The Shift Away from Traditional SANs
For decades, enterprise virtualization relied on the traditional 3-tier architecture: computation servers (running VMware or Hyper-V), storage arrays (expensive SANs from EMC or NetApp), and the complex Fibre Channel switches that connected them together. This architecture is highly resilient but incredibly expensive to purchase, scale, and maintain.
The modern data center has shifted toward Hyper-Converged Infrastructure (HCI). In HCI, standard x86 servers are filled with direct-attached NVMe drives and high-speed network cards. The hypervisor clusters these servers together, pooling their internal disks into one massive, highly available software-defined storage array. If you need more compute or storage, you simply slide another 1U server into the rack and join it to the cluster.
Microsoft’s native solution for HCI is Storage Spaces Direct (S2D), a feature built directly into the Windows Server Datacenter edition.
Step 1: Hardware and Network Prerequisites
Because S2D eliminates the dedicated storage fabric, your underlying network must be flawless. S2D replicates massive amounts of block-level data between the nodes synchronously over the network to guarantee zero data loss.
- Nodes: A minimum of two (and up to 16) identical Windows Server Datacenter nodes.
- Storage: At least two solid-state drives per node (NVMe or SAS SSDs) for the cache tier, plus capacity drives (HDD or SSD). RAID controllers must be flashed to IT Mode (HBA); S2D must see the raw, unconfigured disks.
- Networking: At least 10GbE network adapters, preferably 25GbE or 100GbE, supporting RDMA (RoCEv2 or iWARP). RDMA allows the servers to move data directly from the memory of one node to the memory of another, bypassing the CPU entirely and ensuring SAN-like latency.
Step 2: Installing the Required Roles
Log into one of your nodes (e.g., Node01) and open an elevated PowerShell prompt. You must install the Hyper-V, Failover Clustering, and File Server roles on all nodes.
$nodes = @("Node01", "Node02", "Node03", "Node04")
Invoke-Command -ComputerName $nodes -ScriptBlock {
Install-WindowsFeature -Name Hyper-V, Failover-Clustering, Data-Center-Bridging, RSAT-Clustering-PowerShell, FS-FileServer -IncludeManagementTools -Restart
}Step 3: Validating the Cluster
Before building the cluster, you must run the Microsoft Cluster Validation wizard. This runs hundreds of automated checks to ensure your network cards, firmware, and storage meet the strict S2D requirements.
Test-Cluster -Node $nodes -Include "Storage Spaces Direct", "Inventory", "Network", "System Configuration"Review the resulting HTML report carefully. If the storage or network tests fail, do not proceed. S2D will fail catastrophically if deployed on unvalidated hardware.
Step 4: Creating the Cluster and Enabling S2D
Once validation passes, create the Failover Cluster. (You do not need to assign a static IP to the cluster if you are running Server 2019/2022, as it defaults to a distributed network name).
New-Cluster -Name S2D-Cluster01 -Node $nodes -NoStorageNow, execute the single command that transforms the cluster into a hyper-converged array. This command will automatically claim all eligible raw disks across all nodes, analyze their performance characteristics, and build the storage pool.
Enable-ClusterS2DPowerShell will output a summary showing the total raw capacity pooled across the cluster.
Step 5: Creating the Virtual Disks (Volumes)
With the raw storage pooled, you can now carve out Virtual Disks. Because S2D is software-defined, you choose the resiliency level (mirroring or parity) at the volume level, not the disk level.
For high-performance SQL databases or Hyper-V VMs, create a Three-Way Mirror volume. This writes every block of data to three different physical servers simultaneously, allowing the cluster to survive the simultaneous hardware failure of two entire servers without data loss.
New-Volume -FriendlyName "VM_Storage_1" -FileSystem CSVFS_ReFS -ResiliencySettingName Mirror -PhysicalDiskRedundancy 2 -Size 5TBNotice the filesystem is CSVFS_ReFS. S2D relies on the Resilient File System (ReFS), which provides real-time block cloning (creating fixed-size VHDXs instantly) and proactive data corruption repair.
Conclusion
Windows Server Storage Spaces Direct democratizes the datacenter by moving the intelligence out of proprietary SAN hardware and into the operating system. By utilizing standard x86 servers and RDMA networking, IT administrators can build highly available, blisteringly fast hyper-converged clusters capable of supporting the most demanding enterprise virtualization workloads.