9.1 Azure VM Sizing, Series Selection & Managed Disk Tiers

Azure VM Sizing, Storage & Managed Disk Management

Administering Windows Server workloads in Microsoft Azure Infrastructure as a Service (IaaS) requires aligning workload compute and storage demands with Azure virtual machine architectures. Enterprise applications—such as Active Directory Domain Services (AD DS), Microsoft SQL Server, Internet Information Services (IIS), and enterprise file services—exhibit distinct CPU, memory, and disk input/output (I/O) profiles. Selecting the incorrect VM series, managed disk tier, or host caching strategy leads to severe performance bottlenecks, unnecessary licensing expenditure, or catastrophic database corruption.

1. Azure Virtual Machine Sizing & Series Selection for Windows Server

Azure organizes virtual machine sizes into families and series optimized for specific computational, memory, and storage profiles. The presence of the lowercase letter s in a VM size name (e.g., Standard_D4s_v5 vs. Standard_D4_v5) indicates support for Premium Storage (Premium SSD and Ultra Disk).

+-----------------------------------------------------------------------------+
|                   AZURE VM SERIES WORKLOAD TAXONOMY                         |
|                                                                             |
|   [D-SERIES: General Purpose]      [F-SERIES: Compute Optimized]            |
|   - Ratio: 1 vCPU : 4 GiB RAM      - Ratio: 1 vCPU : 2 GiB RAM              |
|   - Use: Domain Controllers,       - Use: Batch Processing, Web Servers,    |
|     File Servers, App Tiers          CPU-Bound Analytics                    |
|   - Examples: Dsv5, Dadsv5         - Examples: Fsv2                         |
|                                                                             |
|   [E/M-SERIES: Memory Optimized]   [L-SERIES: Storage Optimized]            |
|   - Ratio: 1 vCPU : 8 GiB RAM      - High-Throughput Direct Local NVMe      |
|   - Use: SQL Server OLTP, In-      - Use: NoSQL (Cassandra, MongoDB),       |
|     Memory Caches, Large ERP         Data Warehousing, High-Write Logs      |
|   - Examples: Esv5, Edsv5, M-series- Examples: Lsv2, Lsv3                   |
+-----------------------------------------------------------------------------+

Primary VM Families for Windows Server Workloads:

1. General Purpose (D-Series - Dsv5, Dadsv5):

   • Architecture: Features a balanced 1:4 vCPU-to-memory ratio running on 3rd Generation Intel Xeon Platinum or AMD EPYC processors.
   • Workload Alignment: Ideal for Active Directory domain controllers, enterprise file servers, small-to-medium relational databases, and general application hosting.
   • Note: Sizes with an a designate AMD processors; sizes with a d include a local temporary SSD.

2. Compute-Optimized (F-Series - Fsv2):

   • Architecture: Features a 1:2 vCPU-to-memory ratio powered by Intel Xeon Platinum 8168 processors with high Turbo clock speeds.
   • Workload Alignment: Designed for compute-heavy batch processing, web server farms with high request volume but low memory footprint, video encoding, and analytics nodes.

3. Memory-Optimized (E-Series - Esv5, Edsv5 & M-Series):

   • Architecture: Features a 1:8 vCPU-to-memory ratio. The M-series scales up to 128 vCPUs and 4.0 TiB of RAM (or Mv2 up to 12 TiB of RAM).
   • Workload Alignment: Mandatory for memory-intensive enterprise database workloads, including Microsoft SQL Server OLTP, SAP HANA, and large in-memory cache architectures.

4. Storage-Optimized (L-Series - Lsv2, Lsv3):

   • Architecture: Provides high-throughput, low-latency direct-mapped local NVMe solid-state storage directly attached to the host hypervisor.
   • Workload Alignment: Tailored for big data engines, distributed NoSQL databases (e.g., Apache Cassandra, MongoDB), and data warehousing solutions requiring extreme local disk throughput.

5. Constrained vCPU Sizes for Database Licensing:

   • Sizing models such as Standard_E4-2s_v5 or Standard_E8-4s_v5 constrain the active vCPU count to 2 or 4 while preserving the full memory (32 GiB / 64 GiB), network bandwidth, and storage I/O limits of the parent 4-vCPU or 8-vCPU chassis.
   • Cost Benefit: SQL Server core-based licensing costs are reduced significantly while maintaining the extensive RAM and storage throughput required for database operations.

2. Azure Managed Disks: Performance Tiers & Characteristics

Azure Managed Disks are block-level storage volumes managed by Microsoft that abstract underlying physical storage accounts. They provide 99.999% (five nines) availability, automatic local redundancy (LRS or ZRS), and automated alignment with VM Fault Domains.

+-----------------------------------------------------------------------------+
|                     AZURE MANAGED DISK TIERS COMPARISON                     |
|                                                                             |
|   TIER             MAX IOPS    MAX THROUGHPUT   TYPICAL LATENCY  BEST FOR   |
|   ---------------+-----------+----------------+----------------+------------|
|   Standard HDD   | 2,000     | 500 MB/s       | High (>10ms)   | Backup/Dev |
|   Standard SSD   | 6,000     | 750 MB/s       | Single-digit ms| Web/Light  |
|   Premium SSD v1 | 20,000    | 900 MB/s       | Low single-ms  | SQL/AD Prod|
|   Premium SSD v2 | 80,000    | 2,000 MB/s     | Sub-millisecond| IOPS Tuning|
|   Ultra Disk     | 400,000   | 10,000 MB/s    | Sub-millisecond| Extreme DB |
+-----------------------------------------------------------------------------+

Detailed Disk Tier Characteristics:

• Standard HDD: Backed by magnetic spinning platters. Suited exclusively for non-critical dev/test environments, infrequently accessed backup repositories, and archival workloads where cost is the primary driver.
• Standard SSD: Backed by solid-state drives. Delivers consistent entry-level performance for web frontends, lightweight application servers, and non-critical production workloads with moderate I/O demands.
• Premium SSD (v1): Fixed-tier architecture (P1 through P80). Performance scales strictly with provisioned disk capacity (e.g., P30 provides 1,024 GiB capacity, 5,000 IOPS, and 200 MB/s throughput). Recommended for standard production Windows Server workloads, Active Directory databases, and production SQL Server instances.
• Premium SSD v2: Advanced block storage enabling independent, granular provisioning of Disk Size (1 GiB–64 TiB), IOPS (up to 80,000), and Throughput (up to 2,000 MB/s) without paying for unwanted capacity. IOPS scale at 500 per GiB above a free 3,000 IOPS baseline, and throughput scales at 0.25 MB/s per provisioned IOPS above a free 125 MB/s baseline. Delivers consistent sub-millisecond latency. Does not support host caching; host caching must be set to None.
• Ultra Disk: The highest-performance Azure block storage. Scales to a maximum of 400,000 IOPS and 10,000 MB/s throughput per disk with sub-millisecond latency, sized from 4 GiB to 64 TiB. IOPS scale at 1,000 per GiB and throughput at 0.25 MB/s per provisioned IOPS. Ultra Disks allow administrators to dynamically adjust IOPS and throughput on the fly without detaching the disk or restarting the virtual machine.
  • Constraints: Ultra Disks cannot be used as OS disks, do not support host caching, availability sets, or zone-redundant storage, and require the VM to be provisioned with Ultra Disk compatibility enabled. Performance can be changed at most four times in any 24-hour window.