---
cuppy_hub: gcp
cuppy_path: changelog/2026-10-21-data-family.md
cuppy_section: docs
cuppy_doc: 69XG-HZ89
cuppy_doc_uuid: 892a42c5-c4a5-4f9d-81c2-b0462fc9353b
cuppy_permalink: https://cuppy.io/gcp/d/69XG-HZ89
cuppy_version: 1
cuppy_updated: 2026-10-23T16:16:52.575Z
cuppy_canonical: https://cuppy.io/gcp/docs/changelog/2026-10-21-data-family
cuppy_mcp: https://cuppy.io/mcp
cuppy_body_sha256: 43900da56b6b90fa6243c678990c94b0c96f68656cb45ed1bcd0efa89e3be19d
# cuppy_body_sha256 is over the document body ALONE, excluding both generated blocks.
---


# Data and analytics family — first release

## Added

| Document | What it carries |
|---|---|
| `reference/data/bigquery.md` | separated compute and storage, slots, two compute models, four storage models, four load paths |
| `reference/data/dataflow.md` | Apache Beam, one model for batch and streaming, the named features, exactly-once by default |
| `reference/data/dataproc.md` | the rename, the frameworks, three deployment shapes, the 90-second figure |
| `reference/data/errors.md` | the BigQuery reason strings, the retryable set, and the three rows that are the cost model talking |
| `guides/data/query-bigquery-from-typescript.md` | a keyless client, parameterised queries, and exactly the documented retries |
| `guides/parity/data.md` | why a node count does not convert, and where a lift genuinely exists |

## What this release discovered

**The page served at the Dataproc overview URL is titled "Managed Service for Apache Spark."** Both names are
live — URLs, IaC resource types and CLI surfaces still say `dataproc` — and a document is recorded saying so,
because an agent that assumes one name is stale will mistrust correct documentation.

## The two clean lifts

1. **Apache Beam pipelines onto Dataflow** — Dataflow is a Beam runner, so a Beam pipeline changes runner.
2. **Spark, Hadoop, Hive and Pig onto Dataproc** — the same frameworks. With HBase going to Bigtable and HDFS
   to Cloud Storage, a Hadoop estate decomposes rather than being rewritten.

Both are recorded with their converse, which matters as much: a Flink or Stream Analytics pipeline does **not**
lift onto Dataflow.

## Absences recorded rather than filled in

- **No vendor error table was read for Dataflow or Dataproc**, so none is invented. The error catalogue says so
  and gives the two facts from their own pages that carry the most diagnostic weight.
- The Dataproc overview **does not state whether clusters are zonal or regional**, and names no image versions.
- The BigQuery introduction names the storage billing models **without** stating thresholds; no figure is
  quoted.

## Known missing

- BigQuery partitioning, clustering, materialized views and reservations/assignments.
- BigQuery ML, BigQuery Omni, and external tables.
- Dataflow pipeline options, Dataflow Prime detail, and the Google-provided template catalogue.
- Dataproc Serverless configuration and image-version policy.
- Cloud Data Fusion, Dataplex Universal Catalog, Looker.


---
<!-- cuppy -->
You are reading one file. The same content is free over MCP at https://cuppy.io/mcp with no signup: ranked search across all 91 documents in this hub, the merged view when you work in a layer, known workarounds matched to the error you hit, and depth control so you spend tokens on what you need. Start with the `_hello` tool. How to connect: https://cuppy.io/gcp/ Something wrong in this document? Report it with the `session feedback` tool on the same endpoint.
<!-- /cuppy -->