Solving the Order Service with an Asynchronous Request-Response Pattern

Oliveyoung's existing order system suffered from response delays and server memory occupation issues when processing large volumes of data due to its synchronous processing method. In particular, the process of running dozens of validation checks per product was identified as the bottleneck of the entire process.

To solve this, the request DTO structure was first improved from a 'single record' to a 'group-based' structure, minimizing duplicate validation. Next, the service architecture was converted to a Kafka-based asynchronous method, dramatically reducing user wait time by returning a response immediately upon request.

The reasons for choosing Kafka are as follows.

• Preventing message loss: Data durability is guaranteed through Replication and Log-based storage
• Automated reprocessing: Recovery from a specific point in time is possible in the event of a failure through offset adjustment
• Distributed processing: Improved processing performance through partition-based horizontal scaling

To solve the 'result delivery problem' that arises from asynchronous processing, the introduction of a request-response architecture utilizing Enterprise Integration Patterns (EIP) was reviewed.