How Yelp modernized its data infrastructure with a streaming lakehouse on AWS

This article describes how Yelp modernized its data infrastructure by implementing a streaming lakehouse (streamhouse) architecture using AWS services and open-source technologies.

• Yelp reduced analytics data latencies from 18 hours to minutes using Apache Paimon and S3
• Replaced proprietary CDC format with industry-standard Debezium format for better compatibility
• Decoupled ingestion from storage, reducing storage costs by over 80% versus Kafka-based approach
• Migrated from custom tools to Apache Flink, Paimon, Amazon MSK, and S3 for simplified architecture
• Implemented phased, per-use-case rollout strategy to minimize risk during migration
• Enabled SQL-based access to streaming data, democratizing real-time analytics across teams
• Achieved automatic schema evolution, time travel queries, and built-in data management features
• Integrated AWS Lake Formation for fine-grained access control and governance

Yelp's streamhouse architecture successfully balanced real-time processing needs with cost efficiency by treating streaming as a first-class citizen while leveraging cloud storage economics.