Single service comparison

Service boundary: Purview is a specialized governance control plane that scans metadata but does not run workloads. Huawei's nearer equivalent is the DataArts Catalog + DataArts Architecture + DataArts Security + Data Map components inside DataArts Studio (DGC), which are delivered together as a packaged governance suite and bundled with ETL/development components. DGC's governance modules are not separately purchasable control-plane products; they are gated to the Integration, Lightweight, or Enterprise-class package tiers, so a Purview-like standalone catalog must be carved out operationally.

Cataloging, lineage, and classification model: Purview offers an automated, connector-driven data map with system and custom classification rules, automated sensitive-data labeling, and lineage harvested from supported sources. DataArts Catalog provides technical/business metadata, lineage, and data profiling; DataArts Security adds classification grading and masking. However, Huawei's classification taxonomy, automated classifiers, and out-of-the-box connector breadth for Azure/Microsoft 365/Synapse/Power BI/Fabric sources differ materially; do not assume equivalence without re-validating which source types are supported connectors in each target region.

Integration, scaling, and operational responsibility: Purview scales via capacity units on its managed data map and uses self-hosted integration runtimes for private-network scanning, with policy integrated through Microsoft Defender for Cloud and Purview policy. DGC governance modules are workspace-scoped, depend on DataArts Migration/CDM for ingestion rather than a purview-style scan runtime, and rely on IAM, Obs object policies, and DataArts Security for enforcement. Fabric is a separate serverless SQL/Ray lakehouse for executing queries and ML, not a governance catalog; it coexists with DGC for catalogs/lineage but cannot replace Purview's classification or policy enforcement.

Assessment first: inventory Purview collections, classification rules, lineage exporters, sensitivity labels, and the scanned source list. Classify each dependency as catalog/metadata, classification/policy, or lineage. Map catalog and classification to DataArts Studio's Enterprise-Class package (DataArts Catalog, Architecture, Security, Data Map); reserve DataArts Fabric only for downstream query/ML execution, since Fabric is a compute lakehouse and cannot host the governance catalog. Verify supported data connections against DGC's 30+ source list before assuming parity, especially for Microsoft 365, Synapse, Fabric, and Power BI which may have no native Huawei connector.

Catalog and metadata migration: there is no documented one-click Purview-to-DGC migration tool, so export business glossary terms, classification taxonomy, and asset metadata to CSV/JSON and re-import into DataArts Catalog where the importer supports it. Re-register each source as a DGC data connection, re-run profiling, and reconstruct lineage from job metadata rather than expecting Purview's harvested lineage to transfer. Re-apply sensitive-data classification rules under DataArts Security; expect to rewrite custom regex/classifiers because Purview's rule schema is not portable.

Policy, access, and downstream workload migration: convert Purview access-policy intent into Huawei IAM plus DGC workspace/role permissions and, where row/column masking is needed, DataArts Security masking policies. If Purview was also gating Synapse/Power BI query access, those enforcement paths do not exist in Huawei; downstream query consumers must be re-pointed at DLI/DWS/Fabric and re-authorized. Migrate ETL pipelines that supplied harvested metadata to DataArts Factory jobs so lineage continues to be captured by DGC going forward.

Validation and cost model: validate completeness by comparing asset counts, classification coverage, and lineage completeness between Purview exports and DGC catalog; run sample audits on sensitive-data discovery. DGC is billed by package tier (Integration, Lightweight, Enterprise-class) plus underlying CDM/DLI/DWS/MRS runtime and storage, whereas Purview bills capacity units and scanned-asset volume plus integration-runtime compute. Recalculate TCO on package tier, expected scan frequency, profiling volume, retention, and cross-service runtime before cutover, and confirm region availability of the Enterprise-class package before committing.