Columnar MPP execution
A share-nothing, massively parallel engine built on the proven Apache Cloudberry (Incubating) architecture. Every node scans its own slice of the data; the planner fans work out, gathers results, and gets out of your way.
Product · SynxDB Cloud
A columnar MPP warehouse that runs on the AWS you already use.
SynxDB Cloud takes the share-nothing MPP heritage of Apache Cloudberry (Incubating), decouples storage and compute onto S3, and wraps it in the Postgres wire protocol. Spin up a cluster, point your tools at it, query at warehouse scale.
What's inside
A small number of big decisions.
Instead of a feature list, here are the design choices that shape the product.
Postgres-compatible wire protocol
Any driver that speaks Postgres, speaks SynxDB Cloud. psql, dbt, Metabase, your ORM — point them at a connection string and they just work.
Storage on S3 by default
Tables live on S3. Bring Parquet in place, query it alongside native tables, and pay S3 prices for cold data. No separate data lake to maintain.
Decoupled scaling
Compute and storage scale independently. Resize the cluster for a heavy Monday, shrink it back on Friday. No rebalancing windows, no downtime.
Vectorized query engine
Column-aware execution that stays in cache. Combined with PAX hybrid row/column storage, scans touch only the bytes your query actually needs.
Materialized views & HTAP
Dynamic materialized views for fast dashboards; hybrid transactional/analytical processing for the workloads that refuse to sit cleanly on one side.
Architecture
Compute and storage, each doing one job.
A cluster is a set of compute nodes that run queries. Data sits on S3 — durable, cheap, yours. Compute reads what it needs, caches hot data on local NVMe, and writes results back.
Because the two scale independently, you can add compute for a busy quarter without touching your data layout, and add storage without taking query capacity offline. The planner handles parallelism across every node in the cluster.
There's no new query language to learn. The Postgres planner runs the show, extended with MPP-aware optimizations for partitioning, join distribution, and columnar scans.
AWS integration, at a glance
- Compute
- Managed EC2 / EKS in your region
- Storage
- S3 with KMS-encrypted buckets
- Network
- VPC peering or PrivateLink to your existing VPCs
- Identity
- IAM roles, SSO via SAML / OIDC
- Observability
- CloudWatch metrics + structured logs to your destination
Runs in our account (managed) or yours (bring-your-own-VPC). Full feature parity either way.
Spin up a warehouse. Run a query. See for yourself.
Early access is open. A starter cluster is free while we're in preview — no credit card, no sales call.