like live sports betting.

Problem

“The project is particularly geared towards live premium sports betting,” says Giladi, where even a millisecond of delay can impact user experience and decision-making. He also points out that “some aspects are applicable to any multi-channel or multi-viewer linear streaming use case and will improve viewers’ experience.”

Another challenge addressed is the unreliability of wi-fi networks. “Most of our customers have less-than-ideal wireless signal,” he explains, which can significantly impact the viewing experience. This project seeks to overcome those challenges by ensuring good performance even on lossy networks.

Unique approach and technological integration

One of the major points of differentiation is the project’s focus on using open standards and aiming for a cost-effective solution. “The project tries to build the cheapest solution possible without using proprietary technology,” says Giladi. “There are multiple solutions allowing sub-2s delivery. Most of them are based on non-HTTP protocols such as WebRTC and Media over QUIC (MoQ),” he says. “These protocols are better optimised for lower latencies then traditional HTTP streaming but require a wholesale change of infrastructure. Hence, they are a fairly expensive proposition. This project is trying to get to a solution that will be orders of magnitude cheaper and ‘good enough’ for the purpose.”

The key to this cost-effectiveness is its reliance on HTTP-based adaptive streaming protocols like MPEG DASH and HLS, which are already widely used and do not require overhauling the entire streaming ecosystem.

Additionally, the project focuses on combining a range of existing and emerging technologies, rather than pushing for an entirely new system. “We are trying to combine a basket of technologies,” Giladi explains, “and trying to use them to provide the best economically achievable premium sports experience.” This blend of open standards and cutting-edge technology makes the project not only unique but also highly practical for large-scale deployment.

Technical Innovation

One of the project’s standout features is its focus on utilising open standards and open-source technologies, such as the widely used Shaka player. However, this isn’t merely an open-source exercise. The project also supports commercial product integration, ensuring that broadcasters can deploy these innovations on a large scale without overhauling their entire infrastructure.

Key innovations include:

1. L3D DASH: L3D DASH (extensions to the MPEG DASH standard, coming in 2025 in its 6th edition). It allows fast video start (sub-120ms in the project’s implementation) as well as a low-latency operation mode compatible with LL-HLS.

2. DASH-IF Live Ingest: By bypassing the traditional contribution encoding process, the project reduces latency by uploading encoded segments directly to the origin server. This eliminates the multiplexing delays inherent in MPEG-2 TS (Transport Stream) systems, which are still commonly used in streaming workflows today.

The most significant technological innovation is the use of L3D DASH — an extension to the MPEG DASH standard that allows for ultra-low-latency operation. “L3D DASH allows fast video start and low-latency operation mode,” explains Giladi. “This extension is particularly important for maintaining synchronization between multiple streams, ensuring that viewers receive a consistent and smooth experience across different devices and networks.”

Watch now Accelerator session - Scalable Ultra-Low Latency Streaming for Premium Sports

The use of DASH-IF Live Ingest, reduces latency by bypassing traditional contribution encoding processes. “Distribution encoding is done at a point closest to production,” notes Giladi, adding that this step eliminates the multiplexing delay that is typical in MPEG-2 TS systems.

Additionally, the project harnesses the power of the QUIC protocol and HTTP/3, which improve the efficiency of video delivery by allowing parallel stream processing. Unlike traditional HTTP/2, which processes requests sequentially, HTTP/3 allows for independent, parallel processing of streams. This eliminates the head-of-line blocking problem, where a delay in one stream can hold up the entire process.

Both L4S (Low Latency Low Loss Scalable Throughput) and QB (Queue-Bounded Delay) also play a role in reducing latency, particularly in networks where queuing delays can introduce hundreds of milliseconds of delay. L4S allows for faster queuing of real-time traffic, reducing the chances of delay caused by network congestion. QB, on the other hand, is designed for real-time traffic that operates at low bitrates. Giladi highlights the importance of these IP-level technologies, stating, “When you are trying to provide a low-latency service, this actually matters.”

Showcase at IBC2024

The project showcased its work at IBC2024 with a live demonstration of its ultra-low latency streaming system. As Mickaël Raulet, CTO at Ateme, a participant in the project, previously explained, the team simulated an “end-to-end delivery of a live sports event over a 5G network using product-grade technology.” The demonstration relied on L3D DASH and a single-gen architecture to showcase just how effective the project’s innovations are in a real-world setting.

The IBC demonstration provided a tangible example of how these technologies work in harmony to reduce latency and improve synchronisation in live sports broadcasting. The team will discuss the experimental results and explore further directions for the technology post-IBC. Giladi sums it up well, stating that the ultimate goal is to “take newer technologies that have come into the mainstream over the last few years and see how we can improve what we’re already providing.”

Scalable Ultra-Low Latency Streaming for Premium Sports

Champions

Associated Press

BT Media & Broadcast

Comcast

Globo

Participants

AMD

HP

Ateme

Castlabs

Videoclarity