How to Ensure System Stability During Mega Sales Events like 618

1. Background

The 618 shopping festival generates about 10% of annual sales, with 2022 reaching 379.3 billion CNY, averaging 1.463 million CNY per minute. Such volume makes service stability critical.

2. Challenges

Key factors affecting stability include massive traffic spikes, huge data volume, complex promotional scenarios, long delivery chains, and low tolerance for errors.

Traffic magnitude: traffic can be dozens of times normal, amplifying small issues.

Data volume: billions of orders make queries challenging.

Complex scenarios: multiple promotions and platforms increase load.

Long delivery chain: many services combined reduce overall availability.

Low tolerance: users expect fast response and error‑free experience.

3. Stability Assurance

3.1 Application perspective

3.1.1 Service modularization

Deploy applications as independent units to reduce risk of whole‑service outage, simplify troubleshooting, and allow independent scaling.

3.1.2 Monitoring & alerting

Implement multi‑layer monitoring (middleware, RPC, method, machine, system, business, workflow) with appropriate granularity and sensitivity, and define clear alert handling procedures.

3.1.3 Log management

Standardize log format, level, output, archiving, and trace‑ID usage while avoiding redundant logs.

3.1.4 Fast failure

Configure thread‑pool timeouts, leverage middleware timeout features, and apply rate limiting to fail fast and protect resources.

3.1.5 Rate limiting

Set limits based on system capacity, validate thresholds through load testing, and prioritize critical services during extreme load.

3.1.6 Business degradation

Use defensive degradation strategies to preserve core functionality when resources are constrained.

3.2 Storage perspective

3.2.1 Database

Adopt master‑slave architecture, read‑write separation, appropriate transaction isolation, sharding, and slow‑query optimization.

3.2.2 Cache

Use primary‑replica setups, capacity scaling, hot‑key handling, and avoid large keys to maintain performance.

3.2.3 Elasticsearch

Deploy dual clusters, monitor slow requests, control write rates, and manage storage thresholds.

3.3 Operations perspective

Form a dedicated response team for rapid issue handling.

Conduct full‑scale rehearsal (stress testing) to simulate real traffic.

Freeze code releases during the event to reduce change‑induced failures.

Implement daily health checks and on‑call duty schedules.

Prepare emergency response plans with clear procedures.

4. Conclusion

The article analyzes the background and importance of preparing for large‑scale sales events, and outlines concrete technical and operational measures to ensure system stability.