Scalaxy
Use cases

Use Cases

Where Scalaxy fits: workloads, deployment shapes, and the problems it solves.

Scalaxy is a graph database written in Common Lisp. A database stores property graphs queried with openCypher, replicates writes to configured peers, and serves an HTTP console from every node. The same process also exposes a key/value API.

Graph workloads

Graph records use the same ring, durability log, and replication path as key/value records. There is no separate graph service to operate:

  • Relationship queries: MATCH (a:Person)-[:KNOWS]->(b:Person), OPTIONAL MATCH, variable-length paths (-[:TRIP*1..3]->), named paths, and full aggregation over grouped results.
  • Entity stores with links: profiles, orders, devices, and other linked entities. CREATE, MERGE, SET, REMOVE, and DETACH DELETE operate on the same replicated ring.
  • Bulk analysis: the shipped NYC taxi benchmark loads 2.9M relationships in ~20 s on a laptop, so large scans are tractable on a single node and shard across a cluster like any other data.

Session and user data

Store sessions, profiles, and per-user state behind an API gateway. Keys such as user:42 are assigned to ring owners automatically. Configure SCALAXY_REPLICATE_TO to keep a synchronous copy on another node:

curl -X PUT http://scalaxy:8080/api/keys/user:42 -d '{"value":"{\"theme\":\"dark\"}"}'
curl http://scalaxy:8080/api/keys/user:42

Feature flags and configuration

Keep flags and runtime settings in one store. Prefix scans are available in the console, and values are shown as UTF-8 and hex.

Cache and hot storage tier

Use it as a shared cache when restart recovery matters. A write is appended to the durability log before acknowledgement. Evict entries with delete.

Edge and embedded deployments

The SBCL binary can run on small machines, single-board computers, and air-gapped networks. The HTTP console is served by the database process, so there is no frontend build to maintain.

Microservice shared state

For a small fleet of services, the key/value API provides shared state without a separate cache or database cluster:

Embedded modeThe in-process cluster API (make-cluster) is also usable as an embedded library. It runs a database inside your Lisp process.

Demos, labs, and education

The repository includes a test suite with 9,018 checks, an openCypher TCK runner, and a small binary protocol. make test exercises TCP and HTTP, replication, failover, crash recovery, and the Cypher engine end to end.

When to consider something else

  • Deep analytics over huge scans: aggregation materializes rows today. For terabyte-scale OLAP, use a columnar engine.
  • Multi-datacenter consensus: replication is synchronous and leader-based. For cross-region strong consistency, use a consensus system.