Storage Spaces Direct requirements you need to know about
Microsoft's architecture uses commodity hardware to create highly available, software-defined storage. But success depends on following the vendor's rules and recommendations.
Microsoft's Storage Spaces Direct is a feature in Windows Server 2016 that uses commodity hardware to create highly available and scalable software-defined storage as a lower-cost alternative to a SAN. Like any enterprise-class storage architecture, planning is key to meeting Storage Spaces Direct requirements and achieving a successful deployment. This is especially true for the disks attached to the Storage Spaces Direct nodes.
Storage Spaces Direct provides a low-cost way of building scalable, highly available storage. But to do so, you must be sure to select storage hardware that complies with the Storage Spaces Direct requirements.
Storage Spaces Direct is based on the use of commodity servers that act as cluster nodes. Storage Spaces Direct requirements include at least two servers. The architecture can be scaled to include as many as 16 servers, each with as much as 100 TB of raw storage capacity. However, there's a trade-off between node count and server capacity. The nodes used in a Storage Spaces Direct deployment can collectively support as much as 1 petabyte of raw storage, meaning a 16-node deployment is limited to supporting a maximum of 62.5 TB of storage per node.
Additional factors must be considered beyond raw storage capacity. For instance, each node must be outfitted with at least four drives that provide storage capacity. Microsoft calls these capacity drives. You couldn't simply install a single 100 TB drive into a node and call it a day. The capacity has to be spread across four or more drives.
Brien Posey, Microsoft MVP, discusses Storage Spaces Direct features
Cache drives recommended
Storage Spaces Direct also enables the use of cache disks. They aren't mandatory, but they are recommended and can improve storage performance. If you do use cache drives, you'll need a minimum of two per server in addition to the capacity drives. The cache drives should be SSDs with a durability rating of at least three full disk writes per day. Any SSDs installed for cache or capacity storage must also be equipped with power loss protection.
Planning is key to meeting Storage Spaces Direct requirements and achieving a successful deployment.
If you install cache drives into your nodes, they will also play a role in the capacity drive architecture. Although not a requirement, Microsoft recommends that the number of capacity drives be an even multiple of the number of cache drives installed. For example, if you install two cache drives, then you might install four or six capacity drives for symmetry. Microsoft claims that this will provide better performance than if you used, say, three or five drives.
As you plan the storage architecture and delve into Storage Spaces Direct requirements, remember that storage hardware must be identical in each node. The nodes must have the same number of disks, and the capacity of each disk must be the same from one node to the next.
Storage Spaces Direct is the virtualized storage component of Windows Server 2016.
Windows gets control
When it comes to selecting the node-level storage architecture, the rule of thumb is to install disks in a way that gives the Windows OS direct control over each disk, as opposed to configuring disk structures at the hardware level. You can attach disks to a RAID controller, but they must be presented to Windows as JBOD. Attaching a RAID array to Storage Spaces Direct isn't supported.
Because Storage Spaces Direct is using commodity hardware to lower storage costs, many organizations use direct-attached SATA disks. Microsoft also supports the use of direct-attached nonvolatile memory express disks. And you can use SATA or SAS disks attached to a SAS host bus adapter, as long as the host bus adapter provides pass-through access to the disks.
Microsoft script that cleans disks in preparation for use in Storage Spaces Direct.
This brings up other Storage Spaces Direct requirements. The disks must be directly attached to the server, although the use of external enclosures is OK. Disks can't be shared among servers, as would be typical of a Cluster Shared Volume, nor can disks be attached through a networking protocol, such as iSCSI or Fibre Channel.
