Introduction
Actionbase is a database for serving user interactions. It is designed to serve interaction-driven data in real time and is production-proven across Kakao services.
Overview
Section titled “Overview”Actionbase is built for high-throughput, low-latency OLTP workloads where user interactions are continuously written and queried. Actionbase focuses on serving interaction-derived data—such as recent views, likes, reactions, and follows—that power product listings, recommendations, feeds, and other interaction-driven surfaces in large-scale services.
User interactions naturally form actor→target relationships with associated properties. Actionbase models these relationships using a graph data model and materializes read-optimized structures at write time, enabling fast and predictable queries without expensive read-time computation.
When backed by HBase, Actionbase inherits strong durability and horizontal scalability, and provides a higher-level abstraction tailored for real-time interaction serving.
Design Goals
Section titled “Design Goals”-
Shared Interaction Layer Provide a unified platform for storing and serving user interactions as a shared, canonical interaction graph.
-
Natural Interaction Modeling Model interactions as actor→target relationships with schema-defined properties, closely reflecting how user interactions appear in real applications.
-
Write-Time Optimization Optimize for reads at write time, enabling fast and predictable queries without expensive read-time computation.
-
Leverage Proven Storage Build on the strengths of existing storage engines (for example, HBase), allowing Actionbase to focus on interaction modeling and serving rather than reimplementing durability, scalability, or distribution.
Key Features
Section titled “Key Features”-
Write-Time Materialization Pre-compute the data required for fast, predictable reads at write time, eliminating service-specific indexing and counting logic.
-
Interaction-Oriented Graph Model Represent user interactions as actor→target edges with schema-defined properties for efficient querying.
-
Unified REST API Expose a simple API for querying and mutating interaction-driven data.
-
WAL / CDC Integration Emit write-ahead logs and change data capture streams for recovery, replay, asynchronous processing, and downstream data pipelines.
Datastore
Section titled “Datastore”Actionbase currently uses HBase as its primary storage backend, leveraging its durability and horizontal scalability. HBase is suited for large-scale deployments. SlateDB is planned as an additional storage backend for small to mid-size deployments in future releases.
Production Usage
Section titled “Production Usage”Actionbase is used across Kakao services—including KakaoTalk and KakaoShopping—to power real-time user interaction serving at scale. It has been running in stable production for years, delivering predictable reads, consistent writes, and reliable handling of multi-terabyte datasets.