Apache Cloudberry

Use it when

• You want an MPP warehouse you can run yourself, with no vendor contract, under open Apache governance.
• You are on Greenplum and Broadcom's 2024 closed-sourcing forces a decision; Cloudberry is where the community and the lineage continued.
• PostgreSQL compatibility matters; the PostgreSQL 14 kernel keeps common extensions, tools, and SQL working.
• Data must stay on your own hardware or cloud account.

Think twice when

• Nobody wants to operate a coordinator, segments, and mirrors; the managed cloud warehouses exist precisely to remove that work.
• Elasticity is the requirement; storage is coupled to segments, so scaling is a resize rather than a slider, unlike Snowflake or BigQuery.
• You need time travel, managed streaming ingest, or a vendor SLA; those arrive via ecosystem builds like SynxDB rather than the project.

How it runs

Self-hosted MPP in the coordinator-and-segments model: the coordinator plans and distributes queries, segment instances each own a data slice, and mirror segments cover failover. Storage options span heap, append-optimized row and column, and PAX; PXF and FDWs reach external data. Resource groups police CPU, memory, and I/O per workload.