← Lessons

quiz vs the machine

Platinum1780

System Design

Live Streaming Architecture

Getting a live feed from a camera to millions of viewers with low delay.

6 min read · advanced · beat Platinum to climb

The Live Path

Live streaming pushes a continuous feed to viewers in near real time. The pipeline must ingest, transcode, and fan out segments fast while keeping delay low.

Stages

  • Capture and contribution send the feed from the source to an ingest server, often over a low latency protocol
  • Transcode produces multiple renditions for adaptive playback
  • Packaging chops the stream into segments and writes manifests
  • Delivery fans segments out through CDN edges to viewers

Latency Versus Scale

There is tension between low latency and massive scale.

  • Larger segments cache well and scale to millions but add delay
  • Smaller segments and chunked transfer cut delay but stress the CDN
  • Low latency variants stream partial segments as they encode

Reliability

  • Redundant ingest so a dropped contribution does not end the show
  • DVR windows let viewers rewind by keeping recent segments
  • Failover encoders avoid a single point of failure

Key idea

Live streaming ingests a feed, transcodes to renditions, packages segments, and fans out through CDN edges; segment size trades latency against scale, while redundant ingest and failover encoders keep the show running.

Check yourself

Answer to earn rating on the learn ladder.

1. What is the core tension in live streaming segment size?

2. What does a DVR window provide in a live pipeline?

3. Why use redundant ingest and failover encoders?