Azure Managed Disks are block-level storage volumes managed by Azure that provide persistent, high-performance storage for Azure Virtual Machines. This chapter covers the four main disk types — HDD, Standard SSD, Premium SSD, and Ultra Disk — their performance characteristics, use cases, and how to choose the right type for your workload. Understanding managed disk types is critical for the AZ-104 exam, as approximately 10-15% of questions touch on storage, with many focusing on disk performance, sizing, and tier selection. You will learn the exact SLA metrics, IOPS/throughput limits, and configuration options that appear on the exam.
Jump to a section
Think of Azure Managed Disks like a hotel buffet. The buffet (Azure storage infrastructure) offers different tiers: the economy HDD tier is like a self-serve cereal station — cheap, slow, and shared; the Standard SSD is like a hot-food line with pre-made dishes — faster but still shared; Premium SSD is like a chef-attended omelet station — dedicated, fast, and consistent; and Ultra Disk is like a private chef cooking to order — ultra-fast, fully customizable, and reserved just for you. The hotel (Azure) manages the buffet layout, replenishment, and maintenance — you just pick your plate (disk type) and size. You don't worry about the kitchen logistics (storage backend), but you must choose the right tier for your appetite (workload). If you pick cereal for a VIP guest, they'll be unhappy (poor performance). If you pick a private chef for a budget traveler, you waste money (over-provisioning). The hotel also offers options like caching (like a hot plate to keep food warm), bursting (allowing a quick second helping), and encryption (like a locked cover). Just as you can't change a buffet station's speed after it's set, you can't change a disk's type without downtime — you must create a new disk and migrate. That's why choosing the right managed disk type upfront is critical.
What Are Azure Managed Disks?
Azure Managed Disks are block-level storage volumes that are managed by Azure and attached to Azure Virtual Machines (VMs). Unlike unmanaged disks, which require you to manage storage accounts, managed disks handle all storage account management automatically, including replication, scaling, and encryption. They provide durable, high-availability storage with a 99.999% availability SLA when used with availability sets or zones.
Why Managed Disks Exist
Before managed disks, administrators had to create and manage storage accounts, set replication, and ensure storage account limits were not exceeded. This was error-prone and time-consuming. Managed disks abstract away this complexity, allowing you to focus on the VM workload. They also offer built-in encryption (Azure SSE), integration with Azure Backup, and support for availability sets and zones.
Disk Types Overview
Azure offers four managed disk types, each designed for different workload requirements:
Ultra Disk: Highest performance with sub-millisecond latency. Supports up to 300,000 IOPS and 2,000 MB/s throughput per disk. Configurable IOPS and throughput independently. Currently available in limited regions and requires specific VM series (e.g., E, F, M series).
Premium SSD v2: Next-generation Premium SSD with higher performance and lower cost than standard Premium SSD. Supports up to 80,000 IOPS and 1,200 MB/s throughput. Configurable IOPS and throughput independently. Lower latency than standard Premium SSD.
Premium SSD: High-performance, low-latency disks for production workloads. Supports up to 20,000 IOPS and 900 MB/s throughput per disk (depending on size). Designed for IO-intensive workloads like databases and real-time analytics.
Standard SSD: Consistent performance with lower latency than HDD. Supports up to 6,000 IOPS and 750 MB/s throughput per disk. Suitable for web servers, lightly used databases, and dev/test environments.
Standard HDD: Lowest cost, highest latency. Supports up to 2,000 IOPS and 500 MB/s throughput per disk. Ideal for backup, archive, and infrequently accessed data.
How Managed Disks Work Internally
Managed disks are built on Azure’s distributed storage fabric called Virtual Hard Disk (VHD) storage. Each disk is a page blob in Azure Storage, but managed disks use a dedicated storage infrastructure that isolates disk I/O for performance consistency. The disk is attached to a VM via the SCSI or NVMe protocol (NVMe for Ultra and Premium SSD v2). The VM’s hypervisor (Hyper-V) presents the disk as a local drive. All I/O operations go through the Azure Storage stack, which handles replication (LRS, ZRS, GRS) and encryption.
For Premium SSD and Ultra Disk, Azure uses a reservation model to guarantee IOPS and throughput. When you provision a disk, Azure reserves the required storage cluster resources to meet the performance SLA. For Standard SSD and HDD, performance is shared among tenants but with baseline limits.
Performance Metrics: IOPS, Throughput, and Latency
The exam tests your understanding of these metrics:
IOPS (Input/Output Operations Per Second): The number of read/write operations per second. Each operation is typically 4 KB or 8 KB. High IOPS is critical for transactional databases.
Throughput (MB/s): The amount of data transferred per second. Throughput = IOPS × I/O size. High throughput is important for large sequential reads/writes (e.g., video streaming, data warehousing).
Latency: The time it takes for a single I/O operation to complete. Ultra Disk offers sub-millisecond latency; Premium SSD offers single-digit milliseconds; Standard SSD offers low tens of milliseconds; HDD offers tens of milliseconds.
Disk Sizing and Performance Tiers
Each disk type has a performance tier that scales with disk size. For Premium SSD, the performance tier is tied to disk size. For example:
P4 (32 GB): 120 IOPS, 25 MB/s
P10 (128 GB): 500 IOPS, 100 MB/s
P30 (1024 GB): 5,000 IOPS, 200 MB/s
P50 (4096 GB): 7,500 IOPS, 250 MB/s
P80 (32767 GB): 20,000 IOPS, 900 MB/s
For Ultra Disk, you configure IOPS and throughput independently of size, up to the maximums.
Bursting
Premium SSD disks smaller than 512 GB support bursting, which allows them to temporarily exceed their provisioned IOPS and throughput for up to 30 minutes. Burst credits are accumulated based on idle time. This is useful for boot operations and periodic spikes.
Caching
Azure VM caching can improve read performance by caching data locally on the host. Options are:
ReadOnly: Caches reads. Good for read-heavy workloads.
ReadWrite: Caches reads and writes. Can improve write performance but risks data loss on host failure.
None: No caching. Use for write-heavy workloads or when caching is not beneficial.
Caching is set at the disk level. For Premium SSDs, ReadOnly caching is common to boost read IOPS.
Encryption
All managed disks are encrypted at rest using Azure Storage Service Encryption (SSE) with platform-managed keys by default. You can also use customer-managed keys (CMK) via Azure Key Vault. Azure Disk Encryption (ADE) provides OS-level encryption using BitLocker (Windows) or DM-Crypt (Linux).
Availability and Redundancy
Managed disks support:
Locally Redundant Storage (LRS): Three copies within a single datacenter.
Zone Redundant Storage (ZRS): Three copies across availability zones in the same region. Available for Premium SSD and Standard SSD.
Geo-Redundant Storage (GRS): Replicates to a paired region (not directly for managed disks; use Azure Backup for geo-redundancy).
Interaction with VM Sizes
Not all VM sizes support all disk types. For example, Ultra Disk requires VMs that support Premium Storage and are specifically enabled for Ultra Disk (e.g., E, F, M series with certain configurations). Premium SSD v2 requires VMs that support NVMe. Standard SSD and HDD are supported on all VM sizes.
Configuration and Verification Commands
To create a managed disk using Azure CLI:
az disk create --resource-group myRG --name myDisk --size-gb 128 --sku Premium_LRSTo attach a disk to a VM:
az vm disk attach --resource-group myRG --vm-name myVM --name myDiskTo view disk performance metrics:
az vm disk show --resource-group myRG --vm-name myVM --name myDisk --query 'diskSizeGb,sku,diskIopsReadWrite,diskMBpsReadWrite'In the Azure portal, you can see disk metrics under the VM's Monitoring blade, including IOPS and throughput.
Key Defaults and Limits
Maximum managed disk size: 32,767 GB (32 TiB).
Maximum number of disks per VM: Varies by VM size (e.g., 64 for D64s v3).
Maximum storage account capacity for unmanaged: 500 TB (not applicable to managed).
Disk size increments: 1 GB for all types.
Ultra Disk: Minimum size 4 GB, maximum 65,536 GB (64 TiB).
Exam-Relevant Details
You cannot change a disk's performance tier without deallocating the VM (for Premium SSD, you can change the tier without deallocation if the new tier is within the same size family).
Disk snapshots are incremental and capture only changes since the last snapshot.
Managed disks support shared disks (for clustered applications) but only with specific disk types (Premium SSD and Ultra Disk) and specific VM sizes.
Disk bursting is automatic for eligible Premium SSDs; you cannot manually trigger it.
When resizing a disk, you must increase the size; you cannot decrease it. After resizing, you may need to extend the partition within the OS.
Summary
Choosing the right managed disk type is a balance of cost, performance, and workload requirements. The exam will test your ability to match disk types to scenarios, understand performance limits, and configure disks correctly. Always consider caching, bursting, and VM size compatibility.
Identify Workload Requirements
First, determine the performance requirements of your workload: required IOPS, throughput, and latency. For example, a production SQL Server database may need 10,000 IOPS and 200 MB/s throughput with sub-5ms latency. A file server might need high throughput (500 MB/s) but moderate IOPS (2,000). A dev/test environment may tolerate higher latency (10-20ms). Also consider cost: HDD is cheapest, Ultra is most expensive. This step defines which disk types are viable.
Select Disk Type and Size
Based on requirements, choose a disk type. For high IOPS and low latency, select Premium SSD or Ultra. For consistent performance at moderate cost, Standard SSD. For low cost and infrequent access, Standard HDD. Then choose a disk size that meets or exceeds the required IOPS/throughput. For Premium SSD, performance scales with size. For Ultra, you set IOPS/throughput independently. Ensure the VM size supports the chosen disk type (e.g., Ultra requires specific VM series).
Configure Caching and Encryption
Set caching based on workload: ReadOnly for read-heavy, None for write-heavy, ReadWrite cautiously. For Premium SSD, ReadOnly caching can significantly boost read IOPS. Enable encryption: Azure SSE is default; if needed, configure customer-managed keys (CMK) in Key Vault. For OS-level encryption, enable Azure Disk Encryption (ADE) via the portal or CLI. Note: ADE requires a Key Vault and may impact performance.
Attach Disk to VM
Use Azure CLI, PowerShell, or portal to attach the disk to the VM. The disk must be in the same region and availability zone as the VM. For shared disks, ensure the VM supports SCSI-3 Persistent Reservations. After attachment, you must initialize the disk inside the OS: create a partition, format with a file system (e.g., NTFS, ext4), and mount it. For Linux, you may need to use parted and mkfs.
Monitor and Optimize Performance
After deployment, monitor disk performance using Azure Monitor metrics: Disk IOPS, Disk Throughput, Disk Latency. If performance is insufficient, consider upgrading to a higher disk tier (e.g., P10 to P20) or enabling bursting if eligible. For Ultra Disk, you can dynamically adjust IOPS/throughput without detaching. For other types, you must deallocate the VM to change the tier. Also check for VM-level bottlenecks (e.g., network bandwidth, CPU).
Enterprise Scenario 1: High-Performance Database
A financial services company runs a mission-critical SQL Server database with 50,000 transactions per second. They need consistent sub-5ms latency and high IOPS (20,000+). They choose Premium SSD v2 disks with 32,000 IOPS and 500 MB/s throughput. To maximize performance, they enable ReadOnly caching on the data disks and None on the transaction log disk. They use a VM series that supports Premium Storage (e.g., E64s_v4). They implement LRS for redundancy within the datacenter. If performance degrades, they can increase IOPS/throughput dynamically without downtime. Misconfiguration: if they accidentally set caching to ReadWrite on the log disk, they risk data loss during host failure. Monitoring shows disk latency consistently below 2ms.
Enterprise Scenario 2: Large-Scale File Server
A media company hosts a video editing platform with 10 TB of project files accessed by 50 editors. The workload is sequential reads/writes with high throughput (1 GB/s) but moderate IOPS (5,000). They choose Standard SSD disks (each 4 TB, P40) striped across multiple disks using Storage Spaces to aggregate throughput. They use ZRS to protect against zone failures. They set caching to ReadOnly on the data disks. They monitor throughput and add disks as needed. Misconfiguration: if they use HDD, editors experience long load times and video stuttering. They also ensure the VM (e.g., M32ms) has enough network bandwidth to handle 1 GB/s.
Enterprise Scenario 3: Cost-Effective Backup
A healthcare provider needs to store nightly backups of patient records for 90 days. The backups are accessed rarely (only for restore). They choose Standard HDD disks (each 2 TB, S60) with LRS. They use Azure Backup to automate snapshots and retention. The low cost of HDD keeps storage bills down. They accept higher latency (50ms) because backups are not time-sensitive. Misconfiguration: if they accidentally use Premium SSD, costs increase 10x with no performance benefit. They also ensure the VM used for backup processing is deallocated when not in use to save compute costs.
What AZ-104 Tests on This Topic (Objective 3.1)
The exam focuses on selecting the appropriate disk type based on workload requirements. Key objective codes: 3.1.1 – Configure storage accounts (but managed disks are separate), 3.1.2 – Configure Azure Files, 3.1.3 – Configure blob storage, and 3.1.4 – Configure disk storage. Specifically, you must know:
Performance characteristics (IOPS, throughput, latency) of each disk type.
When to use each type (e.g., Premium SSD for production databases, Standard HDD for backups).
Disk bursting limits and conditions.
Caching options and their impact.
VM size compatibility (e.g., Ultra Disk requires specific VM series).
How to resize a disk and the implications.
Common Wrong Answers and Why Candidates Choose Them
Choosing Standard HDD for a web server: Candidates think HDD is fine for web servers, but web servers often have moderate IOPS requirements and need consistent performance. Standard SSD is the correct choice for most web servers.
Selecting Premium SSD for archival data: Candidates over-provision for performance, ignoring cost. HDD is sufficient.
Assuming all VM sizes support Ultra Disk: Many candidates think any VM can use Ultra Disk, but only specific series (E, F, M) with certain configurations support it.
Confusing disk caching with Azure Cache for Redis: Disk caching is a host-level feature; it is not a separate service.
Specific Numbers and Values on the Exam
Ultra Disk: up to 300,000 IOPS, 2,000 MB/s throughput.
Premium SSD: up to 20,000 IOPS, 900 MB/s.
Standard SSD: up to 6,000 IOPS, 750 MB/s.
Standard HDD: up to 2,000 IOPS, 500 MB/s.
Bursting: lasts up to 30 minutes, credit accumulation based on idle time.
Maximum disk size: 32,767 GB (32 TiB) for most types; Ultra up to 64 TiB.
Edge Cases the Exam Loves
Changing disk type: You cannot change from HDD to Premium without deallocating the VM (or using a snapshot/migration).
Shared disks: Only Premium SSD and Ultra Disk support shared disks; requires SCSI-3 PR.
Disk encryption: SSE is enabled by default; ADE is optional.
Disk snapshot: Incremental snapshots are faster and cheaper; full snapshots are deprecated.
How to Eliminate Wrong Answers
If a question asks for the lowest cost for a backup workload, eliminate all SSD options.
If a question requires sub-5ms latency, eliminate HDD and Standard SSD.
If a question mentions bursting, look for Premium SSD disks smaller than 512 GB.
If a question says 'without downtime,' you cannot change disk type without deallocation (except Ultra IOPS/throughput adjustments).
Ultra Disk offers the highest performance: up to 300,000 IOPS and 2,000 MB/s throughput.
Premium SSD performance scales with disk size; bursting is available for disks <512 GB for up to 30 minutes.
Standard SSD provides consistent performance at a lower cost than Premium, suitable for many production workloads.
Standard HDD is the cheapest option, best for infrequently accessed data like backups.
You cannot decrease the size of a managed disk; you can only increase it.
To change a disk's type (e.g., from HDD to Premium), you must deallocate the VM or use a snapshot/migration.
Ultra Disk and Premium SSD v2 require specific VM series and regions; always verify compatibility.
Disk caching (ReadOnly, ReadWrite, None) can significantly impact read performance; use ReadOnly for read-heavy workloads on Premium SSD.
All managed disks are encrypted at rest by default with Azure SSE; you can add customer-managed keys or Azure Disk Encryption.
Shared disks are supported only on Premium SSD and Ultra Disk, and require SCSI-3 Persistent Reservations.
These come up on the exam all the time. Here's how to tell them apart.
Premium SSD
Up to 20,000 IOPS per disk
Up to 900 MB/s throughput
Performance tied to disk size
Supports bursting (disks <512 GB)
Available in all regions
Lower cost than Ultra
Ultra Disk
Up to 300,000 IOPS per disk
Up to 2,000 MB/s throughput
IOPS and throughput configurable independently of size
No bursting needed (always high performance)
Limited regional availability
Higher cost
Standard SSD
Up to 6,000 IOPS
Up to 750 MB/s throughput
Consistent single-digit millisecond latency
Suitable for web servers, dev/test
Higher cost than HDD
Standard HDD
Up to 2,000 IOPS
Up to 500 MB/s throughput
Variable latency (10-50ms)
Suitable for backups, archives
Lowest cost
Mistake
You can change a managed disk's type without deallocating the VM.
Correct
For most disk types, you must deallocate the VM to change the SKU (e.g., from Standard_LRS to Premium_LRS). The exception is changing the performance tier of a Premium SSD (e.g., P10 to P20) which can be done without deallocation, but only if the new tier is within the same family. Ultra Disk IOPS/throughput can be adjusted dynamically without deallocation.
Mistake
Standard SSD is faster than Premium SSD because it's newer.
Correct
Premium SSD is significantly faster than Standard SSD. Premium SSD offers up to 20,000 IOPS and 900 MB/s, while Standard SSD caps at 6,000 IOPS and 750 MB/s. Standard SSD is an upgrade from HDD but not a replacement for Premium.
Mistake
All VM sizes support Ultra Disk.
Correct
Ultra Disk requires VMs that support Premium Storage and are specifically enabled for Ultra Disk. Only certain VM series (e.g., E, F, M) in specific regions support Ultra Disk. Always check compatibility before provisioning.
Mistake
Disk bursting is available on all Premium SSDs.
Correct
Disk bursting is only available on Premium SSDs smaller than 512 GB. Larger Premium SSDs do not support bursting. Burst credits are accumulated based on idle time and can be used for up to 30 minutes.
Mistake
You can decrease the size of a managed disk.
Correct
You cannot decrease the size of a managed disk. You can only increase it. If you need a smaller disk, you must create a new disk and copy the data. This is a common exam trap.
Reveal each answer, then mark whether you got it right. Score 60%+ to unlock the next chapter.
For most disk type changes (e.g., from Standard HDD to Premium SSD), you must deallocate the VM, which causes downtime. However, you can change the performance tier of a Premium SSD (e.g., from P10 to P20) without deallocation. Ultra Disk allows you to adjust IOPS and throughput dynamically without downtime. If you need to change disk type without downtime, consider taking a snapshot and creating a new disk of the desired type while the VM is running, then swapping the disks.
Azure Storage Service Encryption (SSE) encrypts data at rest at the storage platform level using 256-bit AES encryption. It is enabled by default for all managed disks and cannot be disabled. Azure Disk Encryption (ADE) provides OS-level encryption using BitLocker (Windows) or DM-Crypt (Linux), and it encrypts data inside the VM before it reaches the disk. ADE is optional and requires a Key Vault. SSE is transparent and has no performance impact; ADE may have a slight overhead.
Disk bursting is available for Premium SSDs smaller than 512 GB. These disks accumulate burst credits based on idle time below their provisioned IOPS/throughput. When the workload spikes, they can use these credits to burst above the provisioned limit for up to 30 minutes. The burst limit is typically 3,500 IOPS and 170 MB/s for the smallest disks. Once credits are exhausted, performance returns to the baseline. Bursting is automatic and cannot be manually triggered.
Yes, but only with shared disks, which are supported on Premium SSD and Ultra Disk. Shared disks use SCSI-3 Persistent Reservations to coordinate access between VMs. The VMs must be in the same availability set or zone, and the VM size must support shared disks. Common use cases are Windows Server Failover Cluster or Linux Pacemaker clusters. Standard SSD and HDD do not support shared disks.
If the disk does not support bursting (size >=512 GB) or burst credits are exhausted, exceeding the provisioned IOPS will result in throttling. I/O requests will be queued and latency will increase. For disks that support bursting, excess IOPS are covered by burst credits. To avoid throttling, ensure the workload stays within the provisioned limits or upgrade to a larger disk or higher tier.
For most production databases, Premium SSD is sufficient and cost-effective. Choose Ultra Disk only if you need more than 20,000 IOPS per disk, sub-millisecond latency, or the ability to independently configure IOPS and throughput. Ultra Disk also requires specific VM series and regions. If your database requires 50,000 IOPS, Ultra Disk is the only choice. Otherwise, Premium SSD is recommended.
Yes, you can convert an unmanaged disk to a managed disk without downtime. This is done by migrating the VM to use managed disks. In the Azure portal, you can 'Migrate to Managed Disks' under the VM's Disks blade. The VM will be stopped briefly during the migration. After migration, you can take advantage of managed disk features like encryption, snapshots, and availability sets.
You've just covered Azure Managed Disk Types: HDD, SSD, Premium, Ultra — now see how well it sticks with free AZ-104 practice questions. Full explanations included, no account needed.
Done with this chapter?