Dell PowerStore: Best Practices Guide

July 2024

H18241.8

White Paper

Abstract
This document provides best practices for installing and configuring Dell PowerStore
for optimal performance and availability.

Internal Use - Confidential

 Copyright

The information in this publication is provided as is. Dell Inc. makes no representations or warranties of any kind with respect
to the information in this publication, and specifically disclaims implied warranties of merchantability or fitness for a particular
purpose.
Use, copying, and distribution of any software described in this publication requires an applicable software license.
Copyright © 2020-2024 Dell Inc. or its subsidiaries. Published in the USA July 2024 H18241.8.
Dell Inc. believes the information in this document is accurate as of its publication date. The information is subject to change
without notice.

2 Dell PowerStore: Best Practices Guide

 Contents

Contents
Executive summary ........................................................................................................................ 4

Introduction ..................................................................................................................................... 4

Hardware considerations ............................................................................................................... 7

Network considerations.................................................................................................................. 9

PowerStore storage resources .................................................................................................... 12

PowerStore features and layered applications .......................................................................... 14

External host considerations ....................................................................................................... 16

Conclusion ..................................................................................................................................... 17

References ..................................................................................................................................... 18

Internal Use - Confidential

 Executive summary

Executive summary

Overview This white paper provides best practices guidance for using Dell PowerStore in a mixed-
business environment. It focuses on optimizing system performance and availability, and
on maximizing the usability of the automated storage features.

These guidelines are intended to cover most use cases. They are recommended by Dell
Technologies but are not strictly required. Some exception cases are addressed in this
white paper. Less common edge cases are not covered by these general guidelines and
are addressed in use case-specific white papers.

For questions about the applicability of these guidelines in a specific environment, contact
your Dell Technologies representative to discuss the recommendations.

Audience This document is intended for IT administrators, storage architects, partners, and Dell
Technologies employees. This audience also includes any individuals who may evaluate,
acquire, manage, operate, or design a Dell networked storage environment using
PowerStore systems.

Revisions Part number/

Date Description
revision

April 2020 Initial release: PowerStoreOS 1.0

June 2020 Terminology updates

September Terminology updates

2020

June 2021 PowerStoreOS 2.0 updates

January 2022 PowerStoreOS 2.1 updates; template update

July 2022 H18241.5 PowerStoreOS 3.0 updates

May 2023 H18241.6 PowerStoreOS 3.5 updates

May 2024 H18241.7 PowerStoreOS 4.0 updates

Removed references to PowerStore X

July 2024 H18241.8 Minor updates

Introduction

Document This document introduces specific configuration recommendations that enable optimal
purpose performance from PowerStore.

PowerStore PowerStore achieves new levels of operational simplicity and agility. It uses a container-
overview based microservices architecture, advanced storage technologies, and integrated
machine learning to unlock the power of your data. PowerStore is a versatile platform with

Internal Use - Confidential

 Introduction

a performance-centric design that delivers multidimensional scale, always-on data

reduction, and support for next-generation media.

PowerStore brings the simplicity of public cloud to on-premises infrastructure, streamlining

operations with an integrated machine-learning engine and seamless automation. It also
offers predictive analytics to easily monitor, analyze, and troubleshoot the environment.
PowerStore is highly adaptable, providing the flexibility to host specialized workloads
directly on the appliance and to modernize infrastructure without disruption. It also offers
investment protection through flexible payment solutions and data-in-place upgrades.

Terminology The following table provides definitions for some of the terms that are used in this
document.

Table 1. Terminology

Term Definition

Appliance A solution containing a base enclosure and any attached

expansion shelves. The size of an appliance could be the base
enclosure only or the base enclosure plus expansion shelves.

Base enclosure Refers to the enclosure that contains both nodes (node A and
node B) and 25 NVMe drive slots.

Cluster One or more appliances in a single grouping and management

interface. Clusters are expandable by adding more appliances
to the existing cluster, up to the allowed amount for a cluster.

Embedded module Connectivity card in the PowerStore node that provides ports
for Ethernet connections, and various service and
management ports.

Expansion enclosure Enclosures that can be attached to a base enclosure to

provide additional storage in the form of SAS drives.

Fibre Channel A protocol used to perform NVMe or SCSI commands over a

Fibre Channel (FC) network.

File system A storage resource that can be accessed through file-sharing

protocols such as SMB or NFS.

Internet SCSI (iSCSI) Provides a mechanism for accessing block-level data storage
over network connections.

I/O module Optional connectivity cards that provide additional Fibre

Channel or Ethernet ports.

IOPS I/Os per second, a measure of transactional performance for

small-block workloads.

MBPS Megabytes per second, a measure of bandwidth performance

for large-block workloads.

Network-attached storage A virtualized network-attached storage server that uses the

(NAS) server SMB, NFS, or FTP/SFTP protocols to catalog, organize, and
transfer files within file system shares and exports. A NAS
server, the basis for multi-tenancy, must be created before
creating file-level storage resources. A NAS server is
responsible for the configuration parameters on the set of file
systems that it serves.

Internal Use - Confidential

 Introduction

Term Definition

Network File System (NFS) An access protocol that allows data access from Linux/UNIX
hosts on a network.

Node A storage node that provides the processing resources for

performing storage operations and servicing I/O between
storage and hosts.

NVMe over Fibre Channel Allows hosts to access storage systems across a Fibre
(NVMe/FC) Channel network fabric using the NVMe protocol.

NVMe over TCP (NVMe/TCP) Allows hosts to access storage systems across a TCP network
fabric using the NVMe protocol.

PowerStore Command Line An interface that allows a user to perform tasks on the storage
Interface (PSTCLI) system by typing commands instead of using the user interface
(UI).

PowerStore Q model Container-based storage system that is running on purpose-

built hardware. This storage system supports unified (block
and file) workloads or block-optimized workloads. The
PowerStore Q model supports Quad-Level Cell (QLC) NVMe
SSDs for data storage.

PowerStore T model Container-based storage system that is running on purpose-

built hardware. This storage system supports unified (block
and file) workloads or block-optimized workloads. The
PowerStore T model supports Triple-Level Cell (TLC) NVMe
SSDs for data storage.

Server Message Block (SMB) An access protocol that allows remote file data access from
clients to hosts on a network. This protocol is typically used in
Microsoft Windows environments.

Snapshot A point-in-time view of data stored on a storage resource. A

user can recover files from a snapshot, restore a storage
resource from a snapshot, or provide access to a host.

Thin clone A read/write copy of a volume, volume group, file system, NAS
server, or snapshot that shares blocks with the parent
resource.

Virtual Volumes (vVols) A VMware storage framework that allows VM data to be stored
on individual VMware vSphere Virtual Volumes (vVols). This
framework allows for data services to be applied at a VM-
granularity level while using Storage Policy Based
Management (SPBM).

Volume A block-level storage device that can be shared out using a

protocol such as iSCSI or Fibre Channel.

We value your Dell Technologies and the authors of this document welcome your feedback on this
feedback document. Contact the Dell Technologies team by email.

Author: Ethan Stokes

Contributors: Stephen Wright, Andrew Sirpis, Ryan Poulin, Wei Chen

Note: For links to other documentation for this topic, see the PowerStore Info Hub.

Internal Use - Confidential

 Hardware considerations

Hardware considerations

Introduction At the highest level, design for optimal performance follows a few simple rules. The main
principles of designing a PowerStore system for performance are:
• Distribute workloads across available resources
• Simplify the configuration
• Design for resilience
• Maintain the latest-released PowerStoreOS version

Hardware components are the foundation of any storage system. This section discusses
some key hardware differences between PowerStore models that help determine
performance. It also explains how different configuration options can result in different
performance from the same hardware.

PowerStore The PowerStore platform consists of eleven different models, from the PowerStore 500
deployment model through the PowerStore 9200 model. PowerStoreOS 1.0 launched with the
modes PowerStore 1000T, 3000T, 5000T, 7000T, and 9000T. PowerStoreOS 2.0 introduced the
entry-level 500. PowerStoreOS 3.0 introduced a platform refresh, with the 1200T, 3200T,
5200T, and 9200T models. PowerStoreOS 4.0 introduced the 3200Q model appliance. All
models use a common base enclosure and I/O modules. The models differ by CPU core
count and speed, memory size, and number of NVMe NVRAM drives. These hardware
differences give each model a unique performance profile.

Besides the hardware differences between the models, PowerStore can be installed in
one of two different deployment modes. Each deployment mode has different capabilities,
as detailed in Table 2. Choose the deployment mode that provides the required
capabilities.

Table 2. PowerStore configurations

External block
Deployment mode External file access
access

Unified ✓ ✓
Block optimized ✓ X

The PowerStore system has different performance characteristics depending on the

deployment mode.

PowerStore Deployment Modes

PowerStore appliances can be installed in a unified configuration that provides file and
block access, or in a block optimized configuration that provides only block access.

Deployment Mode: Unified

PowerStore appliances with unified modes can provide access to block and file storage
resources simultaneously. This is the default deployment mode.

Internal Use - Confidential

 Hardware considerations

Deployment Mode: Block optimized

If the PowerStore appliance will not be used for file access, it can be installed in block
optimized mode, which disables the file capabilities. This mode can increase the amount
of block workload that the system can provide because it can devote the additional CPU
and memory that is no longer needed for file capabilities.

Relative performance expectations

This section compares the performance potential of the different PowerStore models
when serving external workloads. Performance scales based on the specific hardware
complement of the model and is also impacted by the configuration type.

In general, the IOPS capability of the PowerStore models scales linearly from PowerStore
500 up to 9200 models. As mentioned previously, deployment mode also impacts
performance capability. A PowerStore T model in block optimized mode can deliver more
block IOPS than the same model in unified mode.

Except for PowerStore 500, PowerStore systems use NVMe NVRAM drives to provide
persistent storage for cached write data. PowerStore 1000 up to 3200 model arrays have
two NVRAM drives per system, while PowerStore 5000 up to 9200 model arrays have
four NVRAM drives per system. The extra drives mean that these systems can provide
higher MBPS for large-block write workloads.

PowerStore cluster
PowerStore systems can be clustered. A PowerStore cluster combines multiple
PowerStore appliances into a single grouping that is managed as a single storage system.
A PowerStore cluster delivers aggregate performance from all appliances in the cluster,
but a single volume is serviced by only one appliance at any given time. While not
required, it is recommended that all appliances in a cluster be of the same model and
have similar physical capacities to provide consistent performance across the cluster.
PowerStore T models and PowerStore Q models can be mixed within the same cluster.

Volumes can be migrated between appliances in a cluster. It is recommended that any

host that is connected to a PowerStore cluster has equivalent connectivity to all
appliances in the cluster. All appliances in a cluster should be physically located in the
same data center and must be connected to the same LAN.

Clustering is applicable to block storage resources only. While a PowerStore appliance in

Unified mode can serve as the cluster’s primary appliance, the file resources cannot
migrate to a different appliance. When deploying multiple appliances for file access, plan
to have multiple clusters.

Drive PowerStore can be configured with NVMe solid-state devices (SSDs) or NVMe storage
configuration class memory (SCM) drives for user data. The 3200Q model only supports QLC NVMe
SSDs, while all other PowerStore model appliances support TLC NVMe SSDs. Mixing
QLC and TLC drives within a single appliance is not supported. SSD-based systems can
be expanded with additional drives to increase the amount of available storage capacity.
PowerStore 1000, 3000, 5000, 7000, and 9000 models can be expanded with SAS SSD-
based expansion shelves. Starting in PowerStoreOS 3.0, all PowerStore models,
including the 3200Q, can be expanded with NVMe SSD-based expansion shelves if they
meet the necessary hardware prerequisites defined in the white paper Dell PowerStore:

Internal Use - Confidential

 Network considerations

Introduction to the Platform. It is recommended that all drives within a PowerStore system
be the same size, which can maximize the usable capacity of each drive.

PowerStore Dynamic Resiliency Engine (DRE) is used to manage the drives in the
system. All drives are automatically used to provide storage capacity. DRE groups the
drives into resiliency sets to protect against drive failure. User configuration of the drives
is not necessary, and dedicated hot spare drives are not required in PowerStore. Spare
space for rebuilds is automatically distributed across all drives within each resiliency set.
This configuration provides better resource utilization and enables faster rebuilds if there
is a drive failure.

At initial installation of the PowerStore system, DRE can be configured with either single-
or double-drive failure tolerance. To provide the greatest usable capacity from the same
number of drives, it is recommended to initially install PowerStore with a minimum of ten
drives for single-drive failure tolerance, or nineteen drives for double-drive failure
tolerance. PowerStore 3200Q models require a minimum of eleven drives, and double-
drive failure tolerance is enforced on the appliance.

SCM drives
PowerStore can use SCM drives either by having only SCM drives installed in the system,
or, with PowerStoreOS 2.0 or higher, by mixing SCM and TLC model SSD drives.

Systems with all SCM drives are recommended for small-block workloads that require the
absolute lowest latencies. A system with all SCM drives will place both data and metadata
on the SCM drives.

Systems with mixed SSD and SCM drives will use the SCM drives for metadata
acceleration; the SCM drives will store metadata for faster lookups. This can reduce
latency on read operations in systems with large physical capacities. When mixing SCM
and SSDs, it is recommended that at least 5 percent of the system’s physical capacity
should be SCM.

Network considerations

Introduction External hosts send and receive data from PowerStore through Fibre Channel, Ethernet,
or both networks. These networks play a large role in determining the performance
potential of PowerStore. This section discusses considerations for the external network,
and for the PowerStore network ports.

Host networking recommendations are covered in External host considerations.

General network It is recommended to use redundant switch hardware between the PowerStore system
performance and and external clients. For more details about configuring a redundant network, see the
high availability document PowerStore Host Configuration Guide on Dell.com/powerstoredocs.

Fibre Channel fabrics

For Fibre Channel connectivity, configure dual redundant fabrics, with each PowerStore
node and each external host having connectivity on each of the fabrics. Minimize the
number of hops between host and PowerStore.

Internal Use - Confidential

 Network considerations

For performance, load balancing, and redundancy, each host should have at least two
paths to each PowerStore node (four paths per PowerStore appliance). It is
recommended that a host should have no more than eight paths per volume.

Ethernet networks
For Ethernet connectivity, use multiple switches that connect to Virtual Link Trunking
interconnect (VLTi) and Link Aggregation Control Protocol (LACP) or equivalent
technologies. Each PowerStore node should have connectivity to all linked switches.

The first two ports of the embedded module 4-port card on each PowerStore node are
bonded together within the PowerStoreOS. For the highest performance and availability
from these ports, it is recommended also to configure link aggregation across the
corresponding switch ports.

With the introduction of PowerStoreOS 3.0, user-defined link aggregations are supported
for Ethernet ports. Link aggregations can include 2 to 4 different Ethernet ports. The ports
must be on the same node and operate at the same speed. A mirror link aggregation will
automatically be created on the peer node. In PowerStoreOS releases prior to 4.0, user-
defined link aggregations only support NAS server interfaces. In PowerStoreOS 4.0 and
later, user-defined link aggregations also support iSCSI host connectivity and replication.

When using Ethernet for block access to PowerStore volumes (by means of iSCSI and/or
NVMe over TCP), each host should have at least two paths to each PowerStore node
(four paths per PowerStore appliance). It is recommended that a host should have no
more than eight paths per volume.

PowerStore PowerStore supports Ethernet connectivity through ports on the embedded module, and
front-end ports on optional I/O modules. PowerStore supports Fibre Channel connectivity through ports
on optional I/O modules.

The fastest I/O module should be installed in slot 0. On PowerStore 1000 – PowerStore
9200 models, I/O module slot 0 is 16-lane PCIe Gen3 while I/O module slot 1 is 8-lane. If
the system is configured with a 100 GbE I/O module, it should be installed in slot 0 to
enable the highest bandwidth. If a second 100 GbE I/O module is required, it can be
installed in slot 1. Both I/O module slots on PowerStore 500 models are 8-lane PCIe, and
the 100 GbE I/O module is not supported on this platform.

When a 32Gb Fibre Channel I/O module is being installed, it is recommended to always
use I/O module slot 0 first unless the system currently contains or in the future will contain
a 100 GbE I/O Module. In this case, the Fibre Channel I/O module can be installed in slot
1.

PowerStore Fibre Channel ports

PowerStore Fibre Channel ports support speeds for 32 Gb/s, 16 Gb/s, 8 Gb/s, and 4
Gb/s. This speed depends on the SFP used and the switchport or HBA that is connected.
Because higher speeds allow for greater MBPS and IOPS capabilities, it is recommended
that you use the highest speed supported by the environment.

Fibre Channel ports are available on I/O modules that are inserted into I/O module slots
on the nodes. The Fibre Channel I/O module is 16-lane PCIe Gen3. On PowerStore 1000
– 9200 models, I/O module slot 0 is also 16-lane while I/O module slot 1 is 8-lane. If Fibre

Internal Use - Confidential

 Network considerations

Channel I/O modules are installed in both I/O module slots, it is recommended to cable
the ports in I/O module slot 0 first, due to the PCIe difference. The PCIe lanes in I/O
module slot 1 are only a limiting factor for total MBPS, and only when all four ports on the
Fibre Channel I/O module are operating at 32 Gb/s. Both I/O module slots on PowerStore
500 are 8-lane PCIe and therefore, there is no slot preference.

NVMe over Fibre Channel

The NVMe over Fibre Channel (NVMe/FC) protocol provides connectivity using the same
Fibre Channel ports but can decrease the transport latency between PowerStore and the
host. Note that all parts of the network, including switches and HBAs, must support NVMe
over Fibre Channel.

PowerStore Ethernet ports

PowerStore optical Ethernet ports support speeds of up to 25 Gb/s, based on the SFP
that is used. Copper Ethernet ports support speeds of up to 10 Gb/s. Because higher
speeds allow for greater MBPS and IOPS capabilities, it is recommended that you use the
highest speed supported by your environment.

With PowerStoreOS 3.0, a new 2-Port Ethernet card is introduced that supports speeds of
up to 100 Gb/s. This 100 GbE card is supported on PowerStore 1000-9200 models in I/O
Module 0 slot.

Jumbo frames (MTU 9000) are recommended for increased network efficiency. Jumbo
frames must be supported on all parts of the network between PowerStore and the host.

Ethernet ports for iSCSI

Map additional Ethernet ports for iSCSI to increase system MBPS capabilities. Enable
Jumbo frames for iSCSI by setting the Cluster MTU to 9000 and set the storage network
MTU to 9000.

The embedded module 4-port card and the optional network I/O modules are 8-lane PCIe
Gen3. When more than two 25 GbE ports are used, these cards are oversubscribed for
MBPS. To maximize MBPS scaling in the system, consider cabling and mapping the first
two ports of all cards in the system first. Then, cable and map other ports as needed.

When PowerStore models that are in unified mode are used for both iSCSI and file
access, it is recommended that you use different physical ports for both NAS and iSCSI.

Ethernet ports for NVMe over TCP

The NVMe over TCP (NVMe/TCP) protocol provides connectivity using the same physical
Ethernet ports as iSCSI. NVMe/TCP can be enabled on the same Storage Network as
iSCSI or different Storage Networks can be created to isolate iSCSI and NVMe/TCP
traffic.

The embedded module 4-port card and the optional network I/O modules are 8-lane PCIe
Gen3. When more than two 25 GbE ports are used, these cards are oversubscribed for
MBPS. To maximize MBPS scaling in the system, consider cabling and mapping the first
two ports of all cards in the system first. Then, cable and map other ports as needed.

When PowerStore models that are in unified mode are used for both NVMe/TCP and file
access, it is recommended that you use different dedicated physical ports for both NAS
and NVMe/TCP.

Internal Use - Confidential

 PowerStore storage resources

SmartFabric Storage Software (SFSS)

Dell SmartFabric Storage Software (SFSS) provides Centralized Discovery Controllers
(CDCs) for NVMe/TCP Endpoints. These CDCs facilitate endpoint discovery, registration,
soft zoning, and event notifications. With SFSS, Dell Technologies provides the industry's
first comprehensive connectivity automation solution for NVMe/TCP endpoints such as
Dell PowerEdge, Dell PowerStore, and Dell PowerMax. For more information about SFSS
support, see the SmartFabric Storage Software Deployment Guide on the Storage
Networking Info Hub.

Ethernet ports for NAS

It is recommended that you use bonded ports for NAS connectivity. Prior to
PowerStoreOS 3.0, NAS servers automatically created their interfaces on the two bonded
ports on the embedded module 4-port card. With PowerStoreOS 3.0, user-defined link
aggregations can be used to reserve different physical ports for file access only. In
PowerStoreOS 4.0, user-defined link aggregations also support storage iSCSI and
replication connectivity. For the highest performance and availability from any aggregated
ports, it is recommended that you configure link aggregation across the corresponding
switch ports.

Enable Jumbo frames for NAS by setting the cluster MTU to 9000.

If the PowerStore is also providing block access through iSCSI or NVMe-TCP, or

asynchronous replication over Ethernet, it is recommended that you use different physical
ports for NAS than the ports which are tagged for replication or storage networks.

Fail-Safe Networking (FSN)

PowerStoreOS 3.5 adds Fail-Safe Networking (FSN) support for file interfaces. FSN is a
high-availability feature that enables configuring ports in a primary/backup configuration.
Under normal circumstances, the primary ports are designated as active and are used to
service IO. If all primary ports of an FSN go offline, the backup ports automatically
become active and continue to service IO. This enables redundancy in case of port, cable,
or switch failure. When the primary ports are restored, the system automatically makes
the primary ports active again. For optimal performance in the event of a failure, it is
suggested that the configurations of the active and standby ports or bonds selected be
consistent. For more information about FSN, see the Dell PowerStore: File Capabilities
white paper on the PowerStore Info Hub.

PowerStore storage resources

Introduction PowerStore allows access to block, file, or both storage types. This section provides
recommendations for the different types of storage resources.

Block storage Block storage resources are accessed through Fibre Channel, NVMe over Fibre Channel,
resources iSCSI, or NVMe over TCP protocols. A host should only access a block resource using
one of these protocols at a time. It is not supported for the same host to access the same
block resource using more than one protocol.

Internal Use - Confidential

 PowerStore storage resources

Appliance balance for block workloads

PowerStore block storage resources are accessed using ALUA/ANA active/optimized or
active/non-optimized paths between the host and the two nodes within the PowerStore
appliance. I/O is normally sent on an active/optimized path. PowerStore automatically
chooses one of the nodes for the active/optimized path, when the volume is mapped to
the host, to maintain a balanced workload across the nodes. This characteristic is called
node affinity and can be viewed in PowerStore Manager and modified with PSTCLI or
REST. These changes take effect immediately and are nondisruptive if the host is
correctly configured for multipathing.

Dynamic Node Affinity

Introduced in PowerStoreOS 2.1, the node affinity of block storage resources will be
dynamically rebalanced between nodes to maintain relatively consistent utilization,
latency, and performance between both nodes of an appliance.

This feature works on block resources that have not had the node affinity manually set by
means of PSTCLI or REST. If the node affinity was manually set, the volume must be
unmapped and then remapped to the host, which will reset the affinity back to “system
selected”. This impacts multipathing only; the system does not need to trespass any
volume between nodes.

In PowerStoreOS 4.0, dynamic node affinity is expanded to include virtual volumes

(vVols). Data and config vVols are eligible for rebalancing, while unbound vVols are
ignored. When an imbalance is detected, the node affinity of volumes is changed first. If
additional balancing is required after volumes are balanced, vVols are changed next.

This feature is designed to be transparent and seamless to clients. The goal is to achieve
an even throughput, bandwidth, and CPU utilization balance across both nodes within an
appliance, regardless of resource type. The system makes changes automatically without
requiring any admin interaction. This provides balanced hardware utilization, optimizes
performance, and adapts dynamically as workloads change.

Performance policy
All block storage resources in a PowerStore system have a defined performance policy.
By default, this policy is set to Medium. The performance policy does not have any impact
on system behavior unless some volumes have been set to Low Performance Policy, and
other volumes are set to Medium or High. During times of system resource contention,
PowerStore devotes fewer compute resources to volumes with Low Performance Policy.
Reserve the Low policy for volumes that have less critical performance needs.

Quality of Service (QoS)

In PowerStoreOS version 4.0, Quality of Service (QoS) can be configured for volumes
and volume groups. A QoS policy, which includes an I/O limit rule, sets the maximum
limits on I/O for volumes and volume groups. These policies are used to ensure that
critical applications get priority over other workloads and provide predictable performance
for each application. QoS policies are interoperable with existing performance policies.

As this feature is an I/O limiting feature, host performance can be impacted. The impact
on host I/O depends directly on the amount of I/O to the resource and the limit configured.
Having a limit too low can cause performance-related issues, which can be seen as

Internal Use - Confidential

 PowerStore features and layered applications

longer than usual response times and queued I/Os on the host. If the performance impact
is too severe, the QoS limit may need to be adjusted.

Maximum limits are enforced only from I/O that arrives from an external host. These limits
are not enforced on internal synchronous or asynchronous replication or migration I/O.
QoS policies are also not replicated to a remote system. When replication is configured, it
is suggested to configure the same QoS limits on each end of the replication session.
Having different limits in a metro volume configuration may change the pattern of I/O, as a
host may prefer paths that have a higher QoS limit.

Dell Technologies recommends setting QoS on workloads that might monopolize system
resources and starve other applications of their required performance. Consider some of
the following opportunities to utilize PowerStore QoS:

• Limit the bandwidth available to large-block applications, such as backup jobs,

which may be increasing the latency on other small-block workloads.
• Limit the IOPS capability of Thin Clones which are used in Test/Dev environments,
so that they do not impact the capability of the associated source objects.
• Limit the IOPS / bandwidth that is available to non-essential applications that are
hosted on the same system as your critical applications.

File storage File storage resources are accessed through NAS protocols, such as NFS and SMB. A
resources NAS server can provide access to a file system using all NAS protocols simultaneously if
configured for multiprotocol access.

Appliance balance for file workloads

A single NAS server uses compute resources from only one node of the appliance. It is
recommended to create at least two NAS servers (one on each node) so that resources
from both nodes contribute to file performance.

NAS servers can be manually moved from one node to the other. This action can be done
to balance the workload if one node is busier than the other. All file systems that are
served by a given NAS server move with the NAS server to the other node.

PowerStore features and layered applications

Introduction This section discusses the features and layered applications available with
PowerStoreOS.

Data reduction PowerStore provides data-reduction capabilities such as zero-detect, compression, and
deduplication. Data reduction is integrated into the PowerStore architecture and is always
active. During periods of high write activity, PowerStore may defer the deduplication of
data, and devote those resources to servicing the client workload. During periods of low
activity, PowerStore will use excess resources to re-examine any data written during
these periods for duplicates, to regain any space savings that were not initially realized.

Replication PowerStore provides native synchronous and asynchronous replication solutions to

protect data and help organizations meet business goals for both data availability and

Internal Use - Confidential

 PowerStore features and layered applications

protection. Supported storage resources for native synchronous and asynchronous

replication are volumes, volume groups, thin clones, NAS servers, and file systems. The
replication itself uses iSCSI or the optimized Dell proprietary TCP-based replication
protocol through Ethernet (LAN) connections that was introduced in PowerStoreOS 3.0.

When configuring replication, it is recommended that you use different physical ports for
replication and front-end Ethernet storage traffic. A specific storage network for replication
can be created and mapped to the desired physical ports for replication.

Metro and synchronous replication also require a low latency network (<5ms). As all host
I/O must be written to the local and remote system before acknowledging the host,
network speeds and the performance of the peer system can impact the latency
encountered by the host. It suggested that the local and remote system models be equal
in case a failover of I/O must occur.

Metro Volume In PowerStoreOS 3.0 and higher, PowerStore appliances support native metro volume
replication. This provides synchronous replication of spanned block storage volumes
exclusively for VMware VMFS Datastores in an active/active configuration across two
PowerStore clusters in metro distance. In PowerStoreOS 4.0, volume group support has
been added, along with Windows and Linux support. For more information about Metro
Volume support, see the white paper Dell PowerStore: Replication Technologies. For best
practices for Metro Volume (latency, distance, uniform/non-uniform configurations, and so
on), see the white paper Dell PowerStore: Metro Volume. Both are available on the
PowerStore Info Hub.

Snapshots and All storage resources in PowerStore are thinly provisioned and space-efficient, including
thin clones snapshots and thin clones. Creation of a snapshot or thin clone requires only a quick
duplication of pointers. After this action, they behave as independent storage resources
and do not impact the performance of the source resource.

Secure In PowerStoreOS 3.5 and later, the optional secure snapshot setting allows you to create
snapshots snapshots for volumes and volume groups. When the secure snapshot setting is enabled,
the snapshot is protected from deletion until the retention period expires. This provides a
cost-effective line of defense against ransom attacks and accidental deletion of
snapshots, volumes, or volume groups. Note that after it is set, the retention period cannot
be reduced by the user. It is therefore important to keep this in mind when defining it.
Secure snapshots can only be terminated after they expire or by customer-authorized Dell
Support.

Storage Direct In PowerStoreOS 3.5 and later, Storage Direct is available. With Storage Direct, users
can back up volumes and volume groups directly to a PowerProtect DD series appliance
running on-premises or in the cloud. This feature eliminates the requirement of a backup
host because all backup traffic is offloaded to the storage appliances. PowerStore uses
the replication tagged storage ports to connect and transfer data to the PowerProtect DD
appliance leveraging the DD Boost protocol. As with replication, it is recommended that
you use different physical ports for data protection and front-end Ethernet storage traffic.

In PowerStoreOS 4.0, an enhancement to storage networks allows users to create

multiple storage networks with defined purposes. These purposes include Storage

Internal Use - Confidential

 External host considerations

(iSCSI), Storage (NVMe/TCP), and Replication. This feature allows storage networks for
host connectivity and dedicated networks for replication. When using Storage Direct, it is
required that the Replication purpose only be added to a single storage network within the
cluster. The Storage (iSCSI) purpose must be assigned to the network with the
Replication purpose and be configured on a port/network that has connectivity to the
remote PowerProtect appliance.

Cluster Block storage resources that are migrating between appliances in a cluster may see
migrations impacted performance from the migration activity. It is recommended to run migrations at
a time when the resource is less busy.

PowerStoreOS New versions of the PowerStoreOS are applied using a nondisruptive upgrade process.
upgrades Because half of the system hardware resources are unavailable during parts of the
upgrade, it is recommended to perform upgrades during planned maintenance windows.
Alternately, perform upgrades when the system is less busy to minimize the impact to
clients. Users should execute a Pre-Upgrade Health Check before their scheduled
maintenance window to ensure that any issues can be resolved.

External host considerations

Introduction This section highlights host configuration changes that may be necessary to access
PowerStore volumes efficiently. For details about the appropriate settings for a host type,
see the Host Connectivity Guides on E-Lab Navigator.

Host Because PowerStore is a new class of storage, host operating systems may not natively
configuration recognize PowerStore volumes and apply the appropriate settings. For optimal
performance, check that the appropriate configuration changes have been applied to all
hosts that are connected to a PowerStore. The Host Connectivity Guides on E-Lab
Navigator have recommendations for the following:
• MPIO settings: Path checker and timeout values
• iSCSI settings: Time-out and queue depth values; disabling delayed ACK
• Fibre Channel settings: Queue depth values
• Network settings: Jumbo frames and flow control
• Unmap operations
• VMware ESXi claim rules

For other recommended configurations for VMware ESXi and vSphere, see the document
Dell PowerStore: Virtualization Integration.

Host file systems When a host is attached to a PowerStore block volume, the host can use this volume as a
raw device, or it can create a local file system on the volume first. When a local file
system is being created, it is recommended to disable SCSI unmap. When PowerStore
creates a volume, all space is already unmapped; the host-based unmap is redundant
and generates unnecessary load on PowerStore.

Internal Use - Confidential

 Conclusion

When creating a local file system, it is recommended to use a file system block size
(allocation unit) that is an even multiple of 4 KB.

It is typically not necessary to perform alignment when creating a local file system. If
alignment is performed, it is recommended to use an offset of 1 MB.

VMware PowerStore is tightly integrated with VMware applications.

For other recommended configurations for VMware ESXi and vSphere, see the document
Dell PowerStore: Virtualization Integration.

Application PowerStore is well integrated with the most widely used enterprise applications. For best
considerations practice recommendations for specific applications, see the solutions-focused white
papers available on the PowerStore Info Hub.

Conclusion

Summary This white paper provides configuration and usage recommendations for PowerStore in
general use cases. For a detailed discussion of the reasoning or methodology behind
these recommendations, or for additional guidance around more specific use cases,
contact your Dell Technologies representative.

Internal Use - Confidential

 References

References
Dell The Dell Technologies Info Hub > Storage site provides expertise that helps to ensure
Technologies customer success with Dell Technologies storage platforms.
documentation
Dell.com/powerstoredocs provides detailed documentation about how to install, configure,
and manage PowerStore systems.

Internal Use - Confidential