What's New in Apache Cloudberry

Apache Cloudberry is built on the open-source lineage of Greenplum Database and is highly backward-compatible with Greenplum. If you are coming from Greenplum, you will feel right at home — most SQL syntax, utilities, and operational workflows work the same way.

At the same time, Apache Cloudberry brings many new capabilities on top of that foundation. It integrates a newer PostgreSQL kernel and delivers features across storage, query optimization, security, and more. This page highlights what's new.

Modern PostgreSQL kernel

Apache Cloudberry tracks PostgreSQL upstream closely across multiple branches:

Branch	PostgreSQL kernel	Status
main	PostgreSQL 16.x	Active development
REL_2_STABLE	PostgreSQL 14.x	Latest stable release, with ongoing minor version upgrades; a newer kernel is planned for upcoming 2.x releases

This kernel upgrade brings a broad range of upstream PostgreSQL capabilities that were unavailable in Greenplum — from multiranges, JSONB subscripting, LZ4 TOAST compression, and incremental sort, to REINDEX CONCURRENTLY, BRIN indexes, hash partitioning, UPSERT support, and much more. You get these out of the box, just as you would with upstream PostgreSQL, all while keeping the MPP architecture you rely on.

Key features in Apache Cloudberry

These are new capabilities brought by the Apache Cloudberry community.

Storage & data formats

Feature	Description
Directory Table	A table type for unified management of unstructured data on local or object storage. Use simple SQL statements to manage and analyze unstructured data objects.
PAX Storage Format	A hybrid row-column storage format that combines batch write performance of row-based storage with read performance of columnar storage, improving cache efficiency. Supports MVCC, z-ordering, bloom filters, sparse filtering, and multiple encoding and compression options. Adapts to both cloud object storage and traditional file systems.

Query performance & optimization

Feature	Description
Incremental Materialized Views	A special form of materialized view that only refreshes changed data — not the entire view — when base tables are updated. Orders of magnitude faster than regular views for large datasets.
Automatic Materialized View Rewrite	During query planning, Apache Cloudberry can automatically use incremental materialized views to answer some or all of a query (AQUMV). Especially effective for aggregation queries on large tables and complex calculations, greatly reducing query processing time.
Dynamic Tables	Database objects similar to materialized views that refresh automatically on a schedule. Useful for accelerating queries in lakehouse setups and processing data pipelines without manual effort.
Parallel Query Execution	On a per-segment level, a single query can use multiple CPU cores to scan heap and AO/AOCO tables in parallel, boosting performance on deployments with a small number of segments per host.
RuntimeFilter for Join	Uses Bloom filters to pre-filter large tables before HashJoin, greatly reducing memory access costs. In suitable scenarios, RuntimeFilter can double join performance.
IndexScan on AO Tables	Enables IndexScan on Append-Optimized tables, greatly reducing the amount of data scanned for queries with ORDER BY and LIMIT clauses. In the most suitable scenario, queries that originally required tens of minutes can be returned within 1 second.
Unique Index on AO Tables	Supports unique indexes and unique constraints on AO/AOCS tables, enforcing data uniqueness at the storage engine level.

Security

Feature	Description
Transparent Data Encryption (TDE)	Data files are encrypted at rest transparently. Supports AES and SM4 algorithms with a two-tier key management architecture. No application changes needed.
Row-Level Security Policy	Table owners can define fine-grained access policies that filter rows per user.
Password Policy	Enforce password complexity, expiration, lockout-after-failures, and reuse rules through named profiles that can be bound to one or more database users.

Deployment flexibility

Feature	Description
Single Node Deployment	Deploy Apache Cloudberry without segment nodes for development, testing, or light production use. Leverage the full Cloudberry feature set in a compact, PostgreSQL-like footprint.
ARM Architecture Support	In addition to x86_64, Apache Cloudberry fully supports ARM/AARCH64 architectures, enabling deployment on a broader range of hardware including cloud instances and ARM-based servers.

Operations & utilities

Feature	Description
gpshrink	Scale in a running cluster by removing segments from underutilized hosts, saving resources when cluster capacity exceeds demand.
gpdemo	Deploy a fully functional multi-segment cluster with a single command — ideal for development, testing, and demos without the complexity of a full production deployment.

Open governance with The Apache Way

All of the above is made possible by the project's open, community-driven governance. Apache Cloudberry is developed under the Apache Software Foundation, following The Apache Way — a proven governance model that ensures the project is vendor-neutral and community-driven. The contributor base spans multiple organizations and independent developers, and decisions are made openly on public mailing lists, not behind closed doors.

The original Greenplum Database followed a single-vendor governance model and was eventually closed as a proprietary product. Apache Cloudberry belongs to the community — no single company can shut it down or take it private.

See our community team page for the full list of contributors and committers.

Summary

Apache Cloudberry is not a fork frozen in time — it is an actively developed, modernized MPP database that preserves Greenplum compatibility while pushing forward with a newer PostgreSQL kernel, new storage engines, smarter query optimization, stronger security, and flexible deployment options. Whether you are a Greenplum user evaluating a migration or a new user looking for a capable open-source MPP database, Apache Cloudberry offers a compelling path.