Cloud Data Warehouse across clouds
The managed analytical database - Redshift, BigQuery, Synapse dedicated SQL - with the endpoint private, access by identity, and the idle cost understood.
9 verified modules, 2 of them live-tested apply→verify→destroy; the rest are static-validated, live-test pending.
Compare by provider
| Provider | Module | Verification |
|---|---|---|
| Alibaba Cloud | A MaxCompute Project That Refuses the Full Table Scan Which Writes the First Bill | static-validated |
| AWS | A Serverless Warehouse Whose Spend Limit Does More Than Log | static-validated |
| AWS | Redshift Cluster (encrypted, private) | ✓ live-tested |
| Azure | A Synapse Workspace with No SQL Login, a Managed Network and a Dedicated Pool Audited | static-validated |
| Google Cloud | BigQuery Dataset & Tables | ✓ live-tested |
| Google Cloud | Slot Reservations with a Ceiling, Assigned to the Projects that Use Them | static-validated |
| Huawei Cloud | A DWS Cluster Whose Snapshots Outlive It, Which the Default Does Not Arrange | static-validated |
| IBM Cloud | A Private Db2 Deployment with a Standby and an Allow List That Is Not Empty | static-validated |
| Tencent Cloud | A ClickHouse Cluster with a Second Replica and a Cold Tier Rather Than Neither | static-validated |
How to choose
The first decision is provisioned or serverless: Redshift and a Synapse dedicated pool bill by the hour while online and need pausing, while BigQuery bills per query (or per reserved slot) and idles at nothing. Then compare how access is granted (IAM, Entra, a SQL login you would rather not have), whether the endpoint can be private, and where the audit of who queried what ends up.
When not to use
A warehouse is a copy of everything sensitive the company owns, in one place, optimised for reading it fast. Row-level security and column masking are the controls that make a wide audience safe, and none of these turn them on by default.