Live streaming is encoding and delivering video in real time as an event happens — sports, news, concerts, gaming, auctions. Unlike VOD, there is a tight latency budget and no second chance: a dropped frame or a stalled encoder during the moment that matters cannot be re-done. The pipeline runs continuously, and every stage must keep up with real time.
A live pipeline starts at ingest, where a contribution feed (RTMP, SRT, or increasingly WebRTC) enters the platform, then transcodes to the encoding ladder, packages into segments, and delivers through the CDN — all within seconds. The hard parts are the latency budget (glass-to-glass delay), absorbing the synchronized demand spike when a premiere or kickoff starts, and graceful failure, since you cannot pause reality to fix a bug.
Latency targets range from ~30s for standard HLS/DASH down to a few seconds with LL-HLS/LL-DASH, or sub-second with WebRTC for interactive use cases like betting and auctions.
