Guide to VM-aware storage

Most enterprises run dozens or hundreds of VMs on shared storage infrastructure. This includes everything from virtual desktops to production databases running on virtual servers. Hosting can take place on-premises or through an infrastructure-as-a-service (IaaS) provider, typically using a logical unit number (LUN – a unit of block storage) or a file-based storage volume.

According to research from Mordor Intelligence , the virtual machine market was valued at USD 13.59 billion in 2025 and is on track to grow to USD 26.94 billion by 2030. As that growth pushes more VMs onto the same shared storage, it gets harder to tell which VM is using the most resources or causing a slowdown using traditional methods. VM-aware storage exists to solve that problem.

VM-aware storage tracks each VM’s activity separately, applying data services such as provisioning at the individual VM level.  It sees past shared volumes or logical unit numbers (LUNs) that traditional storage treats as one unit, tracking VM performance down to details like which VM is generating the most input/output operations per second (IOPS) at a given time.

Rather than leaving VMs to compete for the same pool, the system gives each VM what it needs. It can also assign quality of service (QoS) settings per VM, rules that control how much storage performance each VM is allowed to use. This way, one high-priority application doesn’t have to compete with a lower-priority one for the same storage. Storage gets provisioned when a VM is created, adjusted as workloads shift, and freed up once a VM is or deleted.

Snapshots are handled per VM, too. A snapshot captures the state of a VM’s virtual disk at a specific point in time, so that data can be restored if something goes wrong. In traditional storage, array-level snapshots capture an entire shared volume. VM-aware storage lets the array take a snapshot of individual VMs, so recovering one VM’s data doesn’t mean restoring the whole volume. (Per-VM snapshots have always been possible at the hypervisor level, but VAS enables such granularity at the storage level.)

It’s worth noting that VM-aware storage applies to the block and file storage that hosts VM disks, not the object storage typically used for archival or unstructured data. For a closer look at how block storage compares to other storage types, see object vs. file vs. block storage.

The main difference between traditional storage and VM-aware storage comes down to visibility. Traditional storage relies on block protocols like Fibre Channel and iSCSI, which were designed for physical servers and and have no built-in way to see individual VM activity. It typically presents a shared volume or logical unit number (LUN) from a storage area network (SAN) as one unit. That blind spot leads to over-provisioning: administrators build in extra capacity just in case, and most of that buffer goes unused.

VM-aware storage takes a different approach, giving each VM only what it needs. Tintri was among the first vendors to bring this approach to market, and other storage vendors have since built their own versions of it. 

This shows up most clearly during troubleshooting. On traditional storage, finding a slow VM usually means pulling logs and checking the whole shared volume, which takes time when systems are already struggling. VM-aware storage skips that step. The information is already tied to the individual VM, so the cause is easy to find.

VM-aware storage delivers several key advantages, building on the broader benefits of virtualization:

- Better performance
- Lower costs
- Faster problem-solving
- Easier scaling
- More precise recovery

With VM-aware storage, VMs no longer have to wait on resources tied up by others. That reduces latency, so applications run more consistently, especially when paired with faster underlying hardware like an all-flash array.

Capacity gets used more efficiently instead of sitting unused to cover worst-case scenarios, allowing more VMs to run on the same hardware.

IT teams can pinpoint which VM is causing an issue instead of searching through logs, which cuts down on troubleshooting time.

With VM-aware storage, new VMs are provisioned with storage automatically as they’re added, removing the need to reconfigure capacity by hand.

A single VM’s data can be restored without touching the rest of the environment, limiting the disruption caused by any one incident. That makes VM-aware storage a useful piece of a disaster recovery strategy, since it allows for per-VM replication before an outage and faster, more targeted recovery afterward.

VM-aware storage comes with some tradeoffs worth weighing, such as: 

- Setup complexity
- Vendor lock-in
- Cost

VAS can simplify day-to-day management, but the migration itself can be more complex. It’s not a simple flip of a switch—an organization must plan to move off a LUN-based setup—and often needs storage built to support VM-level visibility. These potential costs require consideration.

Some VM-aware storage systems are also tied closely to a specific hypervisor, the software layer that lets multiple VMs run on one physical computer, or a specific storage vendor, making it harder to switch platforms later without reworking how storage is configured.

VM-aware storage systems can cost more upfront than traditional storage, since the hardware and software must support tracking and managing data at the VM level.

VMware is the leading virtualization platform in enterprise IT, and storage vendors have built around it accordingly. (However, licensing changes that followed Broadcom’s acquisition of VMware drove some customers to competitors.) That support includes:

- VVols
- vSAN
- VAAI

VVols is VMware’s framework for VM-level storage integration. VVols stores each VM’s disks as individual objects on the storage array over block protocols or Network File System (NFS), rather than placing them on a traditional shared datastore.

With VVols, snapshots, cloning and QoS settings can be applied to one VM at a time rather than an entire volume.

Broadcom acquired VMware in 2023 has begun deprecating VVols. New vVols and VASA Provider certifications were discontinued starting with VMware Cloud Foundation (VCF) and vSphere Foundation (VVF) 9.0, released in June 2025, and Broadcom plans to remove the feature entirely in VCF/VVF 9.1.

VMware also offers vSAN, its own storage option that pools server-attached drives into shared storage without a separate SAN. Because it’s built directly into the hypervisor, vSAN can apply VM-level policies for performance and data protection without a separate integration layer like VVols.

There’s also VAAI (VMware vSphere Storage APIs Array Integration), which hands off certain storage tasks, such as provisioning, from the host server to the storage array itself. When a VM is cloned from a template, the array can copy or overwrite data internally instead of sending it through the host. VAAI accelerates operations like provisioning and cloning but doesn’t give the array visibility into individual VMs on its own. Not every vendor supports VAAI the same way, so it’s worth confirming before settling on a platform.

Not every workload runs on virtual machines. Many organizations also use containers, which package applications without a full operating system and often run inside VMs. Containers can use temporary storage or persistent storage like cloud-native storage managed through Kubernetes or other orchestrators. Among organizations that do run VMs, though, VMware isn’t the only option. Alternate hypervisors include Hyper-V, KVM and Nutanix AHV, and each platform handles storage differently:

• Hyper-V is often paired with Windows Server storage features like Storage Spaces Direct (S2D), which pools drives across servers into shared storage without a separate SAN.
  
  
• KVM usually relies on the underlying Linux storage stack, often paired with tools like Ceph for shared storage across a cluster.
  
  
• Nutanix AHV bundles storage and compute together in a single platform, rather than running them as separate systems. Its AHV hypervisor is based on KVM, though Nutanix also supports Hyper-V and ESXi

Switching hypervisors puts VM-level snapshots, QoS settings and recovery policies at risk, since those settings are often tied to the original hypervisor’s tools and don’t always carry over automatically.

Before migrating, map out which protections exist for each VM and check whether the new environment can replicate them or whether they need to be rebuilt. Some vendors support multiple hypervisors on one platform, which makes this easier. Others are built around a single hypervisor, meaning protections may need to be set up again after the switch.

In August 2026, Broadcom removed public download access to the VDDK (Virtual Disk Development Kit), the library that lets third-party tools read VM disks without going through the guest OS. Tools that used it for agentless migration, including Microsoft Azure Migrate, AWS Application Migration Service and Nutanix Move, now require access through Broadcom’s Technology Alliance Program, or must switch to agent-based methods.

Organizations already managing a hybrid cloud environment, one that spans private cloud, public cloud and on-premises environments, may find this transition easier. Moving workloads between different infrastructure is already part of their day-to-day operations. Testing recovery on the new platform before fully cutting over is worth the extra step, since catching a gap in testing costs far less than catching it mid-migration.