Scaling and Performance Optimization of Taobao Shopping Cart

Taobao's shopping cart has undergone a series of technical upgrades to handle increasing user volume and transaction spikes, focusing on capacity expansion and performance optimization.

Key improvements include expanding the cart item limit from 200 to 380, addressing storage bottlenecks that double DB size and increase I/O latency, and mitigating the surge in real‑time computation required for cross‑store promotions.

To solve the database challenge, the team introduced the cloud‑native in‑memory database tairsql as a read‑only replica, achieving read‑write separation and reducing query latency during peak traffic.

Additional optimizations involve pre‑computing promotion eligibility, compressing network payloads, caching frequently accessed data, and redesigning the protocol to eliminate unnecessary state data from upstream packets.

The system also adopted a streaming API model (response streaming) to split large response packets, allowing the client to render immediately while state data is delivered later.

Finally, a parallelization refactor using Java's ForkJoinPool was applied to process per‑item tasks concurrently, dramatically improving end‑to‑end latency for carts with many items.