14.1 Disk Partitioning, Volumes & Enterprise File Systems (NTFS vs ReFS)

Disk Partitioning, Volumes & Enterprise File Systems (NTFS vs ReFS)

Enterprise storage infrastructure in Windows Server forms the bedrock for mission-critical workloads, virtualization fabrics, database engines, and disaster recovery architectures. Designing, configuring, and optimizing storage requires an in-depth understanding of low-level disk partitioning schemes, modern volume management paradigms, and resilient file systems.

This section covers disk initialization (MBR vs. GPT), basic versus dynamic disks, file system selection (NTFS vs. ReFS), and the ReFS virtualization acceleration mechanisms (Block Cloning, Sparse VDL). Data Deduplication — its architecture, usage types, job scheduling, and PowerShell management — is covered in section 14.2.

---

1. Disk Partitioning & Initialization: MBR vs. GPT

When a physical disk or virtual hard disk (VHD/VHDX) is attached to Windows Server, it must be initialized with a specific partition style before partitions and volumes can be created. Windows Server supports two partitioning architectures: Master Boot Record (MBR) and GUID Partition Table (GPT).

+-----------------------------------------------------------------------------------+
|                         DISK PARTITIONING ARCHITECTURES                           |
|                                                                                   |
|   MASTER BOOT RECORD (MBR)                      GUID PARTITION TABLE (GPT)        |
|   +--------------------------+                  +-------------------------------+ |
|   | LBA 0: Master Boot Record|                  | LBA 0: Protective MBR         | |
|   | - Bootstrap Code         |                  +-------------------------------+ |
|   | - 4 Partition Table (64B)|                  | LBA 1: Primary GPT Header     | |
|   | - Boot Signature (0x55AA)|                  +-------------------------------+ |
|   +--------------------------+                  | LBA 2-33: Partition Array     | |
|   | Primary Partition 1      |                  | (128 Partition Entries)       | |
|   +--------------------------+                  +-------------------------------+ |
|   | Primary Partition 2      |                  | User Data Partitions          | |
|   +--------------------------+                  | (Up to 18 Exabytes)           | |
|   | Primary Partition 3      |                  |                               | |
|   +--------------------------+                  +-------------------------------+ |
|   | Extended Partition       |                  | LBA -33 to -2: Backup Entries | |
|   |  - Logical Drive 1       |                  +-------------------------------+ |
|   |  - Logical Drive 2       |                  | LBA -1: Backup GPT Header     | |
|   +--------------------------+                  +-------------------------------+ |
+-----------------------------------------------------------------------------------+

Architectural Comparison: MBR vs. GPT

Deep-Dive Analysis

• The 2 TB MBR Ceiling: MBR utilizes a 32-bit field to record sector addresses. Assuming standard 512-byte sectors, the maximum addressable space is $2^{32} \times 512 \text{ bytes} = 2,199,023,255,552 \text{ bytes} \approx 2.19 \text{ TB}$. Any physical drive capacity beyond 2 TB initialized as MBR becomes unusable unallocated space.
• GPT Redundancy and Self-Healing: GPT writes a primary partition table at the beginning of the disk (LBA 1 through 33) and a duplicate backup table at the very end of the disk. Every GPT header contains a Cyclic Redundancy Check (CRC32) checksum of itself and the partition array. If corruption occurs in the primary table (e.g., due to media degradation or an interrupted write), Windows Server validates the CRC32 checksum, detects the corruption, and automatically reconstructs the partition layout using the backup GPT structures.

---

2. Basic Disks vs. Dynamic Disks (Deprecation Notice)

Windows Server categorizes disk storage into two distinct management types: Basic Disks and Dynamic Disks.

+-----------------------------------------------------------------------------------+
|                         BASIC DISKS VS DYNAMIC DISKS                              |
|                                                                                   |
|   BASIC DISKS (SUPPORTED & STANDARD)            DYNAMIC DISKS (DEPRECATED)        |
|   +------------------------------------+        +-------------------------------+ |
|   | - Uses MBR or GPT Partition Tables |        | - Uses LDM (Logical Disk Mgr) | |
|   | - Simple & Primary Partitions      |        | - Spanned, Striped, Mirrored  | |
|   | - Fully supported in Failover      |        | - Incompatible with S2D/CSV   | |
|   |   Clustering & Storage Spaces      |        | - Cannot host BitLocker OS    | |
|   | - Native ReFS Support              |        | - Deprecated since WS 2012 R2 | |
|   +------------------------------------+        +-------------------------------+ |
+-----------------------------------------------------------------------------------+

Detailed Differences and Volume Types

1. Basic Disks:
   • The standard disk type used by Windows Server.
   • Relies strictly on standard MBR or GPT partition tables to define simple volumes.
   • Supports all Windows Server enterprise features: Storage Spaces, Storage Spaces Direct (S2D), Clustered Shared Volumes (CSV), BitLocker Drive Encryption, and ReFS.
2. Dynamic Disks:
   • Introduced in Windows 2000 using a proprietary Logical Disk Manager (LDM) database stored in the last 1 MB of the physical disk.
   • Permitted software-based multi-disk volumes without hardware RAID controllers:
     • Simple Volume: Portions of a single dynamic disk.
     • Spanned Volume: Combines unallocated space from 2 to 32 disks into a single non-fault-tolerant volume.
     • Striped Volume (RAID-0): Interleaves data across 2 to 32 disks for performance without fault tolerance.
     • Mirrored Volume (RAID-1): Tolerates 1 disk failure using two identical copies.
     • RAID-5 Volume: Stripes data and parity across 3 or more disks.

[!WARNING] Dynamic Disks Deprecation & Exam Trap: Dynamic Disks are officially deprecated in Windows Server. Dynamic disks do not support Storage Spaces, Storage Spaces Direct (S2D), Clustered Shared Volumes (CSV), or ReFS advanced features. Microsoft recommends converting all legacy Dynamic Disks to Basic Disks and implementing software pooling via Storage Spaces or hardware RAID.

---

3. Enterprise File Systems: NTFS vs. ReFS

Windows Server provides two primary file systems for enterprise data storage: NTFS (New Technology File System) and ReFS (Resilient File System).

+-----------------------------------------------------------------------------------+
|                         FILE SYSTEM FEATURE COMPARISON                            |
|                                                                                   |
|   CAPABILITY / FEATURE             NTFS                    ReFS (v3.x)            |
|   -------------------------------+-----------------------+------------------------|
|   Maximum Volume Size            | 256 TB (64 KB cluster)| 35 Petabytes (35 PB)   |
|   Maximum File Size              | 256 TB (64 KB cluster)| 35 Petabytes (35 PB)   |
|   Boot / OS Volume Support       | YES (Mandatory)       | NO (Non-bootable only) |
|   File-Level Security (ACLs)     | YES                   | YES                    |
|   Integrity Streams (Checksums)  | NO                    | YES (Data & Metadata)  |
|   Proactive Data Scrubbing       | NO                    | YES (Self-Healing)     |
|   Block Cloning (Fast Checkpoint)| NO                    | YES (Virtualization)   |
|   Sparse VDL (Rapid VHDX Init)   | NO                    | YES (Instant Zeroing)  |
|   Native File Compression        | YES (LZNT1)           | NO                     |
|   Encrypting File System (EFS)   | YES (File-level cert) | NO (Use BitLocker)     |
|   Disk Quotas                    | YES (User-based)      | NO (Use FSRM Quotas)   |
|   Hard Links / Extended Attr.    | YES                   | NO                     |
+-----------------------------------------------------------------------------------+

Resilient File System (ReFS) Core Innovations

ReFS was engineered from the ground up to address massive storage scalability, resilience against data corruption, and hyper-dense virtualization workloads.

1. Integrity Streams & Proactive Data Scrubbing:
   • ReFS utilizes 64-bit checksums for all file system metadata. Checksums for user file data can be enabled using Integrity Streams (Set-FileIntegrity).
   • When hosted on a mirrored or parity Storage Space, ReFS continuously monitors data integrity. When a read operation detects a corrupted checksum ("bit rot"), ReFS queries the alternate mirror copy, delivers uncorrupted data to the application, and transparently overwrites the damaged block on the failing drive.
   • A periodic background task (Data Scrubber) scans the entire volume proactively to repair latent corruption before applications request the affected blocks.
2. Block Cloning (Metadata Copy-on-Write):
   • In traditional NTFS, creating a Hyper-V VM checkpoint merge (merging a .avhdx differencing disk into a base .vhdx) requires reading gigabytes of data from one file and physically writing them to another, causing substantial disk I/O latency.
   • On ReFS, Block Cloning performs an instant metadata operation: it re-maps the logical cluster pointers of the destination file to point directly to the existing physical disk clusters of the source file. The physical data is never duplicated on disk until a subsequent write modifies a shared block (Copy-on-Write). This reduces VM checkpoint merge times from tens of minutes to sub-second operations.
3. Sparse Valid Data Length (Sparse VDL):
   • On NTFS, creating a fixed-size 1 TB .vhdx file requires physically writing 1 TB of zeros across the drive, locking disk throughput for several minutes.
   • On ReFS, Sparse VDL allows the file system to mark the newly allocated space as zeroed in metadata without writing physical zeros to disk, enabling instant creation of fixed-size VHDX files.

# Initialize Disk 1 with GPT and create an ReFS volume optimized for Hyper-V
Initialize-Disk -Number 1 -PartitionStyle GPT
$partition = New-Partition -DiskNumber 1 -UseMaximumSize -AssignDriveLetter
Format-Volume -Partition $partition -FileSystem ReFS -NewFileSystemLabel 'HyperV_Storage' -AllocationUnitSize 64KB

# Enable Integrity Streams on a specific folder or virtual machine directory
Set-FileIntegrity -FileName 'D:\Hyper-V\Virtual Hard Disks' -Enable $True

---