Buying snapshot
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Decision point | What Microsoft states |
|---|---|
Account and file structure | General Purpose v2 accounts give access to Data Lake Storage along with Block Blobs, Page Blobs, Files and Queues. You choose a hierarchical namespace with directories and folders, or a flat namespace that works like Blob Storage. |
Redundancy | LRS, ZRS, GRS, RA-GRS, GZRS or RA-GZRS. Microsoft states 16 nines of data durability with automatic geo-replication. |
Engines and tools | Designed primarily for Hadoop and frameworks that use the Hadoop FileSystem, such as Spark and Presto. In Azure it works with Data Factory, HDInsight, Databricks, Synapse Analytics and Power BI. |
Network and encryption | Storage firewalls, private endpoints, TLS 1.2 enforcement and encryption at rest with system keys or your own keys. Advanced threat protection is also listed. |
Hadoop migration | Microsoft promotes the WANDisco LiveData Platform for Azure for migrating Hadoop data lakes, and documents how to move an on-premises HDFS cluster. |
Our take
Choose Azure Data Lake Storage when your analytics run on Azure services such as Databricks, Synapse Analytics or HDInsight, and you want analytics files and ordinary objects in one store, with permissions set down to the directory.
Evaluate Azure Data Lake Storage
Comparisons
Where it stands out
Directory operations in one step. With the hierarchical namespace, renaming or deleting a directory is one atomic operation; in an object store, a rename can mean a copy and a delete for every file. This helps analytics jobs that write output to a temporary directory and rename it when they commit.
Permissions down to the folder. Access is authenticated with Microsoft Entra ID and role-based access control. The hierarchical namespace adds POSIX-compliant access control lists, the basis for HDFS permissions, that pass permissions down from parent directories.
Analytics and object data in one store. Analytics data can sit next to object data with no programming changes or data copying. Microsoft says every Blob Storage capability, such as automated lifecycle policy management and high availability and disaster recovery, can be used with Data Lake Storage.
Product screenshot
Trade-offs to weigh
Cooler tiers charge for early deletion. A blob moved to Cool has a 30-day early deletion period on general-purpose v2 accounts. Blobs moved to Cold have a 90-day period and blobs moved to Archive a 180-day period. The charge is prorated: a blob moved to Archive and then deleted or moved to Hot after 45 days is charged an early deletion fee for 135 days of Archive storage. Set lifecycle rules with these periods in mind.
Larger files cost less to process. Files over 4 MB are billed one transaction for every 4 MB block read or written, while a file under 4 MB is charged a full transaction. Microsoft says larger files are more cost effective and give better analytics performance.
Pricing and plan conditions
You pay for the data you store plus the operations you run on it. Storage is priced per GB per month by access tier (Premium, Hot, Cool, Cold or Archive), redundancy and region. Both namespace options pay the same storage price, but the hierarchical namespace is also charged for folder and directory metadata. Usage is measured in binary units: 1 GB is 2^30 bytes and 1 TB is 1,024 GB. Microsoft calls its listed prices estimates that can vary with your agreement, purchase date and exchange rates.
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Item | Flat namespace | Hierarchical namespace |
|---|---|---|
Storage | 120 TB at $0.0184 for the first 50 TB and $0.0177 for the next 450 TB: $2,210.816 | 132 TB, including metadata, at the same rates: $2,428.314 |
Transactions | Writes at $0.05 and reads at $0.004 per 10,000: $264 | Writes at $0.065 and reads at $0.0052 per 10,000: $343.20 |
Directory renames | $0.05 per 100: $5 | $0.065 per 100: $6.50 |
Monthly total | $2,479.816 | $2,778.014 |
Other meters: iterative reads (List Filesystem, List Path), iterative writes (RenameDirectory, RenamePath, RenamePathDir) and all other operations have their own meters, and data retrieval and query acceleration (data scanned and data returned) are billed per GB where the tier offers them. Delete operations are free.
Geo-redundancy: data written to GRS accounts carries a geo-replication data transfer charge, which also applies when you change an account from LRS to GRS or RA-GRS. An account converted from RA-GRS to GRS or LRS is billed as RA-GRS for 30 more days; RA-GZRS to GZRS or ZRS works the same way.
Reserved capacity: one- or three-year commitments in blocks of 100 TB or 1 PB a month for Hot, Cool or Archive, for one region and redundancy, paid upfront or monthly; 1 PB blocks get higher discounts. Reservations do not currently include Premium storage, early deletion charges, operations or data transfer. Exchanges are free; cancellations, up to $50,000 a year, refund the prorated balance minus a 12 percent early termination fee. In Microsoft's example, 100 TB of Hot LRS reserved for one year in US West 2 costs $18,540 upfront or $1,545 a month.
Trial: Microsoft offers free cloud services and a $200 credit to explore Azure for 30 days, or you can pay as you go.
Public signals
About these signals
Verified factual signals from public sources. They indicate observable activity or interest, not total adoption, product quality, or cost.
4.3M PyPI weekly downloads284.5k npm weekly downloads0 vulnerabilities across 2 packages
See all signals from 6 sources
Three-month score against stable baseline terms—not search volume or adoption.
npm · @azure/storage-file-datalake@12.31.0 · PyPI · azure-storage-file-datalake@12.26.1
Alternatives to Azure Data Lake Storage
The reviewed substitutes for Azure Data Lake Storage among the object storage, and what would make each one the better answer.
Other approaches
A different approach to the same problem. Each substitutes only for the workload named beside it.
- Snowflake
- The real question is one tier or two: Snowflake can hold and serve structured and semi-structured data on its own, while a lake-plus-warehouse architecture keeps raw and unstructured data in object storage and loads what analytics needs. Published guidance frames it as whether both are required, which is an architecture decision rather than a substitution.Applies to: Deciding whether raw data needs its own storage tier beside the warehouse.
Related technologies
Normally used together rather than chosen between, so these are not alternatives.
- Databricks
- On Azure, Databricks keeps its tables and files in Azure Data Lake Storage, which is the storage layer, while Databricks supplies the compute, notebooks, SQL and machine learning on top. Unity Catalog governs that data through external locations and managed storage in the same accounts.Applies to: Running a lakehouse on Azure, with the data kept in cloud object storage.