What is Storage Spaces Direct (S2D)?

Every hard disk runs out of space at some point. Put simply, Storage Spaces Direct is a method of grouping together multiple hard disks into a logical volume that resembles a single disk to the user. With S2D, users no longer have to worry about the physical location of their data, since the S2D file system is au­to­mat­i­cal­ly managed by the servers. This means that the com­po­nents used in the system must be highly resilient.

Storage Spaces Direct allows you to combine the data storage devices of different Windows servers into a single storage cluster and subdivide them into virtual disks that are shared with users. The software takes control of the drives, which has certain im­pli­ca­tions for the hardware you choose for S2D.

Storage Spaces Direct (S2D) is a software-defined storage solution in Windows Server 2016, meaning that it doesn’t run on every Windows PC. The tech­nol­o­gy is based on Storage Spaces, which Microsoft in­tro­duced in Windows Server 2012. At the time, Storage Spaces still included Scale-Out File Server (SOFS), a file server with hor­i­zon­tal scal­a­bil­i­ty.

To use Storage Spaces Direct (S2D), you need multiple hard drives on a single server or multiple servers, each with one or more hard drives. The servers can be connected over Ethernet, meaning no special cables are required. Tra­di­tion­al hard drives (HDD), SSD or NVMe memory cards can be used as drives.

All devices must be Windows Server 2016 Certified by Microsoft for use with Storage Spaces Direct. This ensures the hardware you use will interact seam­less­ly. Microsoft provides a list of pre­con­fig­ured servers suitable for Windows Server 2016 – so called hy­per­con­verged systems.

You can also choose your own com­bi­na­tion of suitable com­po­nents for your specific use case. If you choose to do so, use com­po­nents certified by Microsoft for com­pli­ance with the Software-Defined Data Center (SDDC) standard.

Intel Nehalem or com­pa­ra­ble AMD EPYC proces­sors meet the minimum re­quire­ments for server com­pat­i­bil­i­ty. A Storage Spaces Direct node requires 4 GB of RAM per TB per cache drive, in addition to the Memory for a single Windows Server with the same re­quire­ments.

Hard drives and SSDs can be attached via SAS or SATA. USB drives are not supported. Similarly, RAID con­trollers cannot be used since the hardware is con­trolled ex­clu­sive­ly via the Storage Spaces Direct software layer. The drives must be phys­i­cal­ly connected to the server. Network Attached Storage (NAS) cannot be in­te­grat­ed.

If you want to store data that is accessed or changed fre­quent­ly, you should install SSDs in the servers. In this case, you have to use En­ter­prise SSDs that support power loss pro­tec­tion, meaning they will not generate errors in the file system if the power supply fails. When using SSDs or NVMe drives, all drives installed in a node must be of the same type.

The in­di­vid­ual nodes within a cluster require a 10 Gbps or faster network con­nec­tion to com­mu­ni­cate with each other. Network cards must support RDMA (Remote Direct Memory Access) with the RoCE or iWARP protocols. The usual re­quire­ments for Windows server-based networks apply when it comes to con­nect­ing work­sta­tions to the network.

Storage Spaces Direct is ideal when you want to expand network storage capacity in the most flexible way possible and ensure that data is highly secure in the event of hardware failures. S2D allows you to make data available at different company locations at the same time, assuming you have a fast network con­nec­tion between these locations.

You can also use Storage Spaces Direct to make hyper-V machines in your network more scalable. With S2D, multiple virtual machines can access the same data.

Storage Spaces Direct (S2D) is a software-defined storage solution in which a software ap­pli­ca­tion manages the storage hardware. This allows you to pool storage capacity and combine the technical ad­van­tages of the in­di­vid­ual drives, even if they are in different locations and connected to different servers. S2D groups together servers and storage media into a single Storage Space Direct cluster.

If you use different storage tech­nolo­gies, the software au­to­mat­i­cal­ly decides which data is stored on which hardware. Files used fre­quent­ly are stored on fast NVMe storage media, files that are regularly edited are stored on SSDs, and backup files or rarely used files are stored on con­ven­tion­al hard drives. Ideally, SSD or NVMe storage devices should also be used as cache drives. To users in the network, the whole cluster looks like a network share. That means you don’t have to worry about which data is on which hardware.

You can manage servers using the “Server Manager” console in the Windows Server 2016 operating system or from Command Prompt. When you install the Storage Spaces Direct cluster, all hard disks of the different servers are ag­gre­gat­ed into a single storage pool. It’s important to ensure that the cluster includes only those drives that store user data. The operating system drives of the servers remain in­de­pen­dent.

Usable hard disks and solid-state drives are au­to­mat­i­cal­ly detected and added. In this way, you can add more hard drives and servers to the cluster later and scale storage capacity as needed. In addition to the system hard disk, up to four data drives and two cache drives can be installed in each node.

Virtual disks called cluster shared volumes (CSV) can then be created in the new cluster. These CSVs are then formatted using one of two files systems. Microsoft rec­om­mends CSV-ReFS (Resilient File System), which is more suitable for Storage Spaces Direct tech­nol­o­gy than CSV-NTFS, that is based on the default NTFS file system for hard disks. The CSVs can span multiple physical disks, which is of no concern to the user because every­thing is managed by the system.

Each in­di­vid­ual Storage Spaces Direct node can have a total storage capacity of up to 100 TB. A storage pool con­sist­ing of multiple servers can only have a maximum capacity of 1 PB (one petabyte = 1,024 TB), although Storage Spaces Direct (S2D) allows up to 16 servers.

The in­di­vid­ual network nodes com­mu­ni­cate using the SMB3 protocol, an enhanced version of the Server Message Block (SMB) protocol that includes SMB Direct and SMB Mul­ti­chan­nel.

Thanks to the Failover Clus­ter­ing feature in Windows Server 2016, Storage Spaces Direct has built-in methods of pro­tect­ing stored data from hardware failures. If a hard disk or an entire node fails, no data is lost and the system as a whole remains op­er­a­tional. In most cases, users are unaware of the hardware failure.

If only two nodes are combined into a cluster, two-way mirroring is used. This syn­chro­nizes the data of both nodes so one node or one disk could fail com­plete­ly. Three-way mirroring is rec­om­mend­ed for three nodes or more. Then, one of three nodes can fail com­plete­ly and multiple drives can stop working on another node. If 50% of drives are affected at the same time, data con­sis­ten­cy cannot be ensured. Failure of an operating system hard drive is equiv­a­lent to a server failure.

• Good scal­a­bil­i­ty, more servers can be easily added.
• High avail­abil­i­ty through mirroring of data; hardware failures do not result in data loss.
• Unlike similar solutions from other vendors, S2D allows you to use industry-standard server hardware as long as it is approved by Microsoft.
• The system is optimized for SQL servers and vir­tu­al­iza­tion with Hyper-V.