Catalogue / netflix-2023-marken-annotations

Netflix builds Marken on Cassandra, Elasticsearch, and Iceberg to store and search 1.9 billion media annotations

Netflix. Marken generic annotation service.

Organisation
Netflix
Evidence
Reported from production
Source
Scalable Annotation Service — Marken

Netflix Technology Blog, published 2023-01-26

Records

Decision#d1

Combine Cassandra source of truth, rollover Elasticsearch indices, and Iceberg tables for media annotations

Decision#d2

Isolate bulk ML catalog ingestion onto a separate fleet throttled with Java thread pools

Decision#d3

Enforce versioned JSON schemas with multiple inheritance and immutable versioned annotations

Sign in to read what was chosen, what was rejected and what it cost, each with its place in the source. An account is free.