The AZ-900 skills outline has six bullets under Describe Azure storage services. The first one, Compare Azure Storage services, is the one most candidates under-prepare for, because the study material spends nearly all its time on Blob Storage and gives the other services one sentence each. Our own Azure Storage accounts guide does exactly that: a four-row table for Blob, Files, Queue and Table, then a deep dive on blob tiers and redundancy, because that post is about storage accounts. Our moving data into Azure guide covers Azure File Sync, AzCopy and Data Box, because it is about the last two bullets.
This post is about the gap in the middle: what Azure Files, Queue Storage, Table Storage and managed disks actually are, which protocol or API each one speaks, which account type and redundancy option each one can live in, what the exam’s favorite distractors are (Service Bus, Azure Cosmos DB, Elastic SAN, Azure NetApp Files, SQL Database), and how to pick the right one from a scenario sentence. Everything here was checked against Microsoft Learn in November 2026, and the skills measured are the July 20, 2026 version.
Where This Sits in the Exam
The outline for Describe Azure storage services, inside the Describe Azure architecture and services domain (35 to 40% of the exam), reads:
- Compare Azure Storage services
- Describe storage tiers
- Describe redundancy options
- Describe storage account options and storage types
- Identify options for moving files, including AzCopy, Azure Storage Explorer, and Azure File Sync
- Describe migration options, including Azure Migrate and Azure Data Box
This post is the first and fourth bullets as they apply to the non-blob services. Fundamentals questions here are identify questions: given a one-sentence need, pick the service. You will not be asked to mount a share or write a queue client. You will be asked to tell a file share from a blob container, a storage queue from a Service Bus queue, a managed disk from Azure Files, and Table Storage from a relational database.
The Map: What Azure Storage Actually Contains
Microsoft’s own introduction page lists the data services in the Azure Storage platform. Memorize the list, because “which of these is not an Azure Storage service” is a real question shape:
| Service | One-line definition from Microsoft | Protocol or API | Lives in a storage account? |
|---|---|---|---|
| Azure Blobs | ”A massively scalable object store for text and binary data”, including Data Lake Storage for analytics | HTTPS/REST, SFTP, NFS 3.0 | Yes |
| Azure Files | ”Managed file shares for cloud or on-premises deployments” | SMB, NFS 4.1, REST | Yes (classic), or standalone via Microsoft.FileShares (NFS only) |
| Azure Queues | ”A messaging store for reliable messaging between application components” | HTTPS/REST | Yes |
| Azure Tables | ”A NoSQL store for schemaless storage of structured data” | HTTPS/REST, OData | Yes |
| Azure managed disks | ”Block-level storage volumes for Azure VMs” | Attached to a VM as a virtual hard disk | No, Azure manages the backing storage for you |
| Azure Elastic SAN | ”A fully integrated solution that simplifies deploying, scaling, managing, and configuring a SAN in Azure” | iSCSI | No |
| Azure Container Storage | ”A volume management, deployment, and orchestration service built natively for containers” | Kubernetes persistent volumes | No |
Two “specialized storage” services sit beside the platform and appear on the exam only as distractors: Azure NetApp Files (enterprise NAS powered by NetApp, managed through NetApp accounts, NFSv3/NFSv4.1/SMB3.1.x) and Azure Managed Lustre (a parallel file system for high-performance computing). If a question’s answer options include one of these for an ordinary “replace our file server” scenario, the ordinary answer is Azure Files.
Everything in a storage account shares one namespace and one set of endpoints:
| Service | Endpoint |
|---|---|
| Blob | https://<account>.blob.core.windows.net |
| Data Lake Storage | https://<account>.dfs.core.windows.net |
| Files | https://<account>.file.core.windows.net |
| Queue | https://<account>.queue.core.windows.net |
| Table | https://<account>.table.core.windows.net |
The account name is 3 to 24 lowercase letters and numbers and must be unique across all of Azure, which is why it appears in every URL.
Storage Account Types and Which Services Each One Holds
Our storage accounts post covers this table, but it is worth repeating with the non-blob lens, because the account type decides which of these services you can even create:
| Account type | Services | Redundancy | Notes for this post |
|---|---|---|---|
| Standard general-purpose v2 | Blob (incl. Data Lake Storage), Files, Queue, Table | LRS, GRS, RA-GRS, ZRS, GZRS, RA-GZRS | The only account that holds queues and tables. HDD (standard) file shares live here |
| Premium block blobs | Blob only | LRS, ZRS | No files, queues or tables |
Premium file shares (FileStorage) | Azure Files only | LRS, ZRS | SSD (premium) shares, and the only account type that supports NFS shares |
| Premium page blobs | Page blobs only | LRS, ZRS | Page blobs are the random-access type that backed unmanaged VM disks |
Three facts fall out of this table that the exam likes:
- Queue and Table Storage exist only in standard general-purpose v2 accounts. There is no premium queue or premium table.
- NFS file shares need a premium file shares (
FileStorage) account or the newer standaloneMicrosoft.FileSharesresource. A standard GPv2 account gives you SMB shares only. - You cannot change an account’s type after creation. To move from standard to premium you create a new account and copy the data.
Legacy general-purpose v1 and BlobStorage accounts are retired for new work; Microsoft wants everything on GPv2 or a premium type. Classic (ASM) accounts were retired on August 31, 2024.
Azure Files
What it is
Azure Files gives you fully managed file shares that you mount like a network drive, “concurrently from cloud and on-premises clients on Windows, Linux, and macOS — no additional software required”. The exam’s one-sentence cues are lift and shift an application that expects a file share, replace or supplement an on-premises file server or NAS, shared configuration or tools accessed from many VMs, and persistent volumes for containers.
The two things that distinguish it from Blob Storage in a scenario: a file share has a directory hierarchy with file-system semantics (locks, ACLs, drive letters), and it is mounted with a standard protocol rather than accessed by URL through an SDK. If the application calls fopen or expects Z:\, it wants Files. If it uploads objects and fetches them by URL, it wants Blobs.
SMB vs NFS
Azure Files speaks two protocols, never both on the same share (you can have SMB and NFS shares side by side in one account):
| SMB | NFS | |
|---|---|---|
| Versions | SMB 3.1.1, 3.0, 2.1 | NFS 4.1 |
| Clients | Windows, Linux, macOS | Linux only |
| Media tier | SSD and HDD | SSD only |
| Redundancy | LRS, ZRS, GRS, GZRS | LRS, ZRS only |
| Semantics | Win32: case-insensitive, ACLs, mandatory byte-range locks | POSIX: case-sensitive, UNIX permissions, hard and symbolic links, advisory locks |
| Authentication | Identity-based (Kerberos) or storage account key (NTLMv2) | Host-based (network) |
| Internet accessible | Yes (SMB 3.0+ with encryption) | No, VNet only (service or private endpoint) |
The networking fact the exam reuses from the on-premises world: SMB uses port 445, and “many organizations and internet service providers (ISPs) block port 445”. That is why mounting an Azure file share from an office often needs a VPN or ExpressRoute with a private endpoint, and it is the practical reason Azure File Sync (an on-premises cache of the share) exists. Our networking post covers VPN Gateway, ExpressRoute and private endpoints.
Identity-based authentication
For SMB shares Microsoft says to use identity-based authentication “in nearly all cases” instead of the storage account key, because the key is effectively an administrator credential that bypasses every ACL. The three identity sources, which map directly onto the identity services in the AZ-900 identity post:
| Identity source | What it is | Typical scenario |
|---|---|---|
| On-premises Active Directory Domain Services (AD DS) | Domain-join the storage account to your own AD, wherever its domain controllers run | Migrating a Windows file server; users keep their existing accounts and NTFS-style ACLs |
| Microsoft Entra Domain Services | Microsoft-managed domain controllers in Azure | Lift-and-shift apps that need AD permissions but you have no on-premises AD |
| Microsoft Entra Kerberos | Microsoft Entra ID issues Kerberos tickets for hybrid or cloud-only identities | Users on Entra-joined devices accessing shares over the internet without line of sight to a domain controller |
All three use Kerberos. NFS shares use none of them; access is controlled at the network and by UNIX permissions.
SSD vs HDD and the billing models
Microsoft has reorganized how Azure Files tiers are described, and older study guides are out of date here. There are two media tiers and three billing models:
| Media tier | Latency promise | Protocols | Billing models |
|---|---|---|---|
| SSD (premium) | Single-digit millisecond for most I/O | SMB and NFS | Provisioned v2, provisioned v1 |
| HDD (standard) | Cost-effective general purpose | SMB only | Provisioned v2, pay-as-you-go |
| Billing model | What you pay for | Microsoft’s guidance |
|---|---|---|
| Provisioned v2 | Storage, IOPS and throughput, each provisioned separately, “regardless of how much you actually use" | "Use the provisioned v2 model for all new Azure Files deployments” |
| Provisioned v1 | Storage you provision; IOPS and throughput are derived from it in a fixed ratio | SSD only; legacy, use v2 unless you have a reason |
| Pay-as-you-go | Used storage plus transactions plus data transfer | HDD only; legacy, “can be difficult to plan for” |
The access tiers that older material calls “transaction optimized, hot and cool” belong only to pay-as-you-go HDD shares. All three run on the same hardware; cooler tiers have cheaper storage and more expensive transactions. Microsoft’s advice is to start a migration in transaction optimized (migration is transaction-heavy), then move once you know the real usage; you can change a share’s tier up to five times in 30 days and not more than once every 24 hours. A provisioned v2 share has no access tiers at all, because you pay for provisioned capacity rather than per transaction. If an AZ-900 question mentions “transaction optimized”, it is describing a pay-as-you-go HDD share.
Sizes to recognize: a provisioned v2 share runs from 32 GiB to 256 TiB; a provisioned v1 share from 100 GiB to 100 TiB. You cannot move a share between media tiers in place; you create a new share on the other tier and copy.
Redundancy per tier
The redundancy options are the same four the storage accounts post explains, with a restriction: HDD shares support LRS, ZRS, GRS and GZRS; SSD shares support only LRS and ZRS. Azure Files does not read from the secondary Region even if the account is RA-GRS, so a share in an RA-GRS account is billed and behaves as GRS. The exam cue “file share that survives a Regional disaster” therefore implies an HDD share with GRS or GZRS; an SSD share is zonal at best.
Snapshots, soft delete, backup and encryption
- Share snapshots are read-only, incremental, point-in-time copies; up to 200 per share, kept for up to 10 years. Azure Backup schedules them with grandfather-father-son retention and can restore a single file or a whole share.
- Soft delete (on by default for new accounts) turns a deleted share into a recoverable state for a retention period you set.
- Encryption in transit is on by default: only SMB 3.x with encryption or HTTPS is allowed. Disabling it lets SMB 2.1 connect, but only from within the same Azure region.
- Encryption at rest always applies, with Microsoft-managed keys by default and customer-managed keys available for classic shares.
Azure File Sync, in one paragraph
Azure File Sync turns a Windows Server into a cache of an Azure file share, with cloud tiering so that cold files live only in Azure and hot files stay local. It is the answer when the question says “keep the performance of a local file server” or “branch offices” or “SMB over port 445 is blocked”. Our moving-data post covers it in depth; here, the point is that File Sync is a feature on top of Azure Files, not a different storage service.
Queue Storage
What it is
Azure Queue Storage is “a service for storing large numbers of messages” that you reach through authenticated HTTP or HTTPS calls. Its job is decoupling: a producer writes a message, a consumer picks it up later, and the two components scale and fail independently. Microsoft’s canonical example is thumbnail generation: the upload handler drops a message in a queue and an Azure Function picks it up to create the thumbnail, so the customer never waits.
The numbers:
| Fact | Value |
|---|---|
| Maximum message size | 64 KB (48 KB of payload if Base64-encoded) |
| Queue capacity | Millions of messages, “up to the total capacity limit of a storage account” (5 PiB default) |
| Number of queues | Unlimited |
| Default message time-to-live | 7 days; can be any positive value or -1 (never expires) on API version 2017-07-29 and later |
| Queue name | 3 to 63 characters, all lowercase |
| URL | https://<account>.queue.core.windows.net/<queue> |
| Delivery guarantee | At-least-once |
| Ordering | Typically FIFO but not guaranteed |
| Receive model | Peek and lease: a message read by one consumer becomes invisible for a visibility timeout (30 seconds default, up to 7 days); if the consumer crashes, the message reappears |
Queue Storage vs Service Bus queues
This is the comparison the exam reaches for. Service Bus is not part of Azure Storage; it is an enterprise messaging broker with queues, topics and subscriptions. Microsoft publishes a decision list, and the exam borrows its rows:
| Choose Storage queues when | Choose Service Bus queues when |
|---|---|
| You must store over 80 GB of messages in a queue | Your queue stays under 80 GB |
| You want to track processing progress inside the message (in-place update) | You need guaranteed FIFO ordering (via sessions) |
| You need server-side logs of every transaction against the queue | You need duplicate detection, transactions, or at-most-once delivery |
| Basic decoupling between components in the same solution | Messages larger than 64 KB (256 KB standard, 1 MB or up to 100 MB premium) |
| You already use core Azure Storage features and want one programming model across blobs, queues and tables | You want publish/subscribe (topics), dead-lettering, long polling or the AMQP 1.0 protocol |
| Capacity | Storage queues | Service Bus queues |
|---|---|---|
| Max queue size | 5 PiB (the storage account) | 1 GB to 80 GB |
| Max message size | 64 KB | 256 KB / 1 MB / 100 MB by tier |
| Max queues | Unlimited | 10,000 per namespace (Standard) |
Fundamentals cue words: simple, backlog of work, asynchronous, decouple web front end from worker, and very large volume of small messages mean Queue Storage. Ordered, exactly once (Microsoft says this term is not quite accurate, but the exam uses it), transactions, publish/subscribe, enterprise integration, and dead-letter queue mean Service Bus. Event Grid and Event Hubs are separate again (event routing and high-throughput event streaming) and are not storage services at all.
Table Storage
What it is
Azure Table Storage “stores non-relational structured data (also known as structured NoSQL data) in the cloud, providing a key/attribute store with a schemaless design”. It is cheap, fast and scales to terabytes. Microsoft’s listed uses: “user data for web applications, address books, device information, or other types of metadata”, datasets “that don’t require complex joins, foreign keys, or stored procedures”, and anything you can denormalize for fast access.
The vocabulary:
| Term | Meaning |
|---|---|
| Table | A collection of entities. Tables do not enforce a schema; one table can hold entities with different property sets |
| Entity | A set of properties, like a row. Up to 1 MB in Table Storage (2 MB in Azure Cosmos DB for Table) |
| Property | A name-value pair. Up to 252 user properties per entity |
| System properties | Every entity also has a PartitionKey, a RowKey and a Timestamp. PartitionKey + RowKey is the unique, clustered index |
| URL | https://<account>.table.core.windows.net/<table>, queryable with the OData protocol |
Entities that share a PartitionKey can be queried faster and updated together atomically. That single clustered index, and nothing else, is why Table Storage is cheap and why it is wrong for anything that needs secondary indexes, joins or SQL.
Table Storage vs Azure Cosmos DB for Table vs SQL
Microsoft now describes Table Storage as “part of Azure Cosmos DB”, with Azure Cosmos DB for Table as the premium offering: the same table data model and the same unified Azure Tables SDK, plus “higher performance and availability, global distribution, and automatic secondary indexes”, and a serverless consumption mode. The exam distinction:
| Need | Service |
|---|---|
| Cheap, schemaless key/attribute storage at TB scale, single Region, cost first | Table Storage |
| Same data model but turnkey global distribution, single-digit-millisecond latency SLAs, automatic indexing on every property | Azure Cosmos DB for Table |
| Relationships, joins, transactions across tables, SQL | Azure SQL Database (not a storage service) |
| Document JSON with rich queries | Azure Cosmos DB for NoSQL |
Fundamentals cue words for Table Storage: NoSQL, key/value, schemaless, structured but non-relational, large amounts of metadata, low cost.
Managed Disks
What “managed” means
A managed disk is a block-level storage volume for an Azure VM, “like physical disks in an on-premises server, but they’re virtualized”. Managed means you specify only the disk type and size; Azure creates and manages the storage accounts behind the scenes (disks are stored as page blobs, which is why premium page blob accounts exist) and you never hit storage-account IOPS limits or count accounts per VM. Managed disks:
- are designed for 99.999% availability with three replicas; LRS disks give at least eleven nines of durability over a year and ZRS disks twelve nines;
- let you create up to 50,000 disks of each type per subscription per region;
- integrate with availability sets (disks of VMs in a set are placed on different storage stamps) and availability zones;
- support snapshots (a read-only, crash-consistent full copy of one disk, billed on used size, e.g. 10 GiB used of a 64 GiB disk bills 10 GiB), managed images (all disks of a generalized VM, used to stamp out many VMs), Azure Backup, restore points and Azure Site Recovery;
- are encrypted at rest with server-side encryption by default, with Azure Disk Encryption (BitLocker/dm-crypt inside the VM), encryption at host and confidential disk encryption as options;
- can be shared between VMs for clustered applications that bring their own cluster manager (WSFC, Pacemaker).
The exam cue words: persistent storage for a VM, block storage, data that does not need to be accessed from outside the VM, lift and shift an app that writes to a local disk. If several VMs need to see the same files, that is Azure Files, not a disk, unless the question explicitly describes a failover cluster.
The five disk types
Memorize the comparison table, in particular the last row:
| Ultra Disk | Premium SSD v2 | Premium SSD | Standard SSD | Standard HDD | |
|---|---|---|---|---|---|
| Media | SSD | SSD | SSD | SSD | HDD |
| Scenario (Microsoft’s words) | “I/O-intensive workloads such as SAP HANA, top-tier databases" | "Production and performance-sensitive workloads that consistently require low latency, high IOPS, and high throughput" | "Production workloads that require high performance and low latency" | "Web servers, lightly used enterprise applications, and development and test environments" | "Backup and other noncritical, infrequently accessed workloads” |
| Max size | 65,536 GiB | 65,536 GiB | 32,767 GiB | 32,767 GiB | 32,767 GiB |
| Max throughput | 10,000 MB/s | 2,000 MB/s | 900 MB/s | 750 MB/s | 500 MB/s |
| Max IOPS | 400,000 | 80,000 | 20,000 | 6,000 | 2,000 (3,000 with performance plus) |
| Usable as OS disk? | No | No | Yes | Yes | Yes, retiring September 8, 2028 |
Four rules the exam derives from that table:
- Ultra Disk and Premium SSD v2 cannot be OS disks. They are data disks only, and Ultra Disks can only be created empty. The documented pattern is a Premium SSD OS disk with Ultra or Premium SSD v2 data disks.
- Premium SSD requires a premium-storage-capable VM size (the sizes with an “s” in the name, such as Dsv5). Standard SSD and Standard HDD work with every VM size.
- Standard HDD as an OS disk is being retired on September 8, 2028; Microsoft says HDDs “aren’t suitable for workloads that require low latency or frequent small-block access, such as OS disks or transactional databases”.
- Billing shape differs. Premium SSD, Standard SSD and Standard HDD come in fixed sizes (P30 = 1 TiB, E10 = 128 GiB, S4 = 32 GiB) and you are billed for the size tier you round up to. Standard SSD and Standard HDD also bill transactions (with hourly caps). Ultra Disk and Premium SSD v2 let you set capacity, IOPS and throughput independently (Premium SSD v2 includes 3,000 IOPS and 125 MB/s free, then charges above that) and adjust them up to four times per 24 hours without detaching the disk.
Premium SSD and Standard SSD also offer bursting (credit-based, up to 30 minutes for smaller sizes, or on-demand paid bursting for P30 and above), which is the answer to “OS disk boot is slow” style questions in the AZ-104 world and occasionally appears here as a feature name. If you later move on to the AZ-104 exam, this table is where the disk questions get much more detailed.
Disk roles: OS, data and temporary
Every VM has three kinds of disk, and the third is the trap:
| Role | What it holds | Persistent? | Notes |
|---|---|---|---|
| OS disk | The operating system and boot volume; drive C on Windows | Yes (unless an ephemeral OS disk is used) | Max 4,095 GiB, but MBR partitioning limits usable size to 2 TiB; use GPT or a data disk for more |
| Data disk | Applications and data; drive letters you choose, registered as SCSI | Yes | The VM size decides how many you can attach and which types |
| Temporary disk | Page files, swap files, SQL tempdb | No | Drive D on Windows, /dev/disk/azure/resource on Linux. Not a managed disk. Data may be lost on maintenance, redeploy or stop/deallocate; it survives a normal restart. Some VM sizes have no temporary disk |
The scenario “a developer stored application files on drive D and they vanished after the VM was deallocated” is a temporary-disk question. Microsoft’s guidance is to keep the OS disk for the OS and put applications and data on data disks for backup, flexibility, isolation and security.
Snapshots vs images
| Managed image | Managed disk snapshot | |
|---|---|---|
| Source | A generalized (Sysprep’d), deallocated VM | A point-in-time copy of one managed disk |
| Scope | All disks of the VM, OS and data | One disk |
| Use | Create many identical VMs | Back up a disk or create a new disk from it |
| Multi-disk | Captures all attached disks together | Not coordinated across disks |
The compute post on Azure compute services covers VM Scale Sets and availability sets, which is where images matter. Ephemeral OS disks (the OS disk lives on the VM’s local storage, is not persisted and is reset on reimage) are an AZ-104 topic, but recognize the name.
Elastic SAN, Azure NetApp Files and Managed Lustre: Know the Names
These three show up as answer options. Know enough to rule them in or out:
| Service | What it is | When it is the answer |
|---|---|---|
| Azure Elastic SAN | A managed storage area network exposed over iSCSI, consolidating storage for many VMs, AKS clusters and databases into one pool | ”Consolidate storage for many large, I/O-intensive workloads into a single SAN-like back end”, “iSCSI” |
| Azure NetApp Files | Enterprise NAS powered by NetApp ONTAP, managed through NetApp accounts, capacity pools and volumes; NFSv3, NFSv4.1 and SMB3.1.x, including dual-protocol volumes; locally redundant with a 99.99% SLA | ”SAP HANA”, “demanding POSIX-compliant application”, “multiple file protocols on one service”, “difficult-to-migrate enterprise app” |
| Azure Managed Lustre | A managed Lustre parallel file system for HPC and AI, integrated with Blob Storage | ”HPC”, “parallel file system”, “high throughput and low latency for simulations” |
The rule of thumb for Fundamentals: an ordinary file-share requirement is Azure Files; an ordinary VM-disk requirement is managed disks; the three above are only correct when the question uses their specific vocabulary.
Putting It Together: The Service-Selection Table
This is the table to carry into the exam for Compare Azure Storage services:
| The scenario says | Choose | Not |
|---|---|---|
| Images, videos, backups, logs, documents accessed by URL at huge scale | Blob Storage | Files |
| Data lake for big-data analytics | Blob Storage with Data Lake Storage (hierarchical namespace) | Table Storage |
| Many VMs or users need the same shared folder; drive letters; SMB | Azure Files (SMB) | Managed disk |
| Linux-only shared storage, POSIX semantics | Azure Files NFS (SSD, premium file shares account) | SMB share |
| Users must authenticate to a share with their domain accounts | Azure Files with identity-based authentication (AD DS, Entra Domain Services, Entra Kerberos) | Storage account key |
| Keep a fast local copy of the share at a branch office | Azure File Sync | A second share |
| Persistent block storage for one VM; application writes to a local disk | Managed disk | Azure Files |
| Highest-performance VM data disk for SAP HANA | Ultra Disk (data disk only) | Standard SSD |
| Dev/test VM, cost-sensitive, consistent latency | Standard SSD | Premium SSD |
| Backup or archive disk, rarely accessed | Standard HDD (data disk) | Ultra Disk |
| Decouple a web front end from background workers with a simple backlog | Queue Storage | Service Bus |
| Guaranteed ordering, transactions, duplicate detection, pub/sub | Service Bus (not a storage service) | Queue Storage |
| Schemaless key/value data, cheap, non-relational, TB scale | Table Storage | Azure SQL Database |
| Same but globally distributed with latency SLAs and automatic indexes | Azure Cosmos DB for Table | Table Storage |
| Relational data with joins and transactions | Azure SQL Database (not a storage service) | Table Storage |
| Consolidated iSCSI storage for many workloads | Azure Elastic SAN | Managed disks |
| Enterprise NAS with NFSv3/4.1 and SMB on the same volume | Azure NetApp Files | Azure Files |
A Ten-Minute Hands-On Walk-Through
You do not need to do this to pass AZ-900, but seeing the four services in one account makes the names stick. Use Cloud Shell (Bash), which already has the CLI signed in. Replace the account name with something globally unique and lowercase.
-
Create a standard GPv2 account (this is the kind that supports all four services):
az group create -n rg-storage-lab -l eastus az storage account create -n stlab$RANDOM -g rg-storage-lab -l eastus \ --sku Standard_LRS --kind StorageV2Note the output’s
primaryEndpointsblock: four different hostnames, one per service, all carrying your account name. -
Create one of each inside it, using the account key for the lab only:
ACCT=$(az storage account list -g rg-storage-lab --query '[0].name' -o tsv) KEY=$(az storage account keys list -n $ACCT -g rg-storage-lab --query '[0].value' -o tsv) az storage share create -n teamshare --quota 100 --account-name $ACCT --account-key $KEY az storage queue create -n thumbnails --account-name $ACCT --account-key $KEY az storage table create -n devices --account-name $ACCT --account-key $KEY az storage container create -n images --account-name $ACCT --account-key $KEY -
Put a message on the queue and read it back twice to see the lease behavior:
az storage message put -q thumbnails --content "resize photo-001.jpg" \ --account-name $ACCT --account-key $KEY az storage message get -q thumbnails --visibility-timeout 30 \ --account-name $ACCT --account-key $KEY az storage message peek -q thumbnails --account-name $ACCT --account-key $KEYThe second command returns the message and hides it; the peek returns nothing until the 30-second visibility timeout expires. That is at-least-once delivery in action.
-
Insert and query an entity:
az storage entity insert -t devices --account-name $ACCT --account-key $KEY \ -e PartitionKey=sensors RowKey=dev-001 model=TX9 firmware=3.2 az storage entity query -t devices --filter "PartitionKey eq 'sensors'" \ --account-name $ACCT --account-key $KEYThe entity has your two properties plus the three system properties. Insert a second entity with different properties into the same table; it works, because tables are schemaless.
-
Try to create an NFS share in this account with
--enabled-protocols NFS. It fails, because NFS needs aFileStorage(premium file shares) account. That failure is the exam fact. -
Look at a managed disk on any VM you have (or create a small one with
az disk create --size-gb 32 --sku StandardSSD_LRS). Notice it has a resource ID in your resource group and no storage account anywhere in sight. That is what managed means. -
Clean up:
az group delete -n rg-storage-lab --yes --no-wait.
Scenario Drill
Cover the right column, read the cue, answer, then check.
| Cue in the question | Answer |
|---|---|
| Replace an on-premises Windows file server; users keep their AD accounts | Azure Files (SMB) with AD DS identity-based authentication |
| Linux HPC nodes need a shared POSIX file system in Azure Files | NFS share, SSD media tier, premium file shares account |
| Share must survive a Regional outage | HDD (standard) share with GRS or GZRS; SSD shares are LRS/ZRS only |
| Branch office cannot reach the share because port 445 is blocked | VPN/ExpressRoute with private endpoint, or Azure File Sync cache |
| Which protocol works from Windows, Linux and macOS? | SMB |
| ”Transaction optimized” tier mentioned | A pay-as-you-go HDD file share |
| Recommended billing model for a new share | Provisioned v2 |
| Recover a file share deleted yesterday | Soft delete (or an Azure Backup snapshot) |
| Store millions of 10 KB work items for asynchronous processing | Queue Storage |
| Queue must guarantee FIFO and support transactions | Service Bus, not Queue Storage |
| Message is 500 KB | Service Bus (or store the payload in a blob and queue a pointer) |
| Queue will exceed 80 GB | Queue Storage |
| Store device metadata, schemaless, cheapest option | Table Storage |
| Same data, needs single-digit-ms reads from three continents | Azure Cosmos DB for Table |
| Application needs joins and foreign keys | Azure SQL Database, not Table Storage |
| Persistent storage for a VM’s database files | Managed data disk |
| Which disk type can be an OS disk: Ultra, Premium SSD v2, or Premium SSD? | Premium SSD only |
| Files on drive D disappeared after deallocation | Temporary disk is not persistent; use a data disk |
| Create 200 identical VMs from a prepared VM | Managed image (generalized VM) |
| Consolidate storage for 50 database VMs behind one iSCSI target | Azure Elastic SAN |
Mistakes to Avoid
- Reaching for Azure Files when one VM needs a disk, or a disk when many VMs need shared files. Disks attach to one VM (shared disks need a cluster manager); file shares are for concurrent access.
- Saying NFS shares live in a standard account. NFS Azure Files needs a premium file shares (
FileStorage) account or theMicrosoft.FileSharesresource, SSD only, LRS or ZRS only. - Treating “transaction optimized / hot / cool” as the tiers of every file share. They apply only to pay-as-you-go HDD shares; provisioned v2 has no access tiers.
- Picking Service Bus for a simple backlog or Queue Storage for ordered, transactional messaging. Over 80 GB and simplicity point to Queue Storage; FIFO, duplicate detection, transactions and pub/sub point to Service Bus.
- Confusing Table Storage with a relational table. It is a schemaless key/attribute store with one clustered index; no joins, no SQL.
- Putting an OS on an Ultra Disk or Premium SSD v2. Neither can be an OS disk.
- Storing anything you care about on the temporary disk. It is lost on deallocate, redeploy and maintenance.
- Calling Azure NetApp Files, Service Bus, Cosmos DB or SQL Database “Azure Storage services”. They are not part of the Azure Storage platform; the exam uses them as distractors.
- Forgetting that account type is permanent. You cannot convert a standard account to premium.
Practise the Comparisons
Compare Azure Storage services questions are decided by one or two cue words, and the only way to make the mapping automatic is repetition against realistic distractors. The free AZ-900 practice questions give you a starting set, and the AZ-900 Azure Fundamentals mock exam bundle has eight full-length exams aligned to the July 2026 skills outline, with explanations that spell out why Table Storage is wrong when the question wants Cosmos DB and why Queue Storage is right when the options include Service Bus. If you are still planning your study, the 30-day AZ-900 study plan shows where this storage material fits, and the AZ-900 exam guide covers registration, format and scoring.
Frequently Asked Questions
What is the difference between Azure Blob Storage and Azure Files?
Blob Storage is object storage: you upload and fetch objects by URL through REST or an SDK, with a flat namespace (plus an optional hierarchical namespace for Data Lake Storage). Azure Files is a managed file share you mount over SMB or NFS with a directory tree, file locks and ACLs, so applications that expect a file system work without changes.
Which Azure Files protocol should I choose, SMB or NFS?
SMB works from Windows, Linux and macOS, supports both SSD and HDD media tiers and all four redundancy options, and supports identity-based authentication with Kerberos. NFS 4.1 is Linux-only, SSD-only, LRS or ZRS only, uses POSIX semantics, and is reachable only from a virtual network. One share speaks one protocol.
What are the Azure Files tiers now?
Two media tiers (SSD, called premium, and HDD, called standard) and three billing models (provisioned v2, recommended for all new shares; provisioned v1, SSD only; pay-as-you-go, HDD only). The transaction optimized, hot and cool access tiers exist only on pay-as-you-go HDD shares.
When should I use Azure Queue Storage instead of Service Bus?
Use Queue Storage for simple, high-volume asynchronous decoupling, when the queue may exceed 80 GB, when you want to update a message in place to track progress, or when you need server-side transaction logs. Use Service Bus when you need guaranteed FIFO ordering, duplicate detection, transactions, messages larger than 64 KB, dead-lettering or publish/subscribe topics.
What is Azure Table Storage and how is it different from Azure Cosmos DB for Table?
Table Storage is a low-cost, schemaless NoSQL key/attribute store where each entity (up to 1 MB, up to 252 properties) is indexed by PartitionKey and RowKey. Azure Cosmos DB for Table uses the same data model and SDK but adds global distribution, latency SLAs, automatic secondary indexes and a serverless mode, at a higher price.
What are the five Azure managed disk types and which can be OS disks?
Ultra Disk, Premium SSD v2, Premium SSD, Standard SSD and Standard HDD. Only Premium SSD, Standard SSD and Standard HDD can be OS disks, and Standard HDD OS disks retire on September 8, 2028. Ultra Disk and Premium SSD v2 are data disks only.
What is the temporary disk on an Azure VM?
Short-term local storage (drive D on Windows, a resource disk on Linux) meant for page files, swap and tempdb. It is not a managed disk and its contents can be lost when the VM is deallocated, redeployed or moved for maintenance. Keep applications and data on managed data disks.
Is Azure NetApp Files part of Azure Storage?
No. Microsoft lists it as specialized storage alongside the Azure Storage platform. It is enterprise NAS powered by NetApp with NFSv3, NFSv4.1 and SMB support and is the answer only when a question describes demanding, multi-protocol or difficult-to-migrate enterprise file workloads such as SAP HANA. Ordinary file-share requirements point to Azure Files.
Conclusion
The AZ-900 Compare Azure Storage services bullet is a vocabulary test. Blobs are objects by URL; Files are mounted shares over SMB or NFS; Queues are 64 KB at-least-once messages for decoupling; Tables are schemaless key/attribute entities under a PartitionKey and RowKey; managed disks are block volumes for one VM in five types, two of which cannot boot an operating system. Around them sit the distractors: Service Bus for ordered or transactional messaging, Azure Cosmos DB for Table for globally distributed tables, Azure SQL Database for relational data, Elastic SAN for iSCSI consolidation, Azure NetApp Files for enterprise NAS.
Learn which account type holds which service (queues and tables only in standard GPv2; NFS shares only in premium file shares), which redundancy each file tier allows (HDD gets geo options, SSD does not), which disk roles persist (OS and data, not temporary), and the handful of numbers (64 KB, 80 GB, 1 MB, 252 properties, 445, five disk types, 4,095 GiB), and every comparison question becomes a lookup.