Before a new physical disk can be used, you must initialize it with a partition style. The two main styles are Master Boot Record (MBR) and GUID Partition Table (GPT). You must choose GPT for drives larger than two terabytes or if you need more than four primary partitions. GPT also works natively with modern UEFI boot systems, while MBR is kept for older legacy BIOS systems.
After initialization, you can create different volume types depending on your speed and safety needs. A simple volume uses space on a single physical disk but provides no protection if that disk fails. A striped volume spreads data across multiple disks to make reads and writes faster, but a single disk failure destroys all the data. Choose a mirrored volume when you need data redundancy, as it duplicates your files across two or more physical disks.
Storage Pools combine multiple physical drives of different sizes and types into a single virtual pool of storage. From this pool, you can create virtual disks which are known as Storage Spaces to present to the operating system. This setup separates physical hardware from the logical volumes that your applications actually write data to. In hybrid designs, Storage Spaces Direct extends this concept across multiple servers by using local drives to create highly available shared storage.
A layered stack diagram showing how Storage Spaces abstracts physical disks into a storage pool, then into virtual disks with resiliency, and finally into formatted volumes presented to the operating system.
When creating a virtual disk, you must select a resiliency type to protect against drive failures. A mirror resiliency write-path copies data across multiple drives, making it ideal for high-performance workloads like databases. A parity resiliency write-path calculates error-correcting data, which saves physical space but requires more processor effort. Choose simple resiliency only when data can be easily replaced, because it offers no protection against hardware failures.
Storage Replica protects your data by copying volumes between servers or clusters for disaster recovery. You can choose synchronous replication, which writes data to both the local and remote storage before confirming the write is complete. This mode prevents any data loss but requires a high-speed, low-latency network connection to avoid slowing down applications. Alternatively, asynchronous replication confirms the write immediately on the local server and copies data to the destination later, making it better for long-distance connections.
The service supports server-to-server, cluster-to-cluster, and stretch cluster configurations. During active replication, the destination volume remains in a read-only or dismounted state to prevent data conflicts. If the primary site fails, you can switch the replication direction to make the destination volume active and writeable. This orderly transition ensures your applications can quickly recover and run from the secondary location.
Data Deduplication is a service that optimizes storage space by finding and removing duplicate data on a volume. The service breaks files into small chunks, identifies duplicates, and stores only a single copy of each unique chunk. It then replaces the duplicate chunks with pointers that redirect back to the single stored copy. This process is highly effective for virtual machine libraries and general file servers, but it is not recommended for database servers with constantly changing files.
The deduplication process runs in the background using several distinct system jobs. The optimization job processes new files to compress them and move them to the chunk store. The garbage collection job runs weekly to permanently delete chunks that are no longer referenced by any files. Finally, the scrubbing job regularly scans the volume to detect and fix any corrupted chunks using redundant copies.
The SMB Direct feature allows file storage traffic to flow with extreme speed and minimal delay. It requires network adapters that support Remote Direct Memory Access (RDMA), which can use either RoCE or iWARP standards. When configured correctly, the network adapters bypass the traditional operating system network stack. This allows data to move directly between the memory of the client and the server.
By bypassing the standard processor path, this feature drastically reduces CPU usage during large file transfers. This resource savings allows the host processor to focus on running virtual machines rather than handling network packets. You should choose this technology when building high-performance storage networks for Hyper-V or Storage Spaces Direct clusters. The resulting connection behaves like local storage while maintaining the flexibility of a network-attached system.
The Storage QoS (Quality of Service) feature helps administrators monitor and manage storage performance for virtual machines. It prevents a single virtual machine from consuming all the available disk speed, which is a problem known as a noisy neighbor. By managing storage traffic at the hypervisor layer, you can guarantee that critical workloads always get the performance they need. This control ensures that less important systems do not slow down business-critical databases.
Administrators manage these limits by creating policy rules that define minimum and maximum performance levels. A minimum policy guarantees a specific amount of input-output operations per second (IOPS) to a virtual machine. A maximum policy sets a hard cap on IOPS to keep a virtual machine from overloading the storage system. You can apply these policies to a single virtual disk or share them across a group of virtual machines.
When formatting volumes, you must select the file system that best fits your workload. The NTFS file system remains the default choice for system drives and general-purpose file shares because of its compatibility. Conversely, ReFS (Resilient File System) is designed for large data volumes, virtualization, and backup storage. You should choose ReFS for Hyper-V storage because its block cloning feature allows virtual machine checkpoints to merge almost instantly.
File systems manage how data is written and checked for errors over time. ReFS uses integrity streams to monitor files and detect silent data corruption without user intervention. If corruption is found on a mirrored volume, ReFS repairs the damaged file using a healthy copy while keeping the disk online. NTFS requires running utility tools that might require taking the disk offline to repair system errors.
The iSCSI Target Server role allows a Windows Server to share block storage over a standard TCP/IP network. An iSCSI Initiator client connects to the target server to access this storage as if it were a local hard drive. The initiator sends standard SCSI commands wrapped in network packets across the network to the target. This relationship allows you to build a shared storage network without buying expensive fiber channel hardware.
Before a client can connect, the target server must authorize the initiator using its iSCSI Qualified Name (IQN). For extra security, you can configure CHAP authentication to verify the identity of the client before allowing access. You can also encrypt the network traffic between the initiator and the target using IPSec policies. Once the secure connection is established, the storage appears in Windows Disk Management, ready to be initialized and formatted.
Eager to master hybrid server management? Discover how to administer Windows Server Hybrid Core Infrastructure on Azure, setting your path towards the Microsoft Certified: Azure Hybrid Infrastructure Administrator Associate certification!
Prepare and test your skills
Prepare and test your skills
MBR is kept for older legacy BIOS systems, while GPT works natively with modern UEFI boot systems. You must choose GPT for drives larger than two terabytes or if you need more than four primary partitions.
Choose a mirrored volume when you need data redundancy, as it duplicates your files across two or more physical disks. A simple volume uses space on a single physical disk but provides no protection if that disk fails.
Synchronous replication writes data to both the local and remote storage before confirming the write is complete, preventing data loss but requiring a high-speed, low-latency network connection. Asynchronous replication confirms the write immediately on the local server and copies data to the destination later, making it better for long-distance connections.
ReFS is designed for large data volumes, virtualization, and backup storage. Its block cloning feature allows virtual machine checkpoints to merge almost instantly. ReFS also uses integrity streams to monitor files and detect silent data corruption without user intervention, repairing damaged files on mirrored volumes while keeping the disk online.
StudyCerts
StudyCerts