Design and Optimization of the Overseas Hotel Backend System

The article presents a comprehensive case study of the overseas hotel backend system, a newly incubated business within Meituan’s hotel division, covering its lifecycle from initial construction to optimization and future planning.

Initial Build : Emphasizes research, reuse of existing services, and careful creation of new components using mature frameworks. It discusses criteria for reusing company‑level services (e.g., async messaging, cache, risk control) and department‑level platforms (order, payment, customer service).

Custom Service Development : Highlights challenges such as rapid requirement iteration, frequent changes, and fast system construction. The solution focuses on simplicity (clean project structure), flexibility (micro‑service boundaries), and extensibility (future‑proof design).

Technology Selection : Recommends mature frameworks and components to reduce risk, accelerate development, and ease hiring.

Optimization Phase addresses three core aspects: availability, performance, and scalability.

Availability : Implements disaster recovery, rate‑limiting, and circuit‑breaker strategies (Hystrix) for API stability. Manual downgrade via Meituan’s MCC (ZooKeeper‑backed config) is used, with plans for automated downgrade based on monitoring.

Rate Limiting : Uses both local (Java Semaphore, Guava RateLimiter) and distributed (Tair/Redis atomic counters) controls, with a company‑wide limit‑flow component for third‑party API quotas.

Service Backup : Deploys services across multiple data centers, replicates POI cache to both Redis and Tair, and adopts selective data synchronization to reduce maintenance overhead.

Monitoring : Splits monitoring into network, machine, business, and log dimensions, integrating with a unified alert platform. Future work aims to consolidate dashboards for faster incident resolution.

Performance : Focuses on concurrency, asynchronous decoupling, and caching.

Concurrency is achieved through thread‑pool isolation for parallel third‑party calls. Asynchronous decoupling is realized via new threads for long‑running tasks and message‑driven pipelines for data aggregation.

Caching strategies include local caches (Ehcache, Guava) for small, stable data and distributed caches (Redis/Tair) to offload database and third‑party traffic. Cache consistency is maintained through async updates and periodic full syncs.

Cache resilience techniques such as avalanche mitigation (staggered expiration) and penetration protection (storing empty results or using Bloom filters) are described.

Scalability : Covers capacity planning (PV, data volume, CPU/Memory) and service granularity. Services are designed for high cohesion and low coupling, with clear boundaries for POI cache, product center, order center, and pricing services.

Team Building : Discusses challenges of new, inexperienced teams and proposes gradual hiring, cultural onboarding, and the rapid establishment of technical standards. Eight technical specifications were rolled out within six months, improving engineering quality and onboarding speed.

The article concludes that systematic technical thinking—spanning reuse, simplicity, flexibility, performance, and scalability—enables the overseas hotel backend to evolve reliably as business demands grow.