warehouses, graphs, and what we would undo
Data platforms
Warehouses, graph data, and the retros from running them. I ran identity data products at HealthVerity and built Graphite Atlas on a graph database. These are the retros and notes from working with data at scale.
All 1 articles
Common questions
Is Memgraph faster than Neo4j?
Memgraph keeps the whole graph in RAM, and Graphite Atlas moved to it after Neo4j traversal latency hit early. The tradeoff is that memory drives both hardware cost and Enterprise licensing.
How much does Memgraph Enterprise cost compared to Neo4j?
Memgraph Enterprise starts at $25,000 a year for 16GB, metered by memory, and you supply the hardware. Neo4j AuraDB is $65/GB/month on Professional and $146/GB/month on Business Critical, so 16GB for a year is about $12,480 or $28,032. Self-managed Neo4j Enterprise is priced through sales.
What should I check when migrating from Neo4j to Memgraph?
Memgraph doesn't support relationship uniqueness, existence or type constraints, doesn't create indexes automatically, and drops constraint names with a warning. Audit any APOC and GDS usage, and learn how snapshots and startup recovery behave before you rely on them in production.
Do I need a graph database or can I just use Postgres?
Use Postgres alone when traversals are shallow or occasional, with recursive CTEs or Apache AGE if you want Cypher. Add a separate graph engine only when deep, frequent traversals are a product feature and users feel the latency.