Domain 3 of 5 · Chapter 4 of 4

Data Platform

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Professional Data Engineer premium course — no separate purchase.

Included in this chapter:

  • What a data platform is, and the pieces you assemble
  • Govern in place with Knowledge Catalog
  • Federated governance and the data mesh
  • Sharing data products: BigQuery sharing and clean rooms
  • Exam-pattern recognition

Data platform building blocks by the job they do

Building blockJob in the platformWhat it producesChoose when
Knowledge CatalogUnified governance, discovery, metadata, quality, lineageAn org-wide catalog over BigQuery + Cloud Storage assets in placeScattered assets must be found, trusted, and governed across teams
BigQuery + Cloud StorageThe underlying analytical and object storesThe actual data products (datasets, files) domains ownYou need the store that holds a domain's data product
BigQuery sharing (Analytics Hub)Controlled, no-copy distribution between teams or orgsListings in exchanges, read-only linked datasets for subscribersA consumer must access a producer's dataset without a copy
Data clean roomPrivacy-preserving cross-party joinsShared analysis governed by analysis rules and egress controlsSensitive data from multiple parties must be joined, never revealed
Data mesh (organizational model)Decentralized ownership with federated governanceDomain-owned data products under one central policyMany domains should own their data while one team sets the rules

Decision tree

Join data withoutrevealing raw rows?Data clean roomYesConsumer needs adataset, no copy?NoBigQuery sharingYesDomains should owntheir own data?NoData meshfederated governanceYesKnowledge Catalogfind + govern in placeNoAlways: stores stay in BigQuery and Cloud Storage; the platform governs and shares them in place
Cheat sheet

Cheat sheet

  • A data platform is the governed layer over your stores, not another store
  • Govern data in place across BigQuery and Cloud Storage, do not consolidate first
  • Knowledge Catalog does cataloging, discovery, profiling, quality, and lineage
  • Dataplex, Data Catalog, and Knowledge Catalog name the same catalog layer
  • Dataplex auto data quality yields a score that can gate a pipeline
  • A data mesh decentralizes ownership while keeping one central policy
  • Federated governance means central rules, domain-owned data
  • A data product is a logical container of related data resources with a defined interface
  • In a Dataplex data mesh a lake is a domain and an asset is a bucket or dataset
  • Share a dataset with BigQuery sharing, not by granting raw IAM
  • A subscriber gets a read-only linked dataset, never a copy
  • A data exchange can be private or public, and listings can be monetized
  • Use a data clean room to join data without revealing raw rows
  • Clean rooms enforce analysis rules and egress controls
  • Cataloging and sharing solve discovery and governance, not movement
  • Policy-tag taxonomies are regional, so replicate them to apply across regions
  • Split BigQuery column access across Policy Tag Admin, Fine-Grained Reader, and Masked Reader
  • Filter BigQuery rows per user with row-level access policies

Unlock with Premium — includes all practice exams and the complete study guide.

References

  1. Cloud Storage product overview
  2. Knowledge Catalog (formerly Dataplex Universal Catalog) introduction
  3. Introduction to BigQuery sharing (Analytics Hub)
  4. Dataplex auto data quality overview
  5. Architecture and functions in a data mesh Well-Architected
  6. BigQuery data clean rooms