Implementing a Redis-Based Flash Sale System to Prevent Overselling

The article addresses the challenge of preventing overselling during flash‑sale events with limited product inventory and handling massive concurrent user requests to improve user experience and system performance.

The overall solution reduces database access by caching data in Redis: during system initialization, product stock is loaded into Redis; when a flash‑sale request arrives, Redis pre‑decrements the stock, immediately returning failure if stock is insufficient, otherwise queuing the request for asynchronous processing.

An asynchronous queue (simulated with a Redis list) processes queued requests, creates orders, updates the database stock, and sends a success notification to the user via WebSocket, allowing the front‑end to display order details or failure messages.

The implementation steps include loading inventory into Redis, creating a Redis‑based message queue, configuring RedisTemplate serialization, and building an API that spawns many threads (e.g., 10,000) to simulate concurrent purchase attempts.

During testing with 10,000 concurrent threads each attempting to buy 10 items from an initial stock of 100, the system successfully prevented overselling, as only a limited number of threads proceeded to update the database after Redis stock depletion.

Images in the original article illustrate the initialization of Redis stock, the Redis queue setup, serialization configuration, and the testing results.