Performance Optimization Practices for Meituan's Overseas Business

This article, based on the 16th Meituan-Dianping technical salon, shares practical performance‑optimization experiences for Meituan Travel’s overseas services.

Performance Overview – User experience depends heavily on page load speed. Key metrics include first‑paint, white‑screen time, time‑to‑interactive, and full load for web; FPS, crash rate, memory usage for mobile; and response time, TPS, concurrency for backend.

Factors Affecting Performance – Complex product logic, poor network conditions (especially abroad), code quality, database bottlenecks, mobile environment diversity, and hardware/cloud constraints.

Network Optimization

1. DNS Issues : Slow or hijacked DNS adds significant latency. Meituan uses HttpDNS and DNS prefetch (e.g.,

<meta http-equiv="x-dns-prefetch-control" content="on">

) to bypass local DNS.

2. CDN Acceleration : Deploy overseas CDN nodes, pre‑warm caches, and use a Hong Kong intermediate source to improve hit rates.

3. Long‑Connection (Shark) : A proxy‑long‑connection service that selects optimal nodes, maintains TCP connections, and forwards HTTP traffic over dedicated lines, improving success rates for native API calls.

4. Link Optimization : Use dedicated lines between Shanghai and Hong Kong data centers, and a Hong Kong Squid proxy for outbound API calls, dramatically reducing RTT for overseas partners.

5. Dynamic CDN Acceleration : Tested but not widely adopted due to limited overall gains.

Frontend Optimization

• Front‑back separation – static HTML served via CDN, APIs for data.

• Image handling – adaptive sizing, WebP compression, lazy loading.

• Domain consolidation – reduce DNS lookups and request concurrency limits.

• Offline support – AppCache (now replaced by Service Worker) and a custom offline‑package framework.

• Node.js server‑side rendering – combines Java‑side rendering speed with front‑end decoupling, improving first‑screen time.

Backend Optimization

• Hardware upgrades (SSD replacement) eliminated slow‑SQL alerts.

• Caching strategies – multi‑get, async reads, hit‑rate monitoring.

• Product‑logic improvements – limiting export ranges, debouncing UI actions.

• Serviceization – splitting services by business, load type, and client type.

• Asynchrony – thread‑pool tuning, NIO, async HTTP clients.

• Search‑engine usage – handling complex queries outside the DB.

Service Architecture – Dual‑region deployment with synchronized caches and databases, enabling near‑real‑time data consistency and reduced cross‑region latency.

Conclusion – Performance optimization is a systematic effort requiring collaboration among front‑end, back‑end, and SRE teams. Leveraging high‑performance frameworks, BGP networks, cloud platforms, and robust operations has yielded measurable improvements for Meituan’s overseas services.