Amazon Redshift

Redshift speaks the PostgreSQL wire protocol, so AddisDB connects with an ordinary username and password rather than a token. It is treated as a warehouse, though — it opens in the Big Data Console, with a catalog browser and cost estimation.

Who it is for

Redshift is AWS’s columnar warehouse, tuned for large aggregate scans and tightly integrated with the rest of AWS. Redshift Spectrum queries data sitting in S3 without loading it, and Serverless removes cluster sizing from the equation.

On a provisioned cluster two table properties decide most of your performance: the distribution style, which spreads rows across nodes, and the sort key, which lets the scan skip blocks. Getting those wrong shows up as a query that redistributes half the table on every join.

It fits teams already invested in AWS, workloads that read from S3 data lakes, and organizations that want a warehouse inside their existing VPC and IAM boundary.

Set up the cluster

1. In the AWS console, create a Redshift Serverless workgroup or a provisioned cluster.
2. Copy the endpoint host and note the port — Redshift uses 5439, not 5432.
3. Set the admin username and password, or create a dedicated read-only user.
4. Under Network and security, enable public accessibility, or plan to tunnel — a Redshift cluster in a private subnet is not reachable directly.
5. Add your IP to the associated security group’s inbound rules.

Connect from AddisDB

1. New Connection → Amazon Redshift under Big Data / Warehouse. Port prefills to 5439, database to dev and username to awsuser.
2. Enter the endpoint host and your credentials — Redshift uses a real password, not an API token.
3. Set SSL mode to require.
4. For a cluster in a private subnet, open Tunnel (optional) and add your bastion, or use AWS SSM if the instance is SSM-managed.
5. Test, then Save.

Where it stops behaving like Postgres

• No real foreign keys — constraints are declared for the planner and not enforced, so the schema diagram reflects intent rather than a guarantee.
• No unique or primary-key enforcement either, for the same reason.
• The system catalogs are Redshift’s own stv_, svv_ and svl_ tables, not pg_stat_*.
• DDL and function coverage tracks an old PostgreSQL fork, so a modern Postgres statement can be rejected outright.

-- How is this table distributed and sorted?
SELECT "column", type, distkey, sortkey
FROM   pg_table_def
WHERE  schemaname = 'public' AND tablename = 'orders';

What AddisDB gives you

• The Big Data Console — catalog browser, query pane, and a row cap on every preview.
• Cost estimation from the Redshift planner, so you can see the expected row count and width before running a scan.
• Full schema introspection with ⌘K search across every connected database.
• SSH and AWS SSM tunnels for private clusters.
• The Chart view, notebooks, and AI querying grounded in your real schema.