Database guides

Google Cloud Bigtable

Updated 2026-08-11 · 2 min read

Google’s wide-column store, reached over its REST API. None of the connection fields is a hostname.

Cloud Bigtable is Google’s wide-column NoSQL database — the system BigQuery, Search and Maps were built on. AddisDB reaches it over the REST v2 Data API rather than gRPC, so no additional tooling is involved.

Who it is for

Bigtable suits very large, very high-throughput workloads with a known access pattern — time series, IoT telemetry, user activity, financial ticks. It scales horizontally to petabytes and answers single-row lookups in milliseconds.

Its data model is worth stating plainly, because it is unlike a relational database. A table has a row key and column families declared up front. Within a family, individual column qualifiers are created ad hoc per row — two rows in the same table need not have the same columns at all. Everything is bytes, and the row key is the only index. Designing that key IS the schema design.

Set up the token and identifiers

  1. Project ID — from the Google Cloud console. This goes in the username field.
  2. Instance ID — the Bigtable instance. This goes in the database field, as instance or instance/table.
  3. Access token — what gcloud auth print-access-token prints. This goes in the password field.
gcloud auth print-access-token
gcloud bigtable instances list

Connect from AddisDB

  1. New Connection → Google Cloud Bigtable.
  2. Put the project ID in the username field and the instance in the database field.
  3. Paste the access token into the password field.
  4. Test, then Save. There is no host or port to enter.

What AddisDB gives you

Typing a table name reads its rows; a JSON body is sent as a readRows request against the connection’s table, so row-key ranges and filters work. The row key appears as a _key column and each cell as family:qualifier.

Bigtable does not return rows — it returns a stream of cell chunks, where a repeated field is simply omitted and a row ends only when a commit marker arrives. AddisDB assembles those chunks back into rows, carrying the family and qualifier forward, so the grid shows rows rather than fragments.

The schema tree lists tables with their declared column families. It cannot list qualifiers, because Bigtable does not declare them — they appear as the grid reads real rows. Row counts are shown as unknown rather than guessed, since counting would mean a full scan.