How Netflix optimizes use of Apache Cassandra® for massive scale

Join Netflix engineers Vidhya Arvind and Rajasekhar Ummadisetty as they share their multi-year journey implementing abstraction layers for Apache Cassandra® at massive scale. Learn how Netflix transformed their data access patterns to support thousands of use cases while maintaining performance and reliability.

Key Highlights:

Evolution Timeline (2020-2024)
Started with basic V1 API supporting simple puts/deletes/gets
Scaled from 20 production use cases to 3500+
Development of automated shard provisioning and control plane rewrite
Implementation of V2 API with advanced features like chunking and scans

Scale Metrics
Managing 391 shards across the platform
Handling 8M QPS through abstractions
Largest Cassandra cluster operating at 1.8M reads and 4.9M writes per second
83% of all database access now through abstraction layers

Technical Deep Dive: Tombstone Management
Implementation of key-value abstraction layer
Novel approaches to tombstone spreading and SLA-based limiting
Performance optimization strategies and tradeoffs
Real-world solutions for handling deletions at scale

Perfect for engineering teams looking to:
Scale Cassandra deployments efficiently
Implement robust abstraction layers
Manage complex database operations at enterprise scale
Optimize tombstone handling in large deployments

#ApacheCassandra #DatabaseEngineering #Netflix #DistributedSystems #TechnicalArchitecture