VMware Datastore: Architecture and Operation of VMFS, NFS, and Shared Storage

The Fundamental Role of Datastores in Virtualization

In modern virtualization architecture, how computers communicate with their hard drives has changed radically. Instead of each physical server managing its own disk in complete isolation, the VMware vSphere platform utilizes the concept of a datastore. In practice, a datastore is a unified logical repository that aggregates raw storage capacity from a disk array, a SAN network, or a NAS server, turning everything into a clean, ready-to-use space for virtual machines. For the administrator, the datastore hides all the complex engineering behind cables, RAID controllers, and network protocols, presenting itself simply as a large folder where virtual machine files, such as virtual disks known by the VMDK extension, reside.

When configuring a virtualized environment, choosing where these data files live defines the success or failure of the operation. If the storage infrastructure fails, the entire ecosystem of virtual servers halts, regardless of how powerful the processors are or how much RAM is available. This is why understanding datastore anatomy goes beyond being a routine administrative task and becomes an essential survival skill for any engineering team. The flexibility to migrate virtual machines from one physical server to another without shutting them down—widely known as vMotion—is only possible because these machine disks reside on shared storage accessible simultaneously by all servers in the cluster.

Anatomy and Operation of VMFS

VMFS, an acronym for Virtual Machine File System, is the proprietary file system developed by VMware specifically to manage disks in virtualized environments. Unlike traditional file systems found on standard computers, such as Windows NTFS or Linux ext4, VMFS was designed from the ground up to be a high-performance clustered file system. In practice, this means multiple ESXi servers—the hosts running the virtual machines—can read and write to the same datastore at the same time without the risk of corrupting each other's data. This concurrent access capability is the foundational bedrock supporting advanced load balancing and high availability features.

To coordinate this simultaneous access without causing chaos, VMFS uses a sophisticated metadata locking mechanism based on SCSI and structures called heartbeat rings. When an ESXi server needs to alter the file structure of a virtual machine, it applies a temporary lock on that specific region of the disk, instructing other servers to wait their turn. This operation occurs in fractions of a second, ensuring that operations like snapshot creation or virtual disk expansion happen with total safety. Furthermore, VMFS manages dynamic block sizes, optimizing space to accommodate both tiny text files and virtual disk images weighing hundreds of gigabytes without wasting usable capacity.

NFS: Simplicity and File-Based Performance

While VMFS deals directly with data blocks over fiber optic connections or SAS cables, the NFS protocol, standing for Network File System, adopts an entirely different file-based approach. In practice, NFS allows a dedicated storage server to share folders across standard IP networks using traditional networking protocols. For VMware vSphere, an NFS datastore operates much like a VMFS datastore, but with a striking architectural difference: the file system organizing data inside the shared folder is not controlled by ESXi, but rather by the external storage appliance or the operating system hosting the share.

The major advantage of NFS in modern infrastructure is its implementation simplicity and excellent cost-benefit ratio. Because it utilizes standard Ethernet networks, often leveraging existing switches and cables used for other purposes, companies avoid the high financial investment required for special fiber optic adapters known as HBAs and dedicated SAN switches. However, this reliance on the IP network demands rigorous infrastructure planning, ensuring storage traffic has guaranteed bandwidth and low latency. With the evolution of network standards, using NFS v3 and v4.1 has become extremely robust, even supporting multiple paths to prevent connection drops if a network cable fails.

Shared Storage vs. Local Storage

One of the most critical design decisions when structuring a VMware environment involves choosing between keeping datastores locally on each physical server or centralizing them on shared storage. Local storage utilizes hard drives or solid-state drives installed directly inside the ESXi server chassis itself. In practice, this approach offers exceptional raw performance and reduced initial cost since it eliminates the need to purchase an expensive external storage array. However, the major Achilles' heel of local storage is the lack of mobility and shared redundancy: if the physical server housing the disks suffers a critical motherboard or power supply failure, the virtual machines residing there become inaccessible until hardware repair occurs.

Conversely, shared storage centralizes data on robust equipment, such as block-based SANs or file-based NAS storages. Although the initial acquisition cost is considerably higher, centralization unlocks the full operational potential of virtualization. If a physical server fails, the vSphere system can automatically restart affected virtual machines on another healthy server in the cluster, reading the exact same data directly from the shared datastore. This automated resilience drastically reduces unplanned downtime and simplifies routine preventive maintenance tasks, allowing administrators to shut down entire physical servers in broad daylight without interrupting essential organizational services.

Performance and Management Best Practices

Ensuring that datastores operate at peak efficiency requires continuous attention to configuration details and vital metric monitoring. One of the golden rules in VMware environment administration is to avoid totally exhausting a datastore's capacity, always maintaining a safety margin of at least ten to fifteen percent free space. When a datastore completely fills up, virtual machines depending on it can suddenly freeze because they cannot write essential temporary files required for operation. Additionally, constant monitoring of read and write latency helps identify bottlenecks before they impact end-user experience on hosted systems.

Another critical management aspect involves correctly configuring multipathing, ensuring redundant routes exist between ESXi servers and shared datastores. If a network cable or switch port fails, multipathing software automatically reroutes traffic to alternative paths without manual intervention and without perceptible packet drops. Utilizing modern deduplication and compression technologies directly on storage also maximizes physical disk utilization, allowing much more data to be stored using less actual space. Ultimately, mastering VMFS and NFS datastore architecture transforms IT infrastructure from an unpredictable cost center into a solid, scalable, and highly reliable foundation for business growth.