Skip to content

Introduction

Actionbase is a database for serving user interactions in real time, production-proven across Kakao services.

Follows, likes, recent views, related content — they are all interaction data. Every interaction is expressed in the same model — who did what to which target (source → action → target) — and the combination of source and target yields three axes:

  • User–User (U2U) — follow/unfollow, follower/following counts, timeline scans
  • User–Item (U2I) — likes/bookmarks, view history, bidirectional counters
  • Item–Item (I2I) — related products, similar-content graphs, and other precomputed item-to-item relations

Actionbase materializes read-optimized structures at write time. Reads use bounded access patterns (GET, COUNT, SCAN) without expensive computation.

When backed by HBase, Actionbase inherits durability and horizontal scalability.

Focuses onExplicitly avoids
Real-time interactions across U2U / U2I / I2IGeneral-purpose graph queries
Bounded access patterns (GET, COUNT, SCAN)Unbounded traversal or analytics
Continuous writes, immediate readsBatch ingestion or deferred indexing
WAL/CDC to Kafka (yours or ours)Owning downstream processing
Pluggable storage (HBase now, others planned)Building yet another storage engine

Actionbase currently uses HBase as its primary storage backend. Lighter backends are planned for smaller deployments.

Used at Kakao services—primarily KakaoTalk Gift—handling over a million requests per minute. Running in stable production for years.