How Netflix Delivers Key-Value and Time-Series Storage at Any Scale

Explore how Netflix achieves scalable key-value and time-series storage using Apache Cassandra in this 34-minute conference talk by Joseph Lynch and Vidhya Arvind. Learn about innovative techniques for designing reliable APIs and optimizing data layout to handle petabyte-scale datasets. Discover a novel dynamic bucketing approach for managing large partitions, ensuring fast access for small data and linearly scalable latency for large values. Examine strategies for time-series data organization, including table sharding, time and random bucketing, which automatically split large partitions while maintaining aggressive latency goals. Understand how Netflix's approach to data expiration can increase storage efficiency by up to 2x. Gain insights into combining fully-idempotent APIs, advanced table layouts, bucketing algorithms, and compression schemes to dramatically scale Apache Cassandra usage beyond conventional limits.