AZ-305 ยท Data integration and analysis
Data lake architecture
Data Lake Storage as the analytics landing zone, raw/enriched/curated (medallion) layers, file formats such as Parquet and Delta, securing lake data.
Key points
- Azure Data Lake Storage: hierarchical namespace, directories, POSIX-style ACLs.
- Medallion: bronze/raw (as delivered, appended), silver/validated (cleaned, deduplicated, schema-enforced), gold/curated (aggregated for BI, ML).
- Delta Lake: Parquet plus transaction log (ACID, updates, time travel).
- RBAC first: an ACL cannot remove role access; ACLs for groups.
- Shared Key and account/service SAS bypass RBAC and ACLs.
Source: https://learn.microsoft.com/en-us/azure/databricks/lakehouse/medallion
Practice questions
Which storage account setting turns Blob Storage into Azure Data Lake Storage with true directories and POSIX-style ACLs?
- Hierarchical namespace
- Cool access tier
- Geo-zone-redundant storage
- Blob versioning
Show answer
Answer: A. Hierarchical namespace
Enabling the hierarchical namespace gives the account a real directory structure and file/directory ACLs. Access tiers, geo-redundancy and versioning affect cost, durability or recovery, not the namespace model.
Which characteristics describe the bronze (raw) layer of a medallion lake?
- Data is kept in the state the source delivered it
- Records are deduplicated and the schema is enforced
- History is retained so data can be reprocessed or audited
- Data is aggregated into dimensional models for BI
- Little or no cleanup or validation is applied
Show answer
Answer: A. Data is kept in the state the source delivered it, C. History is retained so data can be reprocessed or audited, E. Little or no cleanup or validation is applied
Bronze keeps source data in its original state, grows by incremental appends and retains history so data can be reprocessed or audited, with minimal validation. Deduplication and schema enforcement happen in silver, and dimensional models and aggregates belong in gold.
If a user holds Storage Blob Data Contributor on a Data Lake Storage container, you can still block them from one directory by setting a restrictive ACL on it.
Show answer
Answer: False
False. Role assignments (and their conditions) are evaluated before ACLs. When the role already grants enough access, the ACLs are never consulted, so an ACL cannot narrow access a role has granted. Scope the role more tightly or remove it and use ACLs instead.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
