Amazon DataZone
Amazon DataZone is AWS's data management service for cataloguing, discovering, sharing, and governing data across AWS, on-premises, and third-party sources, organised around projects that publish assets and subscription requests that are approved through a governed workflow.
LicenseCommercial
DeploymentManagedServerless
PricingSubscription
Operational complexityLow
WorkloadInteractiveBatch
Use it when
- Data sharing between teams is the actual problem: projects publish assets, consumers subscribe, and approval creates real Lake Formation or Redshift grants.
- The estate is Glue and Redshift centred, where publishing, lineage, and quality-score import are automatic.
- Per-seat catalog pricing has blocked adoption: billing is per API request, metadata storage, and compute, with no per-user subscription at all.
- Governance must gate data, not just describe it: metadata enforcement rules and row- and column-level subscriptions.
Think twice when
- Sources sit mostly outside AWS; native coverage is Glue, Redshift, and hand-published S3 objects, with everything else indirect.
- You want a documentation-first catalog experience; discovery here is organised around the subscription transaction.
- The organisation is not ready to work in projects and domains, which the model requires.
How it runs
Fully managed and serverless on AWS. Domains hold projects whose members publish assets; analysts search a browser data portal outside the console and open Athena or the Redshift query editor on what they find. Identity comes from IAM Identity Center or IAM; fulfilment for non-native sources goes through EventBridge events your own integration completes.
Details
CompareHow Amazon DataZone answers the questions Data Catalogs turns on.
Data Catalogs
| How it works | |
| Metadata model | Domains organised into domain units, holding projects whose members publish assets, an asset type defines the schema an asset is validated against, and metadata form types add typed fields over boolean, date, decimal, integer, string and glossary values. Every edit creates a new asset version, and a project's inventory stays private to its members until explicitly published |
| Lineage | OpenLineage-compatible and generally available since December 2024, table, column and job level, captured automatically from Glue, Redshift and Spark runs and pushed for anything else through the lineage API, with historical versioning so the graph can be read as it stood at a past moment |
| Search and discovery | A browser data portal outside the AWS console, searched in business terms rather than table names, with generative-AI descriptions enriching assets and glossary terms carried at asset and column level; data products bundle related assets into one subscribable package, and only the latest published version is discoverable |
| Business glossary | Business terms attached to assets and to individual columns, and usable as a field type inside a metadata form, so choosing a term as a form value makes the asset findable by that term rather than only annotated with it |
| Quality checks | Imported, never computed: set autoImportDataQualityResult on a data source and AWS Glue Data Quality scores flow in at up to 100 metrics per asset per run, with an API for third-party tools, and consumers see scores per asset and column and how they move over time. Nothing here evaluates data itself |
| Custom metadata | Custom asset types for ML models, dashboards or on-premises tables, metadata form types for compliance and classification fields, environment blueprints deciding what tooling a project gets, and an HTTPS API over all of it, with EventBridge as the seam for fulfilment you carry out yourself |
| Connections | |
| Connectors | Native data sources are the AWS Glue Data Catalog and Amazon Redshift tables and views, with S3 objects published by hand; everything past that is indirect, Glue crawlers reach RDS and the rest, AppFlow's 100-plus connectors bring Snowflake, Salesforce and Google Analytics, and custom asset types cover ML models, dashboards and on-premises tables through the API |
| Access | |
| Policy and compliance | Two layers: authorization policies inside a domain unit govern the catalog itself, deciding who may create projects, glossaries, metadata forms, custom asset types and environment profiles; fine-grained access control then governs the data, restricting a subscription to particular rows and columns rather than the whole asset, with metadata enforcement rules able to demand specified metadata before a subscription request is even accepted |
| Access requests | The product rather than a feature, subscribe to an asset on behalf of a project, the owner approves, and fulfilment creates the actual Lake Formation or Redshift grants across every environment in that project. Unmanaged assets get an EventBridge event carrying the payload instead, and updateSubscriptionStatus closes the loop when something outside AWS does the granting |
| Cost | |
| Billing unit | The only catalog here that dropped per-seat billing outright, on 1 November 2024. What remains is $10 per 100,000 API requests with 4,000 free a month, $0.40 a GB of metadata storage with 20 MB free, $1.776 a compute unit with 0.2 free, and AI recommendations at $0.015 and $0.075 per thousand input and output tokens with no free tier, while CreateDomain, CreateProject and Search never bill at all |
Alternatives to Amazon DataZone
