Property Sharding: I Want to Shard My Database, Now What?
summary
Explains a working Neo4j Property Sharding system called Infinraph that splits a single topology onto one graph shard and moves all node and relationship attributes into multiple property shards that serve properties in parallel. Topology must still fit on one server, writes commit on the graph shard with ACID and property shards asynchronously pulling the transaction log, reads batch IDs across shards in parallel, creation uses a property_shard_count flag, preview release 2025.101, CDC and property based access control are not yet supported and large MERGE batches can be slow.
AI-written summary · about
chapters (8)
- 0:00Introduction to sharding
- 2:36Why shard your database
- 4:21Architecture overview
- 8:15Reading and writing data
- 11:04Creating a sharded database
- 14:35Querying and limitations
- 17:33Admin commands changes
- 21:25Try it out and resources
related talks
23mInfinigraph: Vector Data at Scale
23mSharding for hyperscale and geographically distributed relational databases
56mSharding: patterns and antipatterns
45mThe Future of Postgres Sharding
45mThe Future of Postgres Sharding
32mA New Sharding Architecture for MongoDB
33mFive Sharding Data Models and Which is Right
47mSharding in MongoDB