The Architecture of Live Video Streaming: How HLS, DASH, and CDNs Actually Work

Live video streaming is the backbone of modern entertainment infrastructure, enabling users to watch everything from football matches to video conferences. However, the technology behind it often remains a mystery to developers, who simply point a video player at a URL and expect it to work seamlessly. The article below delves into the complete architecture of live video streaming, from ingest to last-mile delivery, explaining how a single origin server can deliver 4K video to millions of concurrent viewers without collapsing, how the player can switch between resolutions without buffering, and why the same stream may load in 200ms in Dublin but take 4 seconds in rural Kerry.

The streaming pipeline comprises five distinct stages:

1. Ingest: The raw video signal enters the pipeline via RTMP or SRT protocols. RTMP, originally developed by Macromedia and now officially deprecated, remains the de facto ingest protocol due to its universal encoder support. It operates over TCP port 1935, using a specific handshake sequence. SRT, a modern successor, operates over UDP with built-in encryption and adaptive bitrate recovery, making it more resilient over unreliable WAN links.

2. Transcoding: A single 4K ingest feed must be transcoded into multiple renditions to cater to the diverse range of client devices and network conditions. This is achieved through a set of parallel encodes, known as the ABR Ladder (Adaptive Bitrate Ladder). The FFmpeg transcoding pipeline is used to produce renditions with varying resolutions, bitrates, and codecs suitable for different devices and connection speeds.

3. Packaging: Once transcoded, the renditions are packaged into a streaming format that HTTP-based CDNs can cache and serve. The two dominant packaging protocols are HLS (HTTP Live Streaming) and DASH (Dynamic Adaptive Streaming over HTTP). HLS, developed by Apple, slices the continuous video stream into small, time-based segments and uses a plain-text manifest file (m3u8) as the index.

DASH, an open and vendor-neutral alternative, uses an XML-based Media Presentation Description (mpd) and exclusively uses MP4 containers, reducing overhead.

4. CDN: The packaged segments are delivered through a Content Delivery Network (CDN), which caches and serves the content from edge locations closer to the end-users. This reduces latency and improves the overall viewing experience.

5. Client: Finally, the player on the client device receives the streamed content and handles playback using libraries like ExoPlayer. The player seamlessly switches between resolutions based on the viewer's network conditions and device capabilities, ensuring a smooth and uninterrupted viewing experience.

The article provides code examples and protocol analysis for each stage, offering a comprehensive understanding of how live video streaming functions at a technical level.