Stability Assurance Practices for the 2024 CCTV Spring Festival Gala Live Stream

Each year, the live broadcast of the CCTV Spring Festival Gala serves as a major test for all technical teams involved in iQIYI's live streaming pipeline. The event triggers a massive surge in service interface QPS and puts instant pressure on CDN and core data center bandwidth. Moreover, the stringent real‑time fault handling requirements demand high availability across every link in the streaming chain.

The entire live streaming pipeline can be roughly divided into four stages: signal encoding and segment processing, CDN distribution and origin pull, program request handling, and video segment download and playback. This article introduces the stability assurance practices applied to each of these stages for the 2024 CCTV Spring Festival Gala.

Signal Encoding and Segment Processing – The live streaming cloud handles encoding, RTMP transmission, segment packaging, and upload, ensuring smooth, clear, and stable streams. Upgrades include HDR support, HDR‑to‑SDR transcoding, 4K ultra‑HD real‑time encoding, and frame‑level enhancements, delivering unprecedented visual quality. Five signal sources (one primary and four backups) are employed to guarantee reliability, including satellite signals, CCTV 4K network feeds, and studio feeds.

CDN Distribution and Origin Pull – CDN acts as a massive distributed system to accelerate content delivery. The architecture defines three origin layers (L1, L2, PROXY) and multiple edge nodes (commercial CDN, COC, self‑built CDN). High‑availability measures include multi‑level origin redundancy, origin shield, multi‑CDN isolation, rate limiting, and bandwidth preparation. Specific actions such as reducing TTL, automated failover within two minutes, and dedicated backup links for COC are described.

Program Request Handling – The backend handles a massive QPS spike around the 8 p.m. start time, preparing resources up to ten times the normal peak. HA strategies involve multi‑cluster deployment (QLB, nginx, tomcat, database, middleware) with cache pre‑push on backup clusters, enabling service continuity even if primary clusters fail. Degradation plans ensure at least basic playback by using static program IDs.

Video Segment Download and Playback – To mitigate bandwidth pressure, P2P technology is employed alongside CDN. The P2P system consists of LCache (pulls from CDN), Inside (pulls from CDN/LCache and shares with peers), and Livenet (delivers to players). Measures such as disabling m3u8 dependencies, CRC‑based integrity checks with signed .pct files, and emergency shutdown procedures are implemented to prevent error propagation.

Additional topics covered include live stream quality enhancements (4K, 1080p50/60, HDR), stream management platform for regional and carrier‑based bitrate control, unified retry strategies across clients, large‑scale online stress testing, and comprehensive attack‑defense drills. The article concludes that the extensive preparation and preventive work for the 2024 Spring Festival Gala provide valuable references for future high‑concurrency live events.