WS-013 Azure

Stack HCI

© Copyright Microsoft Corporation. All rights reserved.

Module 3: Planning for
and implementing Azure
Stack HCI storage
Module overview

This module describes how to plan for and implement Azure Stack HCI Storage
 Lessons:
o Overview of Azure Stack HCI storage core technologies

o Planning for Storage Spaces Direct in Azure Stack HCI

o Implementing a Storage Spaces Direct-based hyperconverged infrastructure

o Managing Storage Spaces Direct in Azure Stack HCI

o Planning for and implementing Storage QoS

o Planning for and implementing Storage Replica

Lesson 1: Overview of
Azure Stack HCI storage
core technologies
Lesson 1 overview

This lesson describes the three primary mechanisms Azure Stack HCI uses for storage
Topics:
 Non-volatile storage technologies
 File system technologies
 Overview of Storage Spaces Direct
Non-volatile storage technologies

HDD SSD NVMe PMEM

Price per GB
Throughput/IOPS
MTBF
Latency
Max capacity
File system technologies (1 of 5)

ReFS capabilities (vs NTFS)

Feature ReFS NTFS

ACL-based security Yes Yes
Volume-level encryption (BitLocker) Yes Yes
Volume-level deduplication (Data Yes Yes
Deduplication)
CSV support Yes Yes
Thin provisioning Yes (Storage Spaces only) Yes
Trim/unmap Yes (Storage Spaces only) Yes
Block clone Yes No
Sparse VDL Yes No
Mirror-accelerated parity Yes (Storage Spaces Direct No
only)
File system technologies (2 of 5)

NTFS capabilities (vs. ReFS)

Feature ReFS NTFS

File-level compression No Yes
File-level encryption No Yes
Disk quotas No Yes
Boot volume No Yes
Page file support No Yes
Removable media support No Yes
Extended file system attributes No Yes
File system technologies (3 of 5)

 CSV:
o Serve as a general-purpose clustered file system layered above NTFS and ReFS
o Provide simultaneous read-write access to the same volume from multiple cluster
nodes:
• Each node can independently read from and write to individual files on the volume
• A single node functions as the CSV owner of the volume and handles metadata
orchestration
o Enables faster failover between cluster nodes by:
• Eliminating the need for volume dismount/mount
• Automatically balancing distribution of CSV ownership across nodes
 CSV-supported workloads include:
o Clustered VMs (highly available disk files)
o Scale-out file shares for application data (Hyper-V VM disk files and SQL Server data
in disaggregated scenarios)
 CSV cache:
o Provides block-level caching of read-only unbuffered I/O operations
File system technologies (4 of 5)

 Data deduplication:
o Is available as a Windows Server role service
o Is enabled on a per-volume basis
o Provides deduplication and compression
o Increases usable capacity by:
• Scanning the file system for files meeting the optimization policy
• Breaking files into variable-size chunks
• Identifying unique chunks
• Placing chunks in the chunk store and optionally compressing them
• Replacing parts of the original files with reparse points to the chunk store
o Uses a post-processing approach that:
• Optimizes data on-disk (and does not interfere with writes)
• Runs as scheduled background jobs
• Supports configurable priority based on usage type
File system technologies (5 of 5)

Deduplication usage type Intended workload Configuration settings

Default General purpose file servers Background optimization
Default optimization policy:
• Minimum file age = 3 days
• Optimize in-use files = no
• Optimize partial files = no
Hyper-V VDI Background optimization
Default optimization policy:
• Minimum file age = 3 days
• Optimize in-use files = yes
• Optimize partial files = yes
Backup Virtualized backup apps Priority optimization
(for example, DPM) Default optimization policy:
• Minimum file age = 0 days
• Optimize in-use files = yes
• Optimize partial files = no
Overview of Storage Spaces Direct (1 of 3)
Hyper-V VMs
 Azure Stack HCI Storage Spaces Direct
architecture:
o Storage Spaces Direct workloads (VMs or
scale-out storage) CSV (ReFS) C:\ClusterStorage\
o CSV (including CSV cache) Volumes
Storage spaces
o ReFS-formatted volumes
Cluster
o Storage Spaces
Ethernet +
o Failover Clustering RDMA

o SMB networking

• SMB Direct Serve Serve Serve

r r r
• SMB Multichannel Software storage bus
o Network hardware (RDMA, interconnects)
Storage Pool
o Software Storage Bus (including SB layer
cache)
Local HDD + SSD Local HDD + SSD Local HDD + SSD
o Storage pool (single)

o
Overview of Storage Spaces Direct (2 of 3)

 Software-defined storage components:

o Storage Spaces:
• Use physical disks combined into a storage pool to virtualize storage
• Combine disks into pools (single pool in Azure Stack HCI)
• Enable you to create:
▫ Virtual disks from free space in a disk pool
▫ Volumes from the virtual disks
o Software Storage Bus (including SB layer cache):
• Spans a cluster and establishes a software-defined storage fabric
• Includes local drives of cluster nodes
• Forms a storage-based cache by automatically binding faster and slower drives
Overview of Storage Spaces Direct (3 of 3)

 Network hardware (RDMA, interconnects):

o RDMA provides high-throughput, low-latency communication that minimizes CPU
usage
o RDMA is available as:

• RoCE and RoCEv2 (requires relatively complex switch/server configuration)

• Infiniband (requires proprietary switches)
• iWarp (auto-configured, works over standard TCP network infrastructure,
recommended)
o Interconnects depend on Azure Stack HCI hardware:

• Switched (requires network switches)

• Switchless (requires full mesh of direct connections between nodes)
 SMBv3-related components:
o SMB Direct (optimizes the use of RDMA for SMB traffic, enabled by default)

o SMB Multichannel (required to enable SMB Direct, automatically detects and uses
multiple network connections)
Lesson 1: Test your knowledge

Refer to the Student Guide for review questions

Lesson 2: Planning for
Storage Spaces Direct in
Azure Stack HCI
Lesson 2 overview

Azure Stack HCI optimizes provisioning of hyperconverged infrastructures but still requires a
significant amount of planning necessary to maximize resiliency, capacity, and performance
of Storage Spaces Direct components
This lesson explains how to plan for an optimal Storage Spaces Direct configuration in Azure
Stack HCI
 Topics:
o Plan for Storage Spaces Direct

o Choose drives

o Drive symmetry considerations

o Plan for cluster and pool quorums

o Plan for volumes

Plan for Storage Spaces Direct (1 of 3)

Performance

Storage
Spaces
Direct

Resiliency Capacity
Plan for Storage Spaces Direct (2 of 3)

 General Azure Stack HCI planning considerations that affect storage:

o Number of physical servers per cluster
o Number of fault domains per cluster
o Number and type of processors per server
o Amount and type of memory per server
o Number and types of disks
o Storage resiliency levels
o Tiering and caching configuration
o Volume performance, including the corresponding tiering and caching configuration
o Number and type of network adapters per server
o Number and type of network switches per cluster
Plan for Storage Spaces Direct (3 of 3)

 General Azure Stack HCI Storage Spaces Direct planning considerations:

o Azure Stack HCI supports directly attached SATA, SAS, NVMe, and PMEM drives
o Each cluster node should use the same model, size, number of drives, and drive
sector size
o Drives can be internal to the server or in an external enclosure, providing they're
connected to only an individual cluster node
• RAID controllers and MPIOs are not supported
• HBAs are allowed if they are configured in pass-through mode
Choose drives (slide 1 of 5)

 Cache sizing considerations:

o Each Storage Spaces Direct drive is assigned to either cache or capacity
o Drive assignments occur automatically if the cluster contains at least two types of
drives
o You have the option to configure the cache manually
o Sum of sizes of all capacity drives equals the total raw capacity of the cluster
o Binding between cache and capacity drives occurs dynamically:
• The number of capacity drives should equal a multiple of the number of cache
drives
• The size of the cache should accommodate the working set of Storage Spaces
Direct workloads
• The total size of capacity drives per server shouldn't exceed 400 TB (due to resync
implications)
 Cache behavior considerations:
o Cache behavior occurs automatically depending on the type of capacity drives:
• Write only: SSD capacity drives (including a mix of SSD and HDD capacity drives)
Choose drives (slide 2 of 5)

Deployment Cache drives (default) Capacity Cache behavior

drives (default)
All NVMe None (can be configured manually) NVMe Write only (if
configured)
All SSD None (can be configured manually) SSD Write only (if
configured)
NVMe + SSD NVMe SSD Write only
NVMe + HDD NVMe HDD Read + Write
SSD + HDD SSD HDD Read + Write
NVMe + SSD + NVMe SSD + HDD Read + Write for HDD
HDD Write only for SSD
Choose drives (slide 3 of 5)
All-flash drives, maximized for performance via automatic configuration:

1 O 2 OR 3
R
NVMe

NVMe NVMe
NVMe

NVMe

NVMe

SSD SSD SSD SSD SSD SSD SSD SSD

NVMe for capacity NVMe for cache SSD for capacity SSD for capacity
 Choose drives (slide 4 of 5)
Hybrid configuration that balances performance and capacity:

1 O 2 OR 3
R SSD SSD

NVMe NVMe NVMe NVMe

HDD HDD

HDD HDD HDD HDD HDD HDD HDD HDD SSD HDD SSD HDD

NVMe for cacheHDD for capacity SSD for cacheHDD for capacity NVMe for cache
SSD+HDD for capacity
 Choose drives (slide 5 of 5)
Manual configurations with dual-flash configurations, leveraging slower flash
drives to increase capacity:
1 O 2 OR 3
R
SSD SSD SSD SSD
1 1 1 1

NVMe1 NVMe1

NVMe2 NVMe2 NVMe2 NVMe2 SSD SSD SSD SSD2 SSD HDD SSD2 HDD
2 2 2 2
NVMe1 for cache
NVMe2 for capacity SSD1 for cache
SSD2 for capacity SSD1 for cacheSSD2+HDD for capacity
Drive symmetry considerations (slide 1 of 3)

 Storage Spaces Direct works optimally when every server has the exact same drives
 When implementing and maintaining Storage Spaces Direct:
o Each cluster node should have the same types of drives (capacity and cache)
o Each cluster node should have the same number of drives of each type (capacity and
cache)
o For the individual storage type (capacity and cache), the respective drive models and
firmware versions should match, whenever possible
o For the individual storage type (capacity and cache), the respective drive sizes should
match whenever possible
 Storage Spaces Direct automatically handles capacity imbalances across drives and across
servers
 Mismatches in drive sizes might result in stranded capacity
Drive symmetry considerations (slide 2 of 3)
Stranded capacity:

Stranded capacity
5TB

A A’ B”

B B’ C’

Server 1 Server 2 Server 3

10TB 10TB 15TB

Not all capacity can be used

Drive symmetry considerations (slide 3 of 3)
Optimal placement:

A“

A A’ B’ B”

B C C’ C”

Server 1 Server 2 Server 3 Server 4

10TB 10TB 10TB 15TB

All capacity can be used

Plan for cluster and pool quorums (slide 1 of 5)

 Quorum determines the number of simultaneous failures a highly available entity can
survive
 Quorum considers the number votes associated with resources that form the highly available
entity:
o Cluster nodes (for cluster quorum)
o Clustered disks (for pool quorum)

 Storage Spaces Direct includes two quorum mechanisms:

o Cluster quorum:
• Operates on the cluster level based on votes from nodes and a witness
• Supports dynamic witness (Storage Spaces Direct can use Cloud or File Share
witness) and dynamic quorum
• Requires witness for two-node clusters, witness recommended for three- and four-
node clusters
o Pool quorum:
• Operates on the storage pool level based on votes from nodes and storage
resiliency
• Uses the node that owns the storage pool as the witness (no dynamic witness
Plan for cluster and pool quorums (slide 2 of 5)

Cluster quorum:
Witness
 Server node 2 fails, cluster Vote Vote Vote Vote Vote
stays up Server Server Server Server
 Server node 3 fails, cluster node 1 node 2 node 3 node 4

stays up

Witness
Vote Vote Vote Vote

Server Server Server Server

node 1 node 2 node 3 node 4

Witness
Vote Vote Vote

Server Server Server Server

node 1 node 2 node 3 node 4
Plan for cluster and pool quorums (slide 3 of 5)

Pool quorum:
 Server nodes 3 and 4 go
Vote Vote Vote Vote
down, pool stays up
Server Server Server Server
 A disk in node 2 goes node 1 node 2 node 3 node 4
down, pool goes down
(simultaneous failure)
Storage pool

Vote Vote Vote Vote

Server Server Server Server

node 1 node 2 node 3 node 4

Storage pool
Plan for cluster and pool quorums (slide 4 of 5)

Plan for cluster quorums:

Cluster Can survive 1 node Can survive 2 subsequent Can survive 2 concurrent
nodes failure node failures node failures
2 50/50 No No
2 + Witness Yes No No
3 Yes 50/50 No
3 + Witness Yes Yes No
4 Yes Yes 50/50
4 + Witness Yes Yes Yes
5 and above Yes Yes Yes
Plan for cluster and pool quorums (slide 5 of 5)

Plan for pool quorums:

Cluster Can survive 1 node Can survive 2 subsequent Can survive 2 concurrent
nodes failure node failures node failures
2 No No No
2 + Witness Yes No No
3 Yes No No
3 + Witness Yes No No
4 Yes No No
4 + Witness Yes Yes Yes
5 and above Yes Yes Yes
Plan for volumes (slide 1 of 3)

Volume planning considerations include:

 Number of volumes per cluster: C:\ C:\ C:\
ClusterStorage\ ClusterStorage\ ClusterStorage\
o Up to 64 Volume1 Volume2 Volume3

o A multiple of cluster nodes

 Volume size:
o Up to 64 TB

o Corresponding footprint

 Reserve capacity Volume Volume Volume

1 2 3
 File system:
o NTFS

o ReFS

 Resiliency type
Storage pool
Plan for volumes (slide 2 of 3)

 Resiliency relies on the concept of fault domains

 The choice of resiliency has implications on performance and effective capacity
 Resiliency and, by implication, performance and effective capacity depend on:
o Number of cluster nodes
o Resiliency type

 Resiliency applies to:

o Capacity drives

o Cache drives (data in the cache benefits from the same resiliency as capacity drives)

o Storage pool metadata

 Additional resiliency considerations include:

o Reserved capacity (one capacity drive per server, up to four drives)

o Tier templates (manually-created)

o Delimiting the allocation of volumes (that is, scoped volumes)

Plan for volumes (slide 3 of 3)

Resiliency type Capacity Performance Failure Numbe

efficienc tolerance r of
y nodes
Two-way mirror 50% Best for random reads/writes 1 node or 1 drive 2+
Three-way mirror 33% Best for random reads/writes 1 node and 3+
drives (4)
Mirror-accelerated 33-80% Slower than mirror Up to 2 nodes 4+
parity (1)
Faster than dual parity
Dual parity 50-80% Highest I/O latency and CPU Up to 2 nodes 4+
(2)
usage on writes
Best for large sequential
reads/writes
Nested two-way 25% Best for random reads/writes 2 drives or 2
mirror 1 node and 1
(1)
Depends on the number of nodes and the mirror-parity ratio per volume drive
(2)
Depends
Nested on the number35-40%
mirror- of nodes Slower than nested two-way 2 drives or 2
(3)
Depends onparity
accelerated the number of capacity
(3) drives per node and the mirror-to-parity1ratio
mirror node per volume
and 1
(4)
As long as the drives are on a single surviving node drive
Lesson 2: Test your knowledge

Refer to the Student Guide for review questions

Lesson 3: Implementing
a Storage Spaces Direct-
based hyperconverged
infrastructure
Lesson 3 overview

In this lesson, you will learn about the implementation phases for Storage Spaces Direct in
Azure Stack HCI
 Topics:
o Prerequisites for Azure Stack HCI deployment

o Deploy Storage Spaces Direct in Azure Stack HCI

o Validate Storage Spaces Direct deployment

o Demonstration: Test Storage Spaces Direct by using synthetic workloads

Prerequisites for Azure Stack HCI deployment

 Verify hardware requirements:

o Servers
o Processors

o Memory

o Boot device

o Network interface cards

o Switched or switchless node interconnects

o Storage

 Gather infrastructure-specific information:

o Naming convention for server

o AD DS domain names and domain join procedure

o VLAN IDs
Deploy Storage Spaces Direct in Azure Stack HCI (1 of 2)

1. Assign custom computer names to Windows Server 2019 hosts running on Azure Stack HCI
nodes
2. Join the Windows Server 2019 hosts to an AD DS domain
3. Add Windows Server 2019 roles and features to each host operating system:
o Hyper-V
o Failover Clustering
o RSAT-Clustering-PowerShell
o Hyper-V PowerShell
o Data Center Bridging (optional when using iWARP)
4. Configure network connectivity (SET and RDMA)
5. Configure Storage Spaces Direct:
o Setup Failover Clustering, including the Witness
o Enable Storage Spaces Direct
6. Optionally, deploy SDN
7. If applicable, follow OEM-specific configuration to complete the initial deployment
Deploy Storage Spaces Direct in Azure Stack HCI (2 of 2)

Deploy an Azure Stack HCI

cluster by using:
 Windows Admin Center
 Windows PowerShell
 System Center VMM
Validate Storage Spaces Direct deployment

 Validation of cluster configuration:

o Uses Failover Clustering built-in validation tools
 Validation of DCB configuration:
o Uses the Validate-DCB tool available as a NuGet-based package

 Validation of cluster performance:

o Uses synthetic workloads generated by VM Fleet available on GitHub

o Requires disabling CSV Cache during performance testing

Demonstration:
Test Storage
Spaces Direct by
using synthetic
workloads
 Validate cluster performance by using VM
Fleet-based synthetic workloads
Lesson 3: Test your knowledge

Refer to the Student Guide for review questions

Instructor-led lab A:
Implementing a
Storage Spaces
Direct cluster by
using Windows
PowerShell
 Implementing a Storage Spaces Direct
cluster by using Windows PowerShell
Lab A scenario

One of your objectives is to minimize the effort associated with deployment and management
of on-premises resources. As part of this effort, you want to test the process of implementing
a Storage Spaces Direct cluster in an automated manner by using Windows PowerShell.
Lab A: Implementing a Storage Spaces Direct cluster by using
Windows PowerShell
Exercise 1: Implementing a Storage Spaces Direct cluster by using Windows PowerShell
Lesson 4: Managing
Storage Spaces Direct in
Azure Stack HCI
Lesson 4 overview

This lesson covers managing the Storage Spaces Direct disk technologies that Azure Stack
HCI uses, including the use of Storage Spaces Direct volumes, and deduplication and
compression
 Topics:
o Create volumes

o Extend volumes

o Demonstration: Extend volumes

o Manage deduplication and compression

o Demonstration: Manage deduplication and compression

o Scale Storage Spaces Direct

o Troubleshoot Storage Spaces Direct and replace failed disks

Create volumes (1 of 3)

Use Windows Admin Center to:

 Create a three-way mirror
volume
 Create a mirror-accelerated
parity volume
 Manage files and folders
Create volumes (2 of 3)

In Windows PowerShell:
 Create volumes with the default resiliency settings:
o Run New-Volume with the ResiliencySettingName parameter specifying Mirror
or Parity
New-Volume -FriendlyName "Volume1" -FileSystem CSVFS_ReFS -
StoragePoolFriendlyName S2D* -Size 1TB -ResiliencySettingName Mirror
New-Volume -FriendlyName "Volume2" -FileSystem CSVFS_ReFS -
StoragePoolFriendlyName S2D* -Size 1TB -ResiliencySettingName Parity
 Create volumes with default storage tier templates:
o Run New-Volume with the StorageTierFriendlyNames parameter specifying
Performance and Capacity tier templates
New-Volume -FriendlyName "Volume3" -FileSystem CSVFS_ReFS -
StoragePoolFriendlyName S2D* -StorageTierFriendlyNames Performance, Capacity -
StorageTierSizes 300GB, 700GB
Create volumes (3 of 3)

In Windows PowerShell:
 Create volumes with nested storage tier templates:
o First run New-StorageTier with the MediaType and StorageTierFriendlyNames
parameters
New-StorageTier -StoragePoolFriendlyName S2D* -FriendlyName NestedMirror
-ResiliencySettingName Mirror -MediaType HDD -NumberOfDataCopies 4
New-StorageTier -StoragePoolFriendlyName S2D* -FriendlyName NestedParity
-ResiliencySettingName Parity -MediaType HDD -NumberOfDataCopies 2
-PhysicalDiskRedundancy 1 - NumberOfGroups 1 -FaultDomainAwareness
StorageScaleUnit -ColumnIsolation PhysicalDisk
o Then run New-Volume with the StorageTierFriendlyNames parameter
New-Volume -StoragePoolFriendlyName S2D* -FriendlyName Volume02
-StorageTierFriendlyNames NestedMirror, NestedParity -StorageTierSizes 100GB,
400GB
Extend volumes

1. Ensure there is enough capacity in the storage pool to accommodate the extended
footprint
2. Use either of the following:
o Windows Admin Center:

• Use the Resize option on the volume's details page

o Windows PowerShell:
Cluster Shared Volume
a) Check whether the virtual disk
uses storage tiers: Volume
▫ If no, resize it directly
Partition
▫ If yes, resize tiers
Volume
Disk
b) Resize the partition
Virtual disk

Storage tier Storage tier

Demonstration:
Extend volumes
 Connect to a Storage Spaces Direct
cluster
 Create volumes
 Extend volumes
Manage deduplication and compression

1. Ensure that the Data Deduplication Windows Server role service is installed on all cluster
nodes
2. Use any of the following:
o Windows Admin Center:

• Enable, disable, and select one of three built-in profiles

• Monitor deduplication status
o Windows PowerShell:

• Manage configuration of volume deduplication, including exclusions and custom

settings
• Retrieve deduplication status
• Manage deduplication schedule
• Manage and monitor data deduplication jobs
o Windows Server Manager:

• Manage configuration of volume deduplication, including exclusions and custom

settings
Demonstration:
Manage
deduplication and
compression
 Install the Data Deduplication role service
on cluster nodes
 Enable deduplication and compression on
Storage Spaces Direct volumes
Scale Storage Spaces Direct (slide 1 of 3)

General considerations when adding servers (scaling out):

 Benefits:
o Increase storage and compute capacity

o Can improve performance and storage efficiency

 Procedure:
1. Run cluster validation using:
• Windows Admin Center, Failover Cluster Manager, or Test-Cluster
2. Add cluster node using:
• Windows Admin Center, Failover Cluster Manager, or Add-ClusterNode
Scale Storage Spaces Direct (slide 2 of 3)

Volume-specific considerations when adding servers (scaling out):

 Scaling from two to three nodes
o Typically involves migrating from two-way mirror to three-way mirror

o In-place conversion is not supported

 Scaling from three to four nodes

o Might involve migrating from three-way mirror to dual parity or mirror-accelerated
parity
o In-place conversion is not supported

 Scaling out beyond four nodes

o Offers capacity efficiency benefits due to enhanced parity encoding

o The efficiency benefits apply to new volumes only

Custom fault domain-specific considerations when adding servers (scaling out):

 Fault domain assignment needs to take place before adding a server to the cluster
Scale Storage Spaces Direct (slide 3 of 3)

Adding drives (scaling up):

 Benefits:
o Increased storage capacity

o Can improve performance

 Procedure:
1. Add physical disks to each server

2. Run Get-PhysicalDisks to verify that the CanPool property is set to True

3. Storage Spaces Direct automatically optimizes drive usage:

• Optimization begins 15 minutes after a drive is added to the pool
• Relies on the Optimize and Rebalance background jobs
• To monitor job status, run Get-StorageJob
Troubleshoot Storage Spaces Direct and replace failed disks (1 of
3)
The Windows Server built-in Health Service component simplifies maintenance by providing:
 Fault detection and automatic remediation (via actions):
o Disk lifecycle events

o Retiring disks

o Restoring resiliency

o Disk pooling

 Reporting (via Get-StorageHealthReport):

Get-StorageSubSystem Cluster* | Get-StorageHealthReport -Count <Count>
Get-Volume -FileSystemLabel <Label> | Get-StorageHealthReport -Count
<Count>
Get-StorageNode -Name <Name> | Get-StorageHealthReport -Count <Count>
 Enforcement of hardware compliance (via Supported Components Document):
Enable-ClusterS2D -XML $MyXML
Get-StorageSubSystem Cluster* | Set-StorageHealthSetting -Name
"System.Storage.SupportedComponents.Document" -Value $MyXML
Troubleshoot Storage Spaces Direct and replace failed disks (2 of
3)
The Windows Server 2019 built-in Performance History feature simplifies monitoring by
providing:
 Automatic data collection for Storage Spaces Direct objects, including:
o Clusters, cluster nodes, drives, virtual disks, volumes, network adapters, and VMs

 Automatic data retention for up to 400 days, with diminishing granularity

 Support for data visualization and analysis by using:
o Windows Admin Center

o Get-ClusterPerformanceHistory

The Windows Server 2019 installable System Insights feature identifies future resource needs
by providing:
 Automatic collection of metrics and events
 Predictive analytics based on collected data
 Integration with Azure Monitor and System Center Operations Manager
Troubleshoot Storage Spaces Direct and replace failed disks (3 of
3)
Start troubleshooting with the following steps:
1. Run cluster validation and focus on the Storage Space Direct section
2. Confirm that storage components are part of the Azure Stack HCI Catalog offering
3. Inspect the storage for any faulty drives and replace them
4. Update storage and drive firmware if necessary
5. Update network adapter drivers and firmware
Review additional troubleshooting scenarios if the problem remains unresolved:
 Virtual disk resources with the No Redundancy operational status
 “Unsupported media type” error message when enabling Storage Spaces Direct
 Slow I/O performance
 Slow file copy
Lesson 4: Test your knowledge

Refer to the Student Guide for review questions

Instructor-led lab B:
Managing storage of a
Storage Spaces Direct
cluster by using
Windows Admin Center
and Windows
PowerShell
 Managing storage of a Storage Spaces
Direct cluster by using Windows Admin
Center and Windows PowerShell
Lab B scenario

Now that you have provisioned a Storage Spaces Direct cluster in an automated manner by
using Windows PowerShell, you want to determine whether you can minimize administrative
effort associated with remediating disk failures within a Storage Spaces Direct cluster by
leveraging its resiliency and self-healing capabilities
Lab B: Managing storage of a Storage Spaces Direct cluster by
using Windows Admin Center and Windows PowerShell
Exercise 1: Managing storage of a Storage Spaces Direct cluster by using Windows Admin
Center and Windows PowerShell
Instructor-led lab C:
Managing and
monitoring resiliency
of a Storage Spaces
Direct cluster
 Managing and monitoring resiliency of a
Storage Spaces Direct cluster
Lab C scenario

You want to examine resiliency in situations when there are simultaneous cluster node and
drive failures. You want to understand how resiliency can protect cluster stability and integrity.
To start, you will create tiered volumes and test volume, disk, and cluster resiliency.
Lab C: Managing and monitoring resiliency of a Storage Spaces
Direct cluster
Exercise 1: Managing and monitoring resiliency of a Storage Spaces Direct cluster
Instructor-led lab
D: Managing
Storage Spaces
Direct cluster
tiers
 Managing Storage Spaces Direct cluster
tiers
Lab D scenario

Now that you know more about cluster resiliency, you want to explore additional provisions
that could help you optimize storage capacity and performance. To accomplish this, you will
configure and evaluate storage tiers, including tiers that involve nested resiliency.
Lab D: Managing Storage Spaces Direct cluster tiers

Exercise 1: Managing Storage Spaces Direct cluster tiers

Lesson 5: Planning for
and implementing
Storage QoS
Lesson 5 overview

QoS is a set of networking and storage technologies that allow you to control the flow of
network traffic based on its characteristics so that you can optimize system functionality and
workload performance. This lesson covers Storage QoS in the context of Azure Stack HCI.
 Topics:
o QoS

o Configure QoS policies

o Monitor and troubleshoot QoS policies

QoS

QoS helps you address performance requirements of workloads that rely on shared
infrastructure
Azure Stack HCI relies on Network QoS to provide the following functionality:
 Bandwidth management
 Classification and tagging
 Priority-based flow control
 QoS policies
In Azure Stack HCI clusters, Network Storage Spaces Direct QoS leverages RDMA networking:
 RoCE over UDP/IP with DCB providing flow control and congestion management
 iWARP over TCP/IP, with TCP providing flow control and congestion management
DCB relies on PFC to prioritize traffic based on a class assigned to a payload, such as:
 SMB Direct
 Failover clustering traffic
Configure QoS policies (1 of 2)

1. Install the Data Center Bridging Windows Server feature:

Install-WindowsFeature –Name Data-Center-Bridging -ComputerName
'node1','node2'
2. Configure DCB willing bit:
Set-NetQosDcbxSetting -Willing $false
3. Create QoS rules:
New-NetQosPolicy -Name 'Default' -Default -PriorityValue8021Action 0
New-NetQosPolicy -Name 'SMB' -SMB -PriorityValue8021Action 3
New-NetQosPolicy -Name 'SMB Direct' –NetDirectPort 445 –Priority 3
New-NetQosPolicy -Name 'Cluster' -Cluster -PriorityValue8021Action 7
Configure QoS policies (2 of 2)

4. Configure DCB flow control:

Disable-NetQosFlowControl -Priority 0,1,2,4,5,6,7
Enable-NetQosFlowControl -Priority 3
5. Set the minimum bandwidth:
New-NetQosTrafficClass 'SMB' -Priority 3 -BandwidthPercentage 60 -Algorithm
ETS
New-NetQosTrafficClass 'Cluster' -Priority 7 -BandwidthPercentage 5 -
Algorithm ETS
6. Enable Network QoS on RDMA adapters:
Get-NetAdapterQoS | Enable-NetAdapterQoS
Monitor and troubleshoot QoS policies

Use cluster validation:

 Run cluster validation with:
o Windows Admin Center

o Windows PowerShell (Test-Cluster)

o Failover Cluster Manager

 Review the Validate QoS Settings Configuration section

Use Windows PowerShell QoS cmdlets:

 Get-NetQosDcbxSetting
 Get-NetQosPolicy
 Get-NetQosFlowControl
 Get-NetQosTrafficClass
 Get-NetAdapterQoS
Lesson 5: Test your knowledge

Refer to the Student Guide for review questions

Lesson 6: Planning for
and implementing
Storage Replica
Lesson 6 overview

Storage Replica is a Windows Server technology that enables replication of volumes between
servers or clusters for the purpose of disaster recovery. This lesson covers functionality,
architecture, configuration options, implementation, monitoring, and troubleshooting of
Storage Replica in the context of Azure Stack HCI.

 Topics:
o Storage Replica features
o Azure Stack HCI and Storage Replica
o Implement Storage Replica
o Demonstration: Implement Storage Replica
o Monitor and troubleshoot Storage Replica
Storage Replica features (slide 1 of 2)

 Storage Replica is a volume protection mechanism based on replication:

o Unidirectional
o Storage agnostic Applications
(Primary)
o Between servers or clusters

o Synchronous or asynchronous
1 5
 Storage Replica topologies:
o Stretch cluster
3
o Cluster-to-cluster Server Server
Cluster (SR) Cluster (SR)
4
o Server-to-server

o Server-to-self t
t 2 3
1

Database Log Database Log

Storage Replica features (slide 2 of 2)

Storage Replica includes the following primary features:

 Block-level replication
 Simple deployment and management
 Guest and host
 SMBv3-based
 Security
 High performance initial sync
 Consistency groups
 User delegation
 Network constraints
Azure Stack HCI and Storage Replica (1 of 2)

Azure Stack HCI supports Storage Replica’s cluster-to-cluster topology

Prerequisites include:
 The same or trusted AD DS forest
 2 GB of RAM and two CPU cores per server
 One or more Ethernet connections per server (including RDMA)
 Bi-directional network connectivity via ICMP, SMB (TCP 445 and 5445) and WS-MAN (5895)
 Round-trip latency within 5 ms (when using synchronous replication)
 Two sets of volumes (one for data and the other for logs) on the source and destination
cluster) with the following settings:
o All volumes are based on GPT-initialized disks
o All volumes are formatted with ReFS or NTFS
o Each volume type (data and log) on both clusters has matching sizes and sector
sizes
o Log volumes are based on flash-based storage
o Log volumes are not used for any other workload
Azure Stack HCI and Storage Replica (2 of 2)

Caching considerations:
 Storage Replica in Storage Spaces Direct clusters might result in increased latency in the
following scenarios:
o HDD capacity tier with NVMe-based cache
o HDD capacity tier with SSD-based cache
 To remediate this issue:
o Use a mix of NVMe and SSD drives (rather than HDD)
o Configure NVMe and SSD drives as performance and capacity tiers, respectively
o Place Storage Replica log volume on the performance tier
o Place Storage Replica data volume on the capacity tier

Deduplication considerations:
 Storage Replica supports Data Deduplication
 To implement it:
o Install Data Deduplication on both the source and destination servers
o Enable deduplication on the data volume on the source server only
Implement Storage Replica (1 of 2)

In Windows Admin Center:

1. Identify the source volume
2. Create the destination volume and log
volumes
3. Install on the source and destination
servers:
o Storage Replica role service

o File Server role service (required to

run Test-SRTopology)
4. Run Test-SRTopology
5. Create Storage Replica
Implement Storage Replica (2 of 2)

In Windows PowerShell:
1. Identify the source volume
2. Create the destination volume and log volumes
3. Install on the source and destination servers:
o Storage Replica role service

o File Server role service (required to run Test-SRTopology)

4. Run Test-SRTopology
5. Grant the first cluster full access to the second cluster
6. Grant the second cluster full access to the first cluster
7. Create a Storage Replica partnership
Demonstration:
Implement
Storage Replica
 Create source volumes
 Create destination volumes
 Use Windows Admin Center to implement
Storage Replica
Monitor and troubleshoot Storage Replica

Monitoring:
 Performance: \\Storage Replica Partition I/O Statistics(*) and \Storage Replica
Statistics(*)
 Status via Get-SRGroup and Get-SRPartnership
 Initial replication via Event 1237 message in Storage Replica Admin event log or via Get-
SRGroup
 Replication via the Storage Replica Statistics\Total Bytes Received counter on the
destination server
Troubleshooting:
 Use Test-SRTopology whenever applicable
 Investigate generic error messages displayed by Test-SRTopology, which might indicate
that:
o You are logged on to the source server as a local user (rather than a domain user)
o You specified an incorrect name or IP address of the destination server
o The destination server firewall is blocking access of Windows PowerShell cmdlets
o The destination server is not running the WMI service
Lesson 6: Test your knowledge

Refer to the Student Guide for review questions

Instructor-led lab E:
Identifying and
analyzing metadata
of a Storage Spaces
Direct cluster
(optional)
 Identifying and analyzing metadata of a
Storage Spaces Direct cluster
Lab E scenario

Another resiliency consideration you want to explore is metadata of the storage pool and its
components. You want to ensure that you understand resiliency provisions that must be taken
into account to protect cluster stability and integrity. You also want to be able to identify how
a Storage Spaces Direct cluster maintains information about its data.
Lab E: Identifying and analyzing metadata of a Storage Spaces
Direct cluster (optional)
Exercise 1: Identifying and analyzing metadata for a Storage Spaces Direct cluster
Module-review questions (slide 1 of 2)

1. Which of the following describes SSDs?

a. Contain platters and read/write heads
b. Represent an example of "Storage Class Memory"
c. Offer higher ​input/output (I/O) operations per second (IOPS) and I/O throughput than
any other non-volatile storage technology
d. Use integrated circuit assemblies to persistently store data
2. Which of the following describes HDDs?
a. Contain platters and read/write heads
b. Represent an example of "Storage Class Memory"
c. Offer higher ​input/output (I/O) operations per second (IOPS) and I/O throughput than
any other non-volatile storage technology
d. Use integrated circuit assemblies to persistently store data
Module-review questions (slide 2 of 2)

3. Which resiliency type provides the best performance for random read/writes and highest
resiliency on a two-node Storage Spaces Direct cluster?
a. Two-way mirror
b. Nested two-way mirror
c. Mirror-accelerated parity
d. Nested mirror-accelerated parity
Module-review answers

1. Which of the following describes solid state drives (SSDs)?

d. Use integrated circuit assemblies to persistently store data
2. Which of the following describes hard disk drives (HDDs)?
a. Contain platters and read/write heads
3. Which resiliency type provides the best performance for random read/writes and highest
resiliency on a two-node Storage Spaces Direct cluster?
b. Nested two-way mirror
References (slide 1 of 2)

For more information, refer to the following links:

 Understanding the cache in Storage Spaces Direct
 Delimit the allocation of volumes in Storage Spaces Direct
 Nested resiliency for Storage Spaces Direct
 Find Azure Stack HCI solutions from your preferred hardware vendor
 Hyper-V VM Fleet for S2D
 Validate an Azure Stack HCI cluster
 Fun with VMFleet
 S2D Hyperconverged
 Windows Admin Center
 Install and enable Data Deduplication
References (slide 2 of 2)

For more information topics, refer to the following links:

 Health Service in Windows Server
 Performance history for Storage Spaces Direct
 PrivateCloud.DiagnosticInfo
 System Insights overview
 Troubleshoot Storage Spaces Direct
 Cluster to cluster Storage Replication
Thank you

© Copyright Microsoft Corporation. All rights reserved.