Amazon Keyspaces

Amazon Keyspaces is a serverless, Cassandra-compatible database from AWS. It speaks CQL, so AddisDB connects with the same driver it uses for Cassandra and ScyllaDB — with two differences worth knowing up front: the port is 9142, not 9042, and TLS is mandatory.

Who it is for

Keyspaces suits teams who want Cassandra’s data model — wide rows, partition keys, predictable single-partition reads at scale — without running a ring. There are no nodes to size, repair or compact.

The trade is less control: you cannot tune the things a self-managed cluster lets you tune, tunable consistency is reduced to LOCAL_QUORUM and LOCAL_ONE, and DDL is applied asynchronously — a table can exist before it is ready to be written to.

Billing is per read and write request rather than per node, which changes what a bad access pattern costs. A scan that would merely be slow on your own ring is directly expensive here.

Set up the keyspace

1. In the Keyspaces console, create a keyspace and at least one table.
2. Under your IAM user’s security credentials, generate service-specific credentials for Amazon Keyspaces. These are a separate username and password, not your AWS access keys.
3. Note the regional endpoint, for example cassandra.us-east-1.amazonaws.com.
4. Download the Starfield digital certificate that AWS documents for Keyspaces.
5. Keep the credentials somewhere safe — the password is shown once, exactly like an access key secret.

Connect from AddisDB

1. New Connection → Amazon Keyspaces. Port prefills to 9142 — do not change it to 9042.
2. Enter the regional endpoint as the host.
3. Use the service-specific credentials as the username and password.
4. Point the CA certificate field at the Starfield certificate, and leave verification on. Keyspaces refuses plaintext connections entirely.
5. Set the keyspace you want in the database field.
6. Test, then Save.

How to query it

It is Cassandra’s CQL, with Cassandra’s rules: name the partition key, or the statement is refused. Because billing is per request rather than per node, an unbounded read here is expensive as well as slow.

SELECT ts, value
FROM   app.readings
WHERE  sensor_id = 42 AND bucket = '2026-07-31'
ORDER  BY ts DESC
LIMIT  100;

What AddisDB gives you

• Keyspaces and tables in the sidebar, with column types from the CQL schema.
• CQL run from the editor, with results in the grid.
• Read-only enforcement in the driver: mutating CQL is refused before it is sent on a read-only connection.
• The Chart view, ⌘K search, and AI chat grounded in your real schema.