Inside Baidu’s Epic Spring Festival Red‑Envelope Operation: How 100 000 Servers Powered a Nation‑Wide Live Event

Background

Since 2015, Chinese New Year gala (Spring Festival Gala) viewers have expected a red‑envelope giveaway, turning the broadcast into a data‑traffic tsunami. Baidu decided to seize this opportunity in 2019, announcing a secret "1218 reform" that would unify all core infrastructure under a new "TG" system.

Task Force Formation

Li Yanhong’s internal letter sparked the creation of a cross‑departmental task force led by veteran cloud architect Hou Zhenyu. Key members included operations lead He Feng, performance guru Chen Xiyang, network specialist Song Lei, procurement head Zhang Jiajun, and app experts Wu Yongwei and Wang Mao.

Key Tasks

Precisely calculate expected app traffic and required bandwidth.

Draft a detailed "spring‑gala script" for seamless system hand‑over.

Procure and deploy up to 100 000 servers nationwide.

Upgrade IDC, CDN, and network resources.

Optimize Baidu App connections from 100 to 3 per device.

Strengthen the login (Passport) system for a 2 500‑fold surge.

Server Procurement and Deployment

Engineers estimated a peak of 5 000 000 000 requests per second, requiring roughly 100 000 servers. Baidu coordinated with hardware vendors, accelerated production cycles, and shipped servers in record‑breaking time, even moving entire server cabinets as single units to data centers.

Performance Optimization

Chen Xiyang dissected Baidu App, reducing external connections and trimming data transfer to the theoretical minimum. The login system was reinforced to handle up to 150 000 000 logins per second, and SMS providers were pre‑booked to avoid bottlenecks.

Testing and Execution

Multiple full‑link joint tests simulated the 5 000 000 000 RPS load. A "blue‑team" deliberately injected failures to validate rapid response procedures. On the eve of the gala, a pre‑launch test generated 880 000 RPS, confirming readiness.

Live Event

During the broadcast, Baidu’s red‑envelope system took over from the advertising platform, handling a peak of 5 000 000 000 requests per second without service interruption. The operation involved over 1 000 engineers on site, 100 000 servers, and coordinated effort across three major telecom operators.

Outcome and Reflections

The successful rollout proved Baidu’s cloud and operations capabilities, turning a cultural tradition into a technical triumph. Engineers described the experience as a "life‑changing battle" that validated years of perseverance and highlighted the importance of meticulous planning, cross‑team collaboration, and relentless execution.