Skip to content

Datastore

Actionbase builds on proven storage technologies, leveraging their strengths while adding value through graph modeling and write-time optimization. The datastore abstraction layer allows integration with different backends.

Actionbase abstracts storage operations through a minimal interface, enabling integration with various backends. By building on established technologies, Actionbase leverages their reliability, scalability, and performance while focusing on interaction-specific capabilities.

A datastore backend must support:

OperationDescription
getRetrieve value(s) by key(s)
deleteDelete a value by key
scanRange scan with prefix, start, stop, limit
checkAndMutateAtomic check-and-mutate for consistency
batchBatch mutations (optional, recommended)

This interface is intentionally minimal. Any backend implementing these operations can be integrated.

HBase is the production datastore backend for Actionbase. It’s a distributed, scalable NoSQL database built on HDFS.

CharacteristicDescription
Horizontal ScalabilityShards data across nodes, linear scaling
Strong DurabilityData replicated across nodes
Low-latency AccessOptimized for random reads and writes
Automatic Load BalancingEven distribution across cluster
Proven at ScaleUsed for petabyte-scale data

HBase requires expertise when used directly:

  • Row key design
  • Column family organization
  • Region splitting and compaction
  • Cluster management

Actionbase provides:

  • HBase’s durability, scale, and distribution
  • Simpler usability through high-level abstraction
  • Interaction-specific features (State/Index/Count)

For HBase configuration, see HBase Operations.

Memory is an in-memory backend for development and testing.

CharacteristicDescription
Easy SetupNo configuration required
No PersistenceData lost on server stop
FastIn-memory operations

Ideal for local development, testing, and prototyping.

SlateDB is a planned backend for small to medium-scale deployments.

CharacteristicDescription
Lower ComplexitySimpler setup and maintenance
Resource EfficiencyLess infrastructure overhead
Transactional SupportSuitable for smaller workloads

Actionbase maps its operations to datastore operations:

Actionbase OperationDatastore OperationData Structure
Get QuerygetEdgeState
Scan QueryscanEdgeIndex
Count QuerygetEdgeCounter
Mutation (lock)checkAndMutateLock
Mutation (write)batch / putState, Index, Counter
Mutation (cleanup)deleteOld indexes

By building on proven storage, Actionbase focuses on interaction-specific features while the underlying datastore handles persistence, replication, and scalability.

BackendUse CaseScale
HBaseProductionLarge
SlateDBProduction (planned)Small to Medium
MemoryDevelopmentLocal