Didi's National Carpool Day: Technical Insights into Stability Assurance

Didi held its National Carpool Day event on December 3, 2019, offering rides at a 1‑fold discount across 26 cities, attracting 3.1 million passengers, 680 thousand first‑time users and generating 1.952 million shared seats.

The article details how Didi’s operations team ensured service stability through six pillars: team organization, capacity management, monitoring and alerting, contingency planning, process standards, and third‑party cooperation.

In the team organization phase, relevant business owners were identified early, a kickoff meeting was held, and a stability‑guarantee task force was formed to synchronize progress via regular meetings and weekly reports.

Capacity management began with business‑volume estimation using pilot cities to build a predictive model, followed by service scaling (leveraging Didi’s Docker‑Kubernetes container platform that can expand nearly 60 000 containers within 5‑10 minutes) and capacity pressure testing with a dedicated full‑link pressure‑test model that was refined through multiple trial runs.

Monitoring and alerting combined basic resource metrics, system‑level RPC data, business‑level traffic, and a custom “fire‑fighting map” that aggregates core scenario cards; a monitoring credit mechanism quantified coverage and accuracy to expose blind spots.

Contingency planning focused on completeness, validity and execution efficiency; a pre‑built platform integrated flow‑cutting, throttling and circuit‑breaking capabilities, and regular drills validated limit‑flow/degradation and multi‑active switchover procedures.

Process standards included pre‑event risk inspections, change freezes during the event window, on‑site duty shifts for rapid fault response, and post‑mortem reviews to feed improvements back into the stability system.

Finally, Didi synchronized activity information with third‑party partners to give them time for capacity preparation and to minimize disruptive changes during the peak period.