Designing a High‑Availability Comment System for QQ Music: Architecture, Challenges, and Optimizations

Since its launch, QQ Music has iterated its comment service several times. The latest version (released in 2019) stores comments in a tlist structure and displays them in chronological order. To support a new "comment‑threading" feature, the storage was migrated to MongoDB.

Commenting is a critical social scenario, especially during "artist landing" events where sudden traffic spikes occur. The system must handle massive read‑write pressure, including high‑frequency comment listings and write operations such as posting comments and pinning artist comments.

Key challenges include:

▶︎ Protecting the system against extreme read/write bursts during celebrity events. ▶︎ Ensuring data consistency when multiple operations may fail. ▶︎ Balancing strong consistency (transactions) with high throughput (eventual consistency).

Core design decisions :

▶︎ Use a cache layer for fast reads; fall back to MongoDB for persistence. ▶︎ Decouple write paths with a high‑speed cache and asynchronous MongoDB writes via a message queue, enabling retries and guaranteeing eventual consistency. ▶︎ Split read and write services to avoid interference and allow independent scaling. ▶︎ Introduce priority queues for artist‑centric writes to prevent blocking. ▶︎ Apply rate‑limiting on non‑critical paths to protect core services.

Scalability improvements :

▶︎ Increase MongoDB CPU cores and parallel consumption to boost write throughput. ▶︎ Parallelize read services and shard cache keys (e.g., using uin%10) to disperse hot keys. ▶︎ Migrate comment count and like count from CKV to CKV+ without cold‑storage, adding local cache with versioning. ▶︎ Front‑end safeguards: avoid exposing errors when comment counts or lists fail, and optimize page load times.

Reliability enhancements :

▶︎ Separate global and hot‑comment message queues; hot‑queue congestion only affects a subset of traffic. ▶︎ Conduct regular read/write stress tests to identify bottlenecks and enable rapid horizontal scaling. ▶︎ Monitor and limit usage of cold‑storage (CKV) during peak loads to prevent platform‑wide alerts.

Through these architectural refinements and operational practices, the QQ Music comment system achieves high availability and can sustain the growing demand of celebrity‑driven traffic spikes while maintaining a smooth user experience.