Scaling Muse: How Netflix Powers Data-Driven Creative Insights at Trillion-Row Scale
Netflix's Muse application evolved its architecture to enhance data-driven insights for promotional media, leveraging technologies like HyperLogLog sketches, Hollow feeds, and Druid optimizations to handle complex querying and massive datasets efficiently.
MAIN POINTS
- Muse helps Netflix strategists by providing data insights on effective promotional media.
- The architecture evolved to include React, GraphQL, and Spring Boot GRPC microservices.
- HyperLogLog sketches and Hollow feeds improved performance and data handling.
- Druid optimizations reduced query latencies by approximately 50%.
TAKEAWAYS
- Muse's architecture changes support advanced filtering and grouping capabilities.
- HyperLogLog sketches offer a balance between performance and accuracy for distinct counts.
- Hollow feeds enable efficient in-memory data storage and retrieval.
- Druid optimizations, including broker count and segment size tuning, enhance query throughput.