Technical Architecture and Operational Practices of Alibaba's 2019 Double‑11 "Cat Evening" Live Show

The 2019 "Cat Evening" live show, part of Alibaba's Double‑11 shopping festival, was streamed simultaneously on Youku, Taobao, and Tmall apps, creating a multi‑screen, multi‑device interactive experience. The event featured large‑scale games such as a 4×8 m touch‑screen competition and real‑time score integration that influenced the program flow.

Technical goals were defined to support the business objectives of delivering visual spectacle, consumer benefits, and merchant sales. The goals were broken down into stability, user experience, full rights distribution, and zero financial loss, while also fostering team growth and system knowledge retention.

To ensure a fair and consistent experience across all platforms, the system required synchronized interaction timing, identical UI/UX, and uniform lottery windows. The solution involved a unified codebase deployed on Taobao, Tmall, and Youku, with a lightweight proxy on the Youku side and core services hosted in Alibaba's data centers.

Four mechanisms were introduced to guarantee fairness despite network latency and production delays:

Client‑server clock synchronization via CSN and wireless RPC gateway polling.

On‑site latency measurement devices to capture network and production delays.

Coordinated operation of program‑schedule events with hosts and directors.

Insertion of SEI messages into the live stream to align interaction start times on the client side.

High‑concurrency spikes generated by interactive rounds were mitigated using three main strategies: extensive full‑stack load testing, pre‑warming of services, and anti‑scraping rate limits. Additional optimizations included dynamic routing, peak‑shaving, and downstream protection.

Routing : By default all requests passed through the wireless RPC gateway, but dynamic routing ratios could switch traffic to CSN polling when gateway load approached limits.

Peak‑shaving & off‑peak strategies :

Separate public and private domain interactions in time to avoid simultaneous peaks.

Schedule high‑value “red‑packet rain” events away from hour marks.

Use middleware message channels for private‑domain interactions to reduce backend pressure.

Randomized request distribution and controlled win rates for red‑packet activities.

Pre‑fetch and disperse final‑prize queries to avoid concentrated DB hits.

Downstream protection : The prize‑distribution system relied on an external rights platform. To prevent overload, the team implemented business‑rule‑driven user segmentation, application‑level traffic splitting, and real‑time monitoring to trigger fallback plans, ensuring all prizes were awarded without financial loss.

During the event, a multi‑level contingency plan was executed for large‑screen failures, ranging from pre‑event rehearsals to hot‑standby machines and manual keyboard control, emphasizing the need for executable, rehearsed, and redundant emergency procedures.

Key takeaways for engineers participating in rare, high‑impact projects include:

Formulating clear technical objectives.

Developing technical product‑management skills to anticipate risks and ensure delivery.

Maintaining craftsmanship in performance and experience.

Contributing reusable tools, components, and organizational knowledge.

Continuous post‑mortem analysis from day one.

Broadening perspective beyond one’s specialty to learn from cross‑functional teams.

The presentation concluded with a call to join the upcoming technical salon for Double‑11 Cat Evening.