Contribution Protocols for Live Streaming

Modern live-streaming deployments involve at least 2 stages of video encoding and transmission over IP networks. The first stage is encoding and transmission of a contribution feed from the event production site to a headend or cloud-based video distribution system. The second stage is the transcoding and distribution of the transcoded video content to end users. The technologies for second-stage transmissions have matured in the last two decades. Nowadays, they are commonly implemented by using HTTP-based streaming protocols, such as HLS and DASH, and are supported by content delivery networks (CDNs) at the infrastructure level. The adaptive bitrate mechanisms of HLS and DASH, along with CDNs, effectively address the scalability and reliability of end-user delivery. However, the technologies for supporting the first transmission stage, delivering the streams from the contribution encoder to the head-end, are more fragmented. Examples of such existing contribution protocols include the Real-Time Messaging Protocol (RTMP), the Real-Time Transport Protocol (RTP), the SMPTE 2022-1 and 2022-2 protocols, Secure Reliable Transport (SRT), Reliable Internet Stream Transport (RIST), DASH-IF ingest protocol, Zixi, and several others. The problem they address is also inherently more complex: the physical networks connecting to the production site (such as wireless, satellite, or terrestrial) can be highly diverse and bandwidth-limited. Furthermore, such protocols can only send a single stream, and the means for adapting to network conditions can be limited. In some cases, the transmission may only be one-way. To solve such problems, existing contribution protocols employ a variety of techniques, including long network adaptation buffers, Forward Error Correction (FEC) codes, and Automatic Repeat Request (ARQ). Some implementations are very simplistic, for example, by relying on TCP as an ARQ-based transport. Others are more complex, using UDP transport and implementing all essential algorithms within the contribution protocol logic. Some advanced implementations also include an encoder feedback loop and network-adaptive encoding, modulating the encoder’s bitrate or forcing the insertion of Instantaneous Decoder Refresh frames in response to network events. As a result, the features, performance, and applicability of the existing contribution protocols can vary across different use cases and deployment scenarios. The objective of this work item is to

• study the features of the existing contribution protocols and their existing implementations,
• conduct experiments to confirm the performance and features of the contribution protocols under study
• analyze the experimental results and
• produce a report informing the industry of our findings.

It is envisioned that such a report would serve as a valuable consolidated source of information, providing a comparative analysis of existing contribution protocols, their suitability for different possible applications, deployment scenarios, and related best practices.