SmartStack Reference Archi tec tures
Nimble Storage SmartStackTM
Desktop and Server Virtualization with Cisco and VMware
Combining both Desktop Virtualization and Virtualized Business Critical Applications
on a single converged infrastructure.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2
Table of Contents
REFERENCE ARCHITECTURES .................................................................................................................................................... 5
SUMMARY OF MAIN FINDINGS ................................................................................................................................................ 6
BUSINESS VALUE ...................................................................................................................................................................... 6
SOLUTION OVERVIEW .............................................................................................................................................................. 7
UNIFIED, PRETESTED, AND VALIDATED INFRASTRUCTURE ........................................................................................ 7
BENEFITS ................................................................................................................................................................................ 11
ARCHITECTURE AND DESIGN .................................................................................................................................................. 12
VMWARE VSPHERE, CISCO UCS, NIMBLE STORAGE ............................................................................................ 12 VIRTUALIZING BUSINESS CRITICAL APPLICATIONS ................................................................................................. 13
Availability ..................................................................................................................................................... 13 Performance Optimization ............................................................................................................................ 17 Data Protection ............................................................................................................................................. 22 Operational Management ............................................................................................................................. 24
MODULAR VIRTUAL DESKTOP INFRASTRUCTURE ................................................................................................... 32 Cisco UCS B-Series Blade Servers .............................................................................................................. 33 VMware vSphere .......................................................................................................................................... 33 VMware Horizon View .................................................................................................................................. 34 VMware View Planner .................................................................................................................................. 34 Nimble Storage CS-Series ........................................................................................................................... 34
SOLUTION VALIDATION .......................................................................................................................................................... 35
Server: Cisco UCS ........................................................................................................................................ 35 Storage: Nimble Storage CS-Series ............................................................................................................. 38 VDI: VMware ................................................................................................................................................. 39
TESTING METHODOLOGIES .................................................................................................................................................... 40
CONCLUSION .......................................................................................................................................................................... 43
APPENDIX A: BILL OF MATERIALS ........................................................................................................................................... 44
APPENDIX B: VALIDATION FOR 500-USER BUSINESS CRITICAL APPLICATIONS ENVIRONMENT ............................................ 45
RESOURCE UTILIZATION ...................................................................................................................................... 49
APPENDIX C: STORAGE PERFORMANCE ................................................................................................................................. 54
BOOT STORMS .................................................................................................................................................... 54 STEADY STATE .................................................................................................................................................... 54
APPENDIX D: HORIZON VIEW POOL DETAILS ......................................................................................................................... 56
APPENDIX E: VDI PLANNING AND SIZING QUESTIONS ........................................................................................................... 58
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3
List of Figures
FIGURE 1 - CONVERGED INFRASTRUCTURE REFERENCE ARCHITECTURE ................................................................. 5 FIGURE 2 - HIGH LEVEL PHYSICAL TOPOLOGY ...................................................................................................... 12 FIGURE 3 - HIGH LEVEL LOGICAL TOPOLOGY ........................................................................................................ 13 FIGURE 4 - UCS AVAILABILITY – KEY POINTS ....................................................................................................... 14 FIGURE 5 - REDUNDANT NETWORKS .................................................................................................................... 14 FIGURE 6 - AUTOMATED NETWORK FAILOVER ....................................................................................................... 15 FIGURE 7 – STORAGE MPIO – ROUND ROBIN PATH SELECTION ............................................................................ 15 FIGURE 8 – ESXI STORAGE DEVICE SETTINGS ..................................................................................................... 16 FIGURE 9 - HYPERVISOR HA ................................................................................................................................ 16 FIGURE 10 - VIRTUAL NETWORK SEPARATION ...................................................................................................... 17 FIGURE 11 - OPTIMIZATION LOCATIONS ................................................................................................................ 17 FIGURE 12 - ISCSI STORAGE SUBNETS................................................................................................................ 18 FIGURE 13 – INFRASTRUCTURE SPECIFIC VOLUME ASSIGNEMNT ........................................................................... 18 FIGURE 14 - APPLICATION SPECIFIC VOLUME ASSIGNMENT ................................................................................... 19 FIGURE 15 - APPLICATION AWARE PERFORMANCE POLICIES ................................................................................. 19 FIGURE 16 - VMKERNEL NETWORK ALLOCATION - VSWITCH ................................................................................... 20 FIGURE 17 - VMKERNEL NETWORK ALLOCATION - DETAILS .................................................................................... 20 FIGURE 18 - VM PROPERTIES - EXCHANGE .......................................................................................................... 21 FIGURE 19 - VM PROPERTIES - SQL SERVER ....................................................................................................... 21 FIGURE 20 - VM PROPERTIES - SHAREPOINT ....................................................................................................... 22 FIGURE 21 - BACKUP UCSM ............................................................................................................................... 23 FIGURE 22 - NIMBLE VOLUME COLLECTION PROTECTION SCHEME ......................................................................... 24 FIGURE 23 - NIMBLE STORAGE SNAPSHOT SYNCHRONIZATION - EXCHANGE .......................................................... 24 FIGURE 24 - NIMBLE STORAGE SNAPSHOT SYNCHRONIZATION - SQL SERVER ....................................................... 24 FIGURE 25 - UCS VNIC ALLOCATION ................................................................................................................... 25 FIGURE 26 - ISCSI BOOT FROM NIMBLE ............................................................................................................... 26 FIGURE 27 - VCENTER STORAGE INTEGRATION ..................................................................................................... 26 FIGURE 28 - VCENTER OPERATIONS MANAGER .................................................................................................... 27 FIGURE 29 - VCOPS - APPLICATIONS .................................................................................................................... 27 FIGURE 30 - VCOPS HEALTH ............................................................................................................................... 28 FIGURE 31 - INFOSIGHT WELLNESS OVERVIEW..................................................................................................... 28 FIGURE 32 - INFOSIGHT CAPACITY USAGE............................................................................................................ 29 FIGURE 33 - INFOSIGHT PERFORMANCE VIEW ...................................................................................................... 30 FIGURE 34 - INFOSIGHT PROTECTION COVERAGE VIEW ........................................................................................ 31 FIGURE 35 - INFOSIGHT EXECUTIVE DASHBOARD ................................................................................................. 32 FIGURE 36 – UCS FABRIC INTERCONNECT – APPLIANCE PORT VLAN CONNECTIVITY ............................................ 35 FIGURE 37 - UCS APPLIANCE PORT CONFIGURATION ........................................................................................... 36 FIGURE 38 - CISCO UCS BLADE SERVER VNIC CONFIGURATION .......................................................................... 36 FIGURE 39 - VMWARE NETWORK CONFIGURATION ............................................................................................... 37 FIGURE 40 - VMWARE ISCSI PORT 1 CONFIGURATION DETAILS ............................................................................ 37 FIGURE 41 - VMWARE ISCSI PORT 2 CONFIGURATION DETAILS ............................................................................ 38 FIGURE 42 - VMWARE DESKTOP VIRTUAL MACHINE NETWORK CONFIGURATION DETAILS ...................................... 38 FIGURE 43 - NIMBLE STORAGE ISCSI INITIATOR SETUP ........................................................................................ 39 FIGURE 44 - NIMBLE STORAGE VOLUME MANAGEMENT ......................................................................................... 39 FIGURE 45 - VDI CLUSTER CONFIGURATION – VCENTER ORGANIZATION ............................................................... 40 FIGURE 46 - VMWARE VIEW PLANNER WORKLOAD AND RUN PROFILE ................................................................... 41 FIGURE 47 - NIMBLE STORAGE PERFORMANCE - DAY VIEW ................................................................................... 42 FIGURE 48 - CISCO UCS BLADE - STEADY STATE WORKLOAD - CPU UTILIZATION ................................................. 43 FIGURE 49 - BUSINESS CRITICAL APPLICATIONS ................................................................................................... 45 FIGURE 50 - EXCHANGE 2010 ARCHITECTURE ...................................................................................................... 45 FIGURE 51 - EXCHANGE 2010 DETAILED SETUP ................................................................................................... 46 FIGURE 52 - SQL AND SHAREPOINT DATABASE ARCHITECTURE ............................................................................ 46 FIGURE 53 – EXCHANGE LOADGEN RESULTS ....................................................................................................... 48 FIGURE 54 - DVDSTORE RUN RESULTS ............................................................................................................... 48 FIGURE 55 - EXAMPLE SHAREPOINT PAGE 1 ........................................................................................................ 49
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4
FIGURE 56 - EXAMPLE SHAREPOINT PAGE 2 ........................................................................................................ 49 FIGURE 57 - ESXI SERVER 1 RESOURCE UTILIZATION .......................................................................................... 50 FIGURE 58 - ESXI SERVER 2 RESOURCE UTILIZATION .......................................................................................... 50 FIGURE 59 - EXCHANGE DAG NODE 1 RESOURCE UTILIZATION ............................................................................. 51 FIGURE 60 - EXCHANGE DAG NODE 2 RESOURCE UTILIZATION ............................................................................. 51 FIGURE 61 - SQL DB SERVING DVDSTORE AND SHAREPOINT DATABASES RESOURCE UTILIZATION ....................... 52 FIGURE 62 - NIMBLE STORAGE CS220G ARRAY IOPS AND LATENCY CHART ......................................................... 52 FIGURE 63 - BOOT STORM STORAGE PERFORMANCE............................................................................................ 54 FIGURE 64 – COMBINED WORKING DAY STORAGE PERFORMANCE ........................................................................ 55 FIGURE 65 - VMWARE HORIZON VIEW POOL CONFIGURATION ............................................................................... 56 FIGURE 66 - VIEW POOL STORAGE CONFIGURATION – POOL 1 .............................................................................. 56 FIGURE 67 - VIEW POOL STORAGE CONFIGURATION - POOL 2 ............................................................................... 57 FIGURE 68 - VMWARE HORIZON VIEW ADVANCED STORAGE SETTINGS ................................................................. 57
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5
Reference Architectures
This document describes the Nimble Storage SmartStackTM reference architecture for virtual
desktop infrastructure (VDI) based on VMware Horizon View and virtualized business critical
applications (BCA) from Microsoft, including Exchange, SQL Server and SharePoint. VMware
vSphere serves as the hypervisor for both VDI as well as the business critical applications. The
hardware solution includes Cisco UCS® B-Series Blade Servers and Nimble Storage CS-Series
storage appliance.
This document will highlight design best practices for virtualizing business-critical applications on
SmartStack, and showcase what was validated jointly by Nimble Storage, VMware, and Cisco. If
you want to learn more about Nimble Storage SmartStackTM, please contact your sales
representative or visit our website for links to more resources.
The purpose of the reference architecture is to provide a tested and modular architecture built with
proven best-in-class technologies to create a complete server and desktop virtualization solution,
including the application and desktop software, hypervisor, computing, networking, and storage
elements. These reference architectures accelerate your IT datacenter transformation by
enabling faster deployments, greater efficiency, and lower risk.
Choosing the correct storage solution for your virtualization solution is a critical step, especially
when combining different workloads such as virtual desktop infrastructure (VDI) and virtualized
business applications. Through the use of partner provided tools such as VMware View Planner
and Microsoft specific load testing tools, we can provide more information about the actual running
of the solutions to help assess and plan the proper network and storage configuration for your
environments.
Using this approach, Nimble Storage has achieved a server application and desktop deployment
levels of 500 users on a single three-rack-unit (3RU) Nimble Storage flash optimized storage
system. This configuration is described in this document.
Figure 1 - Converged Infrastructure Reference Architecture
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 6
This reference architecture is intended as a high level guide to understanding what was
configured and tested and is not intended to be a comprehensive guide to deployment and
configuration for every aspect of this or other possible solutions.
Summary of Main Findings
High density: the combination of the Cisco Unified Computing System™ (Cisco UCS)
and Nimble Storage hardware with VMware vSphere and VMware Horizon View software
produces a virtual desktop delivery system with a high density of users per blade and
chassis.
Workload consolidation: The converged infrastructure stack consisting of Cisco UCS
and Nimble Storage arrays eliminate the need for separate infrastructure silos for desktop
and server virtualization, simplifying deployment and management for the completely
virtualized datacenter.
Simplicity: Cisco maintains industry leadership with the new Cisco UCS Manager
software, which makes scaling and maintenance simple and helps ensure consistency.
Performance: The Cisco 10-Gbps unified fabric is also validated on second-generation
Cisco UCS 6200 Series Fabric Interconnects, testing more challenging workloads and
maintaining unsurpassed user response times.
Scalability: Up to 500 users virtual desktops and application support can run on an entry-
level Nimble Storage CS Series storage array.
Business Value
As organizations increasingly move to deploy server and desktop virtualization projects, they
often find that their IT infrastructure is incapable of adapting to the performance requirements that
virtualized business critical applications and desktop virtualization demand for meeting the needs
of the organizations. A well-designed, thoroughly tested server and desktop virtualization solution
can strategically empower a company to respond to the post-PC era, while simplifying
management and reducing cost by deploying these workloads on a single converged
infrastructure stack.
Companies require a scalable and highly available infrastructure on which to deploy their virtual
server and desktop environments. Several new technologies are available to assist them in
designing a virtual desktop solution, but they need to know how to use these technologies to get
the most from their investments, support service-level agreements (SLAs), and reduce their total
cost of ownership (TCO).
This solution builds an example of a common customer server and desktop virtualization
environment and validates the environment for performance, scalability, and capability.
Customers achieve:
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 7
Increased control and security of their global, mobile desktop environment, which is
typically their most at-risk environment
Better end-user productivity with a more consistent environment
Simplified management, with the environment contained in the data center
Better support for SLAs and compliance initiatives
Lower operation and maintenance costs
Solution Overview
This solution uses Cisco UCS, Nimble Storage, and VMware vSphere to provide resources for a
virtual desktop infrastructure and virtualized business critical applications. The virtual desktop
infrastructure is built upon a VMware Horizon View environment of Microsoft Windows 7 virtual
desktops provisioned by VMware Horizon View Composer. The applications are built with
Microsoft specific application of Exchange, SQL Server and SharePoint.
Planning and designing the server, networking, and storage infrastructure for the VMware
environment is a critical step because the server infrastructure should be sized to handle both the
application and the desktop workload. This workload needs to address operations in terms of
density and scale. The infrastructure should be provisioned to handle the burst of data traffic, and
the shared storage must be able to absorb large bursts of I/O traffic that occur during a workday.
To provide a cost-effective and predictable performance for this type of virtualization
infrastructure, the solution must be able to:
Support a high density of virtual desktops per server
Scale linearly with increases in the number of virtual desktops
Support user workload on the selected applications
Provide low latency and high bandwidth for clustering, provisioning, and storage
interconnect networks
Handle the peak I/O load from clients while maintaining a quick response time
Unified, Pretested, and Validated Infrastructure
Cisco’s virtualization solution binds together the three critical elements of an end-to-end
deployment: the end-user, the network, and the data center. It draws on Cisco’s architectural
advantage to provide a solution that supports a diversity of endpoint devices, extends pervasive
security and policy management to each virtual desktop, and uses a new and innovative
virtualization-optimized stateless server computing model (Cisco UCS).
Base Components
The computing platform with Cisco UCS includes:
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 8
o Cisco UCS 6200 Series Fabric Interconnects
o Cisco UCS 2200 Series Fabric Extenders
o Cisco UCS 5108 Blade Server Chassis
o Cisco UCS B-series Blade Servers for applications and virtual desktop hosting
Hypervisor: VMware vSphere
Virtual desktop connection broker: VMware Horizon View
Microsoft Business Critical Applications:
o Exchange
o SQL Server
o SharePoint
Storage platform with Nimble CS-Series
Cisco Unified Computing System
Cisco UCS is the first truly unified data center platform that combines industry-standard, x86-
architecture blade and rack servers with networking and storage access into a single system.
Innovations in the platform include a standards-based unified network fabric; Cisco virtualized
interface card (VIC) support, and Cisco Extended Memory Technology. The system uses a wire-
once architecture with a self-aware, self-integrating, intelligent infrastructure that eliminates the
time-consuming, manual, error-prone assembly of components into systems.
Cisco UCS B-Series Blade Servers provide a comprehensive line of 2 and 4-socket servers to
deliver world-record-setting performance to awide range of workloads. Based on Intel® Xeon®
processor E7 and E5 product families, these servers are excellent for virtualized and non-
virtualized applications. These servers:
Reduce CapEx and OpEx with converged network fabrics and integrated systems
management
Deliver performance, versatility, and density without compromise
Address a broad set of workloads, including IT and web infrastructure and distributed
databases, for both virtualized and non-virtualized environments
Increase IT staff productivity and business agility through just-in-time provisioning and
mobility support for both virtualized and non-virtualized environments
Cisco focuses on three main elements to deliver the best desktop virtualization data center
infrastructure: simplification, security, and scalability. The software in combination with platform
modularity provides a simplified, secure, and scalable desktop virtualization platform.
Simplified
o Virtual desktop density per sever
o Unified management, providing common view of the platform
o Predefined, validated infrastructure
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 9
Secure
o Virtual desktop–aware access and control policies
o Virtual desktop–aware networking and on-demand provisioning
o Segmentation and network security policies across LAN and WAN
Scalable
o Capability to linearly scale up to thousands of desktops in a single domain
o Rapid desktop provisioning through service profiles
o Low-latency, high-bandwidth network for virtual desktop and multimedia delivery
The simplified, secure, scalable Cisco data center infrastructure solution for desktop virtualization
saves time and money. It provides faster payback and ongoing savings (better return on
investment [ROI] and lower TCO) with the industry’s highest virtual desktop density per server,
meaning that fewer servers are needed, reducing both capital expenditures (CapEx) and
operating expenses (OpEx). The solution also has much lower network infrastructure costs, with
fewer cables per server and fewer ports required, through the use of the Cisco UCS architecture
and unified fabric.
The simplified deployment of Cisco UCS for desktop virtualization accelerates time to productivity
and enhances business agility. IT staff and end users are more productive more quickly, and the
business can respond to new opportunities by simply deploying virtual desktops whenever and
wherever needed. The high-performance Cisco system and network deliver a near-native end-
user experience, allowing users to be productive anytime and anywhere.
VMware vSphere
VMware vSphere is the market-leading virtualization platform that is used in thousands of IT
environments around the world. VMware vSphere transforms a computer’s physical resources by
virtualizing the CPU, RAM, hard disk, and network controller. This transformation creates fully
functional virtual desktops that run isolated and encapsulated operating systems and applications
just like physical computers.
The high-availability features of VMware vSphere are coupled with VMware Distributed Resources
Scheduler (DRS) and vMotion, which enables the transparent migration of virtual desktops from
one VMware vSphere server to another with little or no impact on the customer’s use.
This reference architecture uses VMware vSphere Enterprise Plus edition for deploying desktop
virtualization. This provides the full range of features and functions of the VMware vSphere
allowing customers to achieve scalability, high availability, and optimal performance for all their
desktop workloads.
Horizon View
VMware Horizon View is a desktop virtualization solution that simplifies IT manageability and
control while delivering one of the highest-fidelity end-user experiences for devices and networks.
The VMware Horizon View solution helps IT departments automate desktop and application
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 0
management, reduce costs, and increase data security through centralization of the desktop
environment. This centralization results in greater end-user freedom and increased control for IT
departments. By encapsulating the operating systems, applications, and user data into isolated
layers, IT departments can deliver a modern desktop. IT can then deliver dynamic, elastic
desktop cloud services such as applications, unified communications and 3D graphics for real-
world productivity and greater business agility.
Unlike other desktop virtualization products, VMware Horizon View is built on, and tightly
integrated with, VMware vSphere, the industry-leading virtualization platform, allowing customers
to extend the value of VMware infrastructure and its enterprise-class features such as high
availability, disaster recovery, and business continuity. VMware Horizon View includes many
enhancements to the end-user experience and IT control. Some of the notable features include:
VMware Horizon View Storage Accelerator (VSA): In memory host-based read cache that
reduces storage IO operations to the storage subsystem during Login/boot/AV storms.
VMware PCoIP Optimization Controls: Deliver protocol efficiency and enable IT
administrators to configure bandwidth settings by use case, user, or network
requirements and consume up to 75 percent less bandwidth
VMware PCoIP Continuity Services: Deliver a smooth end-user experience regardless of
network reliability by detecting interruptions and automatically reconnecting the session
VMware PCoIP Extension Services: Allow Microsoft Windows Management
Instrumentation (WMI)–based tools to collect more than 20 session statistics for
monitoring, trending, and troubleshooting end-user support problems
VMware Horizon View Media Services for 3D Graphics: Enable VMware Horizon View
desktops to run basic 3D applications such as Microsoft Windows Aero and Office 2010
or those requiring OpenGL or DirectX, without the need for specialized graphics cards or
client devices
VMware Horizon View Media Services for Integrated Unified Communications: integrate
voice over IP (VoIP) and the VMware Horizon View desktop experience for the end user
through an architecture that optimizes performance for both the desktop and unified
communications
VMware Horizon View Persona Management (VMware Horizon View Premier edition
only): Dynamically associates a user persona with stateless floating desktops; IT
administrators can deploy easier-to-manage stateless floating desktops to more use
cases while enabling user personalization to persist between sessions
VMware Horizon View Client for Android: Enables end users with Android-based tablets
to access VMware Horizon View virtual desktops
Support for VMware vSphere uses the latest functions of the leading cloud infrastructure platform
for highly available, scalable, and reliable desktop services.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 1
Nimble Storage CS-Series
Nimble Storage arrays are the industry’s first flash-optimized storage designed from the start to
increase efficiency. Built on the patented Cache Accelerated Sequential Layout (CASL)
architecture, Nimble Storage offers scalable performance, exceptional efficiency, integrated data
protection, and simple push-button management. As a result, customers can run more workloads
and perform more backup operations with less storage infrastructure and empower IT to take on
new projects with higher returns. CASL accelerates applications by using flash memory as a read
cache in conjunction with a write-optimized data layout. It protects data by supporting instant
snapshots for easy backup and restoration, along with efficient replication for disaster recovery.
Nimble Storage’s intuitive, automated tools greatly simplify storage and data management,
reducing operations overhead.
Nimble Storage delivers:
Accelerated performance for greater throughput and IOPS, and sub-millisecond latencies
Greater storage efficiency, reducing the storage footprint needed by 30 to 75 percent
Non-disruptive scaling to fit changing application needs through increased performance or
capacity, or both
Increased data and storage availability with integrated data protection and disaster
recovery
Simplified storage management and reduced day-to-day operation overhead
Benefits
High performance: Nimble Storage arrays with Cisco UCS deliver the adaptive
performance needed to handle comprehensive virtualization workloads delivered by
VMware Horizon View and Microsoft applications, in a compact footprint. This capability
allows IT to maintain a positive user experience through application use, boot and login
storms, patch operations, and updates.
Greater consolidation: VMware Horizon View frees IT departments from the chores of
desktop upgrades by centralizing the creation and deployment of linked clones. This
feature compliments Nimble Storage’s in-line compression and the dense computing and
I/O capabilities of Cisco UCS. IT departments that have deployed the joint platform have
seen 30 to 75 percent reduction in their storage footprint compared to traditional solutions,
with no degradation of the end-user experience.
High availability: Nimble Storage and Cisco UCS both incorporate redundant
components, with no single point of failure, along with proactive monitoring and reporting.
These features complement VMware vCenter’s capability to trigger efficient Nimble
Storage snapshots for instant backup and fast restore operations. The result is less end-
user disruption and fewer calls to the help desk.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 2
Expansion on demand: By integrating computing, fabric, storage, and virtualization
resources in a single architecture, IT can start small and independently scale horizontally
or vertically to large deployments supporting thousands of virtual desktops as needs
grow.
Architecture and Design
The converged infrastructure solution was built with Nimble Storage as the foundation and only
storage for the environment. The Cisco UCS servers and network were connected using only the
Fabric Interconnect methods. VMware vSphere provided the hypervisor platform. The details of
the solution are outlined below.
VMware vSphere, Cisco UCS, Nimble Storage
The following figure shows the basic components and connectivity used to for the server, storage,
and network. A single Nimble Storage array was connected to the Cisco UCS chassis through a
pair of Cisco UCS fabric interconnects using appliance ports. The VMware Infrastructure
components and the VMware Horizon View desktops were all run on the blades in Cisco UCS.
Figure 2 - High Level Physical Topology
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 3
BCA Cluster
Business Critical Applications
Excha ng e
2010loa dg en
Sha rePoint
DB
Excha ng e
DB
Sha rePoint
2013
Horizon Cluster
125 Microsoft Windows7Desktops per
Cisco B-series balde
500 desktops
4 blades
...
Horizon View Desktops
VMware View Infrastructure
AD/DN S/
DH CP
Virtual
Center
View
Composer
View
Connect
View
Pla nnerM S KM S
M S S QL
vCO ps
Figure 3 - High Level Logical Topology
Virtualizing Business Critical Applications
When virtualizing business critical applications, there are four key areas we will want to address.
They are:
Availability
Performance
Data Protection
Operational Management
These will be addressed in the remainder of this section.
Availability
When you virtualize business critical applications, you want to ensure the entire infrastructure has
no single point of failure, for both hardware and software, across all layers (compute, network,
storage, VM and applications). Here is a list of design considerations:
UCS
Cisco UCS blade chassis has redundancy for all components
Two UCS blade servers in case one of them fails
Dual UCS Fabric Interconnect configured as a cluster
UCS fabric NIC failover is used for management and virtual machine traffic
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 4
SLOT
1
SLOT
5
SLOT
3
SLOT
7
SLOT
2
SLOT
6
SLOT
4
SLOT
8
!
UCS 5108
OK FAIL OK FAIL OK FAIL OK FAIL
CISCO UCS 6248UP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
STAT
ID
CISCO UCS 6248UP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
STAT
ID
! Reset Console
UCS B230 M1/M2
!2!1
A03-D0100SSD-LH 100GB SSD SATA>> A03-D0100SSD-LH 100GB SSD SATA>>! Reset Console
UCS B230 M1/M2
!2!1
A03-D0100SSD-LH 100GB SSD SATA>> A03-D0100SSD-LH 100GB SSD SATA>>
1
VMware vSphere
2
3
Figure 4 - UCS Availability – Key Points
Figure 5 - Redundant Networks
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 5
Figure 6 - Automated Network Failover
Storage
Nimble Storage CS-series array has redundancy for all components
All volumes provisioned use SATP_ALUA & PSP_RR for path failover and load
distribution
Figure 7 – Storage MPIO – Round Robin Path Selection
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 6
Set iops=0 for
each volume
All volumes
should have
“iops=0”
Figure 8 – ESXi Storage Device Settings
vSphere
vSphere HA enabled to auto restart VMs in case ESXi server fails
Host monitoring is enabled to monitor heartbeat of all ESXi hosts in the cluster
Admission control is enabled to ensure the cluster has enough resources to
accommodate a single host failure
N+1 configuration to tolerate for one ESXi host failure
vSphere Virtual Switch layout (only single vNIC is needed as UCS Fabric failover is
enabled for each management and virtual machine traffic vNIC; more on the iSCSI
vSwitch later)
Figure 9 - Hypervisor HA
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 7
Figure 10 - Virtual Network Separation
Performance Optimization
This section covers various performance optimization considerations in the configuration of the
SmartStack solution.
SLOT
1
SLOT
5
SLOT
3
SLOT
7
SLOT
2
SLOT
6
SLOT
4
SLOT
8
!
UCS 5108
OK FAIL OK FAIL OK FAIL OK FAIL
CISCO UCS 6248UP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
STAT
ID
CISCO UCS 6248UP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
STAT
ID
! Reset Console
UCS B230 M1/M2
!2!1
A03-D0100SSD-LH 100GB SSD SATA>> A03-D0100SSD-LH 100GB SSD SATA>>! Reset Console
UCS B230 M1/M2
!2!1
A03-D0100SSD-LH 100GB SSD SATA>> A03-D0100SSD-LH 100GB SSD SATA>>
1
2
3 VMware vSphere
OS
APP
DATA
OS
APP
DATA 4
Figure 11 - Optimization Locations
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 8
UCS (#1)
Dual subnet for directly connecting Nimble to Cisco UCS Fabric Interconnect (without
failover of Fabric for the iSCSI vNICs)
Figure 12 - iSCSI Storage Subnets
Storage (#2)
Storage Volume layout
Volumes supporting the infrastructure
Volumes supporting the application
Performance Policy for each storage volume:
Use PSP_RR to distribute I/O across both paths
Change default path IOPS to 0
Infrastructure AD
vCenter Server
vCenter Opreations
OS VMDK for all VMs
VMKD for Sharepoint
(web/app tiers)
Dedicated datastore
for all VM swap
(.vswp)
bizappVMswap
Boot volume
(ESXi1)
Boot volume
(ESXi1)
Figure 13 – Infrastructure Specific Volume Assignemnt
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 1 9
ExchangeDB
ExchangeLog
SQL2012DB
SQL2012Log
Figure 14 - Application Specific Volume Assignment
For VM .vswp swap
For infrastructure VMs and
app VM OS VMDKs
Performance Policy
Block size 4K
Compression: ON
Cache: Disabled
Block size 4K
Compression: ON
Cache: Enabled
Figure 15 - Application Aware Performance Policies
vSphere
One VMkernel port for each of the iSCSI vNIC
Software iSCSI initiator binds to two VMkernel ports
Separate OS, Data, log into its own VMDK, dedicated virtual SCSI adapter, and use
vmnxet3 as the virtual adapter
o Exchange, SQL Server, SharePoint
VM Guest OS:
o If upgraded from Windows 2003, be sure to align the VM (change partition
starting offset to be divisible by 4KB)
o NTFS allocation unit size for data/log partitions should be 64KB
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 0
Figure 16 - vmkernel Network Allocation - vSwitch
Figure 17 - vmkernel Network Allocation - Details
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 1
Separate VMDK
for OS, DB, Log
Separate vSCSI
HBA for each VMDK
Separate vmnic for
MAPI and DAG
replication traffic
Figure 18 - VM Properties - Exchange
Separate
vSCSI HBA
for each
VMDK
Separate VMDK
for OS, DB, Log
Figure 19 - VM Properties - SQL Server
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 2
Separate
vSCSI HBA
for each
VMDK
Separate VMDK for
OS and Application
Figure 20 - VM Properties - SharePoint
Data Protection
Infrastructure
Backup UCSM configuration on a regular basis (service profile templates, service profiles, all
environmental configurations for the Fabric Interconnect), especially after changes have been
made (for example, modification to service profile, configuration of ports/VLANs in the Fabric
Interconnect)
Backup ESXi sever boot volumes and infrastructure VMs (including SharePoint Web/App tier) by
placing all boot volumes into a single Volume Collection with daily snapshot (NOTE: No snapshot
synchronization is needed as crash consistent snapshot is all that’s needed)
Application
Ensure application consistent snapshot can be taken through Nimble and VMware integration
Exchange
o For simplicity, each Exchange mailbox database is configured with circular
logging
o NOTE: The ability to perform log truncation is provided through add-on products
such as CommVault Simpana with Nimble Storage integration or vSphere Data
Protection
o VMware VSS integration is used to properly quiesce Exchange database for
application consistent snapshot
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 3
o VMware vCenter Synchronization is used for the Exchange Volume Collection
(the volume collection contains both Exchange database and log datastores)
SQL
o For simplicity, each SQL database is configured with simple recovery mode
NOTE: The ability to perform full recovery is provided through add-on products
such as CommVault Simpana with Nimble Storage integration or vSphere Data
Protection
o VMware VSS integration is used to properly quiesce SQL database for
application consistent snapshot
o VMware vCenter Synchronization is used for the SQL Volume Collection (the
volume collection contains both database and log datastores)
SharePoint
o The SharePoint Web/Application tier VMDK resides in the Infrastructure volume
which is backed up daily. Note the SharePoint database is backed up through
SQL Server volume collection
Figure 21 - Backup UCSM
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 4
Infrastructure
ESXi1 boot
ESXi2 boot
Infrastructure VolCol
(daily snapshot)
Figure 22 - Nimble Volume Collection Protection Scheme
1
2
3
Figure 23 - Nimble Storage Snapshot Synchronization - Exchange
2
1
3
Figure 24 - Nimble Storage Snapshot Synchronization - SQL Server
Operational Management
In this chapter we will highlight tools and integrations that help making deployment and
operational management simple and easy.
Server Deployment with Cisco UCS Service Profile
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 5
A custom UCS Service Profile template was created for vSphere. It creates a standard for
deploying the vSphere environment serving business critical applications, and simplifies scalability
expansion down the line. We created two service profiles based on this ESXi template, apply it to
each blade, and then modify the boot target for each server. That is it – all subsequent servers
that will be added to the environment serving business critical applications will follow the same
steps. Here’s what the service profile template looks like:
For vNIC
Figure 25 - UCS vNIC Allocation
iSCSI vNICs settings for Boot-from-SAN
A1: subnet A for iSCSI boot vNIC A
A2: iSCSI Discovery IP address for Nimble Array
B1: subnet B for iSCSI boot vNIC B
B2: iSCSI Discovery IP address for Nimble Array
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 6
A1
A2
B1
B2
Figure 26 - iSCSI Boot from Nimble
Storage Management with Nimble Storage vCenter plugin
Don’t want to toggle between different UIs to perform storage related tasks? Just stay in vCenter
Server. Nimble Storage plugin allows for new datastore provisioning, cloning, resizing,
snapshotting, and monitoring performance statistics, space usage, and compression savings:
Performance
/Space
monitoring
Figure 27 - vCenter Storage Integration
Operational Management with vCenter Operations Manager
After the environment has been deployed, use vCenter Operations Manager to monitor health,
workload and faults in the infrastructure. Good practice is to pay attention to any red icon(s) for
Health, Workload and Fault badges, as well as “Alerts”:
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 7
Figure 28 - vCenter Operations Manager
You could also leverage the Group view functionality to look at the current health and workload
status of all the VMs by their grouping folder:
NOTE: It is recommended to create a custom group with all Nimble Storage datastores. Doing so
allows for quick overview of the health and workload status of the Nimble array volumes. Nimble
InfoSight could then be used to look at deeper statistics based on heartbeats sent from the array.
Figure 29 - vCOps - Applications
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 8
Figure 30 - vCOps Health
Deep Data Analytics with Nimble InfoSight
It is a good practice to regularly monitor Nimble InfoSight for storage health, availability,
performance, data protection reports based on heartbeats from the array:
Wellness tab shows alerts from the array (both hardware and software), as well as support cases
that have been open automatically based on criticality of the alerts:
Figure 31 - InfoSight Wellness Overview
Capacity tab shows current array space utilization, as well as projection of when the array would
run out of capacity:
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 2 9
Figure 32 - InfoSight Capacity Usage
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 0
Performance tab shows CPU and cache utilization of the array, as well as average read and write
latency based on heartbeat sent by the array:
Figure 33 - InfoSight Performance View
Data Protection tab shows snapshot/replication configuration for each volume within the Nimble
array:
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 1
Figure 34 - InfoSight Protection Coverage View
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 2
Dashboard tab shows summary reports of space savings through compression, data protection
level for each volume, snapshot retention duration as well as upgrade recommendations based on
workload
Figure 35 - InfoSight Executive Dashboard
Modular Virtual Desktop Infrastructure
The Horizon View desktop virtualization environment was configured with the following
components:
Cisco UCS B-series (10-Gbps Cisco UCS)
Nimble Storage CS200 Series array (10-Gbps iSCSI)
VMs to support the following functions:
o Active Directory, DNS, and DHCP
o Microsoft KMS
o VMware View Planner
o VMware vCenter
o Microsoft SQL Server
o VMware Horizon View Composer
o VMware Horizon View Connect
125 Microsoft Windows 7 Desktops per Cisco UCS B-series blade
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 3
Cisco UCS B-Series Blade Servers
The VMware Horizon View test was performed on a Cisco UCS blade system targeted to host
approximately 125 desktops per blade. For the targeted 500 desktop user test, four blades were
configured into a VMware VDI service cluster. For details on the specific blades used, refer to
Appendix A.
VMware vSphere
For this test, VMware vSphere and Horizon View were deployed. All Cisco UCS blades were
configured to iSCSI boot the VMware ESXi hosts from the Nimble Storage.
A VMware Horizon View Infrastructure host blade was used to support the VDI components,
which include the following Microsoft Windows Server 2008 R2 virtual machines:
Microsoft Windows Active Directory, Domain Name System (DNS), and Dynamic Host
Configuration Protocol (DHCP)
Microsoft Key Management Server (KMS)
VMware vCenter Server
VMware Horizon View Connection Server (View Manager)
VMware Horizon View Composer
Microsoft SQL Server (to support the VMware infrastructure)
VMware View Planner appliance (Linux based) to run the real-world workload tests
The desktop virtual machines were run on a virtual desktop cluster composed of four dedicated
Cisco UCS B-series blades.
The desktop configuration used for this test was a Microsoft Windows 7 Professional (x86 32-bit)
virtual machine with 1 GB of RAM, one virtual CPU (vCPU), and a 24-GB disk.
The software configuration of the desktop application environment consisted of:
Microsoft Office 2007 (Word, Excel, and PowerPoint)
Microsoft Outlook
Adobe Reader
Mozilla Firefox
7-Zip tools
Also configured in the Microsoft Windows 7 golden image were the VMware Horizon View Agent
and the VMware View Planner Agent. The configuration of the golden image followed the
VMware View Planner guide for optimizations, system settings, and other configuration
requirements. For details about the specific setup, refer to Appendix A of the VMware View
Planner Installation and User Guide.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 4
VMware Horizon View
VMware Horizon View was used to construct and manage the desktop virtual machines. This test
used a floating pool of non-persistent linked-clone desktop virtual machines. The details of the
specific VMware Horizon View pool definition are included at the end of this document.
In addition to the linked-clone method of deploying desktop virtual machines, VMware Horizon
View uses VMware VSA technology. This technology allows a small amount of VMware ESXi
host RAM to be used as a content-based read cache for the selected desktops’ read I/O
operations. VMware VSA was configured at 2 GB for each VMware ESXi host and for each
VMware Horizon View pool used for this test.
VMware View Planner
To perform the workload test, the VMware View Planner tool was used to run a sample workload
on each of the 500 active Microsoft Windows 7 desktops. The VMware View Planner tool can be
configured in many ways for configuration testing. This test followed the guidelines of the VMware
Rapid Desktop Program (RDP) testing. The result of the VMware View Planner workload is close
to that for a typical Microsoft Windows–based task worker with a steady-state IOPS rate of about
7 IOPS per desktop.
Nimble Storage CS-Series
The Nimble Storage system used for this testing was a CS220G-X4 model that has 12 TB of raw
capacity (12 x 1-TB HDD), 1200 GB of SSD-based flash memory, and dual 10 Gigabit Ethernet
data network interface card (NIC) connections per controller. The Nimble Storage CS220G-X4
has two controllers operating in an active-standby redundant configuration.
For the 500 desktop virtual machines solution, the storage volumes were configured to have no
more than 125 desktop virtual machines in each VMware Virtual Machine File System (VMFS)
datastore. Each datastore volume was configured with 3 TB, using thin provisioning at the
storage level and providing plenty of room at the linked-clone level for conservative desktop
provisioning methods. The Nimble Storage appliance was then configured with 5 of the 3 TB
iSCSI volumes presenting 5 individual datastores at the VMware level. Each Cisco UCS blade
had shared access to each volume. Through the VMware Horizon View pool balanced allocation,
each volume contained approximately 100 desktop linked-clone virtual machines.
The VMware Horizon View storage component that mapped to the Nimble Storage solution used
a simple and easy-to-manage approach: everything for the desktop was in the same datastore.
With Nimble Storage capabilities, it is not necessary to separate the various VMware Horizon
View storage artifacts into different storage tiers or locations. The CASL file system is optimized
to handle VDI workloads, including boot storms. During a boot storm, the virtual desktop replica
image is automatically cached in the built-in SSDs. As a result, boot times are greatly reduced,
even when large numbers of virtual desktops boot at the same time. During steady-state
operations, in which end users access applications in a random fashion, random write I/O is also
optimized with the CASL write-coalescing technology.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 5
Solution Validation
As part of the reference architecture development practice, we constructed an example
configuration of Nimble Storage, Cisco UCS and VMware ESXi hosts for running the sample load
test utilities from the desired server application and desktop virtualization solution sets.
The next sections will describe some of the detail of the solution configuration.
Server: Cisco UCS
The Cisco UCS configuration was connected to the Nimble Storage as an iSCSI array directly to
the Cisco fabric interconnect with dual, redundant 10-Gbps connections. This architecture
provides the most efficient use of hardware and lowers the total cost (cost per desktop) of the
server, network, and storage solution.
The chassis is connected to each fabric interconnect with four 10-Gbps network connections.
Active/Passive Fabric
Interconnect
Active-Passive Fabric
Interconnect
Figure 36 – UCS Fabric Interconnect – Appliance Port VLAN Connectivity
Active-Passive Fabric Interconnect
For a configuration without an access-layer switch (Cisco Nexus® 5000 Series Switches), the
fabric interconnect ports connected to the Nimble Storage array must be configured to operate in
Appliance Port mode.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 6
Figure 37 - UCS Appliance Port Configuration
The Cisco UCS B-series Blade Servers were configured with multiple virtual NIC (vNIC) interfaces
to allow separation of management traffic, VDI desktop traffic, and iSCSI data traffic. The
following figure shows a typical server service profile.
Figure 38 - Cisco UCS Blade Server vNIC Configuration
Each Cisco UCS blade server is configured with five vNIC interfaces, with the VMware ESX
vSwitch network configuration shown here:
vSwitch0 (management network)
vmnic0
vSwitch1 (iSCSI data traffic)
vmnic1 and vmnic2 (no teaming; VMware ESX software iSCSI multipathing is enabled
with PSP_Round_Robin algorithm)
vSwitch2 (virtual desktop traffic)
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 7
vmnic3 and vmnic4 (teamed for load distribution and failover)
Figure 39 - VMware Network Configuration
Figure 40 - VMware iSCSI Port 1 Configuration Details
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 8
Figure 41 - VMware iSCSI Port 2 Configuration Details
Figure 42 - VMware Desktop Virtual Machine Network Configuration Details
Storage: Nimble Storage CS-Series
The Nimble Storage CS220G-X4 was configured with multiple volumes presented to the Cisco
UCS VDI cluster as shown in the figures that follow. Each of the 5 volumes was configured as 3
TB of thin-provisioned iSCSI SAN storage VMFS datastore.
The iSCSI initiators of the Cisco UCS blades were defined in a single initiator group for ease of
management and presentation. Nimble Storage allows all host initiator iSCSI qualified names
(IQNs) to be grouped into a single group, which eases the need to provision storage for multiple
groups of VMware ESX host initiators. This approach is a best-practice recommendation for
VMware vSphere environments, because it prevents inconsistent performance policy settings and
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 3 9
volume presentations to a VMware ESX host cluster that shares the same group of volumes. The
following figure shows the grouping of all VMware ESX hosts in the VDI environment sharing a
single initiator group. The initiator group is constructed by adding the IQN information for each
blade’s iSCSI initiator in a single combined group representing the VMware ESX cluster formed by
the blades.
Figure 43 - Nimble Storage iSCSI Initiator Setup
The 5 VMFS datastores were then configured into separate VMware Horizon View pools as
described in Appendix C VMware Horizon View Pool Details.
Figure 44 - Nimble Storage Volume Management
VDI: VMware
The VMware vSphere and View environment was set up with a single cluster of VMware ESXi
hosts for the VDI configuration.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 0
Figure 45 - VDI Cluster Configuration – vCenter Organization
Four of the Cisco UCS blades in this chassis were used to host the Microsoft Windows 7 desktop
virtual machines, and a separate blade was used to provide the core virtualization infrastructure
services as described in the following paragraphs.
Testing Methodologies
This section will describe the basic testing approach taken to validate the reference architecture
capabilities and performance.
To validate the capability to run up to 500 desktop virtual machines on the Nimble Storage
CS220G-X4 array, the VMware View Planner tool was used to run sample tests for specific
configurations. The goal was to exercise and measure the storage infrastructure under a
reference workload. The workload consisted of the following:
500 desktop virtual machines configured into five VMware Horizon View pools of 100
desktops each
Four blade servers in a Cisco UCS and VMware cluster, each running approximately 125
virtual machines based on VMware Horizon View pool allocations for the cluster resource
All 500 virtual desktops configured in floating pools (they represent non-persistent
desktops used by task workers, with random workloads throughout the day)
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 1
As shown in the workload profile in the figure below, a typical VMware View Planner configuration
consists of a set of applications to exercise and a set of virtual machines on which to run that
workload profile.
The next figure shows examples of the VMware View Planner Workload Profile and the VMware
View Planner Run Definition used for the test of the Nimble Storage and Cisco UCS VDI solution.
For this test, all applications were selected for the workload profile. The Multimedia Application
value was set to Medium but unchecked for this test scenario. The Iterations value should be set
to at least 7 to allow ramp up and ramp down of the individual desktop workloads during the test
run and still obtain 5 iterations for steady-state measurements. For this test scenario we used an
iteration count of 10 to mimic the ramp up and down of roughly an 8 eight work day and coincided
with the 10 hour runs of Exchange Loadgen and SQL DVD-Store load testing.
The Think Time setting was used to adjust the intensity of the desktop activity. Think Time
dictates the average interval that each user session waits before moving on to the next workload
profile task and can be a value between 5 and 20. For the 500 desktop tests for average use
cases, a Think Time value of 20 was used.
This selection of run profile characteristics is representative of a workload in the range of 7 to 10
IOPS per user, or comparable to a user roughly between a task worker (3 to 7 IOPS) and
knowledge worker (7 to 10 IOPS).
Figure 46 - VMware View Planner Workload and Run Profile
Boot and Login Scenarios
The VMware View Planner tool allows control of an orderly boot and login sequence when you are
performing a test run. For the workload tests performed, the desktop virtual machines were
rebooted each time before the workload was processed. This approach allowed measurement of
the boot and login process in addition to the measurement of the actual desktop workload
processes.
The four Cisco UCS blades are configured in a VMware cluster, and the VMware Horizon View
pools and datastores are shared across the set. This approach provides greater simplicity of
management for the environment and allows better distribution of the computing assets.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 2
The total time to boot 500 desktops and have the VMware View Planner agent log in is
approximately 15 minutes. This time is measured by the completion of the VMware Horizon View
Agent registration with the VMware Horizon View Administrator and the VMware View Planner
Agent registration with the VMware View Planner appliance. Keep in mind that the desktop
sessions are booting to an operational state and automatically logging in the test user
(administrator), who then connects to the VMware View Planner harness. This user registration
takes a little extra time, but it allows you to monitor the ready state of the desktop virtual machines
under test.
The effect on the Nimble Storage appliance during the boot storm period can be seen in the next
figure, which contains two graphs:
MBps throughput
IOPS
These graphs show the performance for the “typical day” test run used and shows that the storage
infrastructure is delivering data well within the operational specifications expected. The system
under test is capable of approximately 15,000 mixed-workload IOPS during the initial boot/login
phase. Throughput is simply a result of the I/O rate and block sizes coming from the desktop
virtual machines.
Figure 47 - Nimble Storage Performance - Day View
Steady-State Scenarios
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 3
As you can see, the transition to desktop user steady state around 8:00 AM causes a dramatic
shift in the number of random read operations required to support the desktop workload. As the
desktop virtual machines shift into various VMware View Planner tasks as defined by the workload
profile, the random write operations become the dominant factor, at three to four times the number
of random read I/O operations.
The following figure from the View Planner results documentation shows that during the steady
state, the CPU utilization of the desktop ESXi host is within the performance bounds desired for
each blade in the Cisco UCS blade system.
Figure 48 - Cisco UCS Blade - Steady State Workload - CPU Utilization
Conclusion
The combined desktop and server solution test conducted with the Cisco UCS B-Series Servers,
Nimble Storage array, and VMware vSphere using Microsoft applications and Horizon View
showed that a robust yet simple configuration is possible for up to 500 users in a mixed workload
environment on the same easy to manage, scalable and well performing virtualization
infrastructure.
When you virtualize business critical applications such as Microsoft Exchange, SQL and
SharePoint, be sure to design the architecture with the four key pillars of requirements in mind:
availability, performance, data protection and operational management. This document highlights
the key design principles and best practices that address the requirements from all four pillars.
Virtualize with confidence using SmartStack, powered by Cisco, VMware and Nimble Storage.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 4
Appendix A: Bill of Materials
Table 1 shows the bill of materials (BOM) for the test configuration.
Vendor Component Model (Quantity) Version
Cisco Cisco UCS B230 M2 Blade Server (6) 2.1(3a)
Cisco UCS 2200 Series Fabric Extenders
Cisco UCS 5108 Blade Server Chassis (1)
Cisco UCS 6248 48-Port Fabric Interconnect (2)
Nimble Storage Nimble Storage CS220G-X4 (1) 1.4.9.0
VMware VMware vSphere ESXi Enterprise Plus 5.1 U1
VMware vSphere vCenter Server Standard 5.1
vCenter Operations Manager 5.7
VMware Horizon View Premier 5.3
VMware View Planner (optional) 3.0.1
Note:
The B.O.M listed above is a reference design of an environment capable of supporting 500+ users
with business critical applications. Customers and partners are welcome to use different models
of equipment from Cisco for compute, and Nimble for Storage. For example, Cisco UCS C-series
rack mountable servers or other blade models, and a Nimble CS400 series could be used in place
of the CS200 series, depending on the workload and capacity needs.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 5
Appendix B: Validation for 500-User Business Critical Applications Environment
High Level Environment Overview:
CISCO UCS 6248UP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
STAT
ID
CISCO UCS 6248UP 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
STAT
ID
SLOT
1
SLOT
5
SLOT
3
SLOT
7
SLOT
2
SLOT
6
SLOT
4
SLOT
8
!
UCS 5108
OK FAIL OK FAIL OK FAIL OK FAIL
! Reset Console
UCS B230 M1/M2
!2!1
A03-D0100SSD-LH 100GB SSD SATA>> A03-D0100SSD-LH 100GB SSD SATA>>! Reset Console
UCS B230 M1/M2
!2!1
A03-D0100SSD-LH 100GB SSD SATA>> A03-D0100SSD-LH 100GB SSD SATA>>
VMware vSphere
VMware vCenter
Nimble Storage InfoSight
vCenter Operations Manager
Figure 49 - Business Critical Applications
Exchange 2010:
1 Database Availability Group
CAS/Hub/Mailbox Server 1
Mailbox DB 1Mailbox DB 2
(copy) Mailbox DB 2Mailbox DB 1
(copy)
250 mailboxes 250 mailboxes
CAS/Hub/Mailbox Server 2
Exchange LoadGen VM
Figure 50 - Exchange 2010 Architecture
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 6
Figure 51 - Exchange 2010 Detailed Setup
SQL 2012 and SharePoint 2013:
SQL2012 Infrastructure SQL2012 Application
vCenter Server DB
DVDStore DB
SharePoint DB
VMware
vCenter Server
DVDStore SQL Load Driver VM
MyNimble Intranet Powered by SharePoint
Web/App Tier
Figure 52 - SQL and SharePoint Database Architecture
In case you are wondering how the SmartStack solution performs with real applications, here are
the details of the validation:
In short, the physical servers, VMs hosting the applications, and the Nimble CS220G-X4 array did
not show any signs of resource starvation. The environment could definitely take on additional
workload. We leverage vCenter Operations Manager to determine the impact of running all
workloads simultaneously, and here are the results:
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 7
Summary of observations:
Mixture of Exchange, SQL and SharePoint workload shows both random and sequential read and
write, with bursts of up to 15,000 IOPS
The SmartStack architecture is well equipped to handle the mixture of workloads without signs of
resource starvation for CPU, memory, network or storage (as shown in vCenter Operations charts
below)
Nimble CS220G-X4 array shows average latency of under 2 ms for both read and write IO
Details:
Application Validation Tool Workload Profile
Microsoft Exchange 2010 LoadGen Version 14. 01.
0180. 003
500 1GB mailboxes (250 in
each DAG node with cross
replication); Outlook_150
action profile (150
messages/day); total of 10
hour test simulating 8 hour
busy work day
Microsoft SQL Server 2012 DVDStore Version 2. 1 Large DVDStore database
with 1 million customers and 2
million DVD products
Microsoft SharePoint 2013 Nimble Storage employees Day-to-day cross functional
usage of SharePoint farms for
page creation, modification,
file upload and sharing
NOTE: Validation was conducted with all three workloads and VMware View Planner 500 user
desktops running simultaneously.
Results:
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 8
Figure 53 – Exchange LoadGen Results
DVDStore Results:
Total test run duration: 36018 minutes (~10 hours)
Total transactions completed: 1892280 orders
Total new customers added: 378376
Total number of browse during run: 5677543
Total number of purchases: 1892280
Average latency per second to login to DVDStore: 6 millisecond
Average latency to add new customer: 1 millisecond
Average latency to browse catalog: 1 millisecond
Average latency to purchase: 9 millisecond
Figure 54 - DVDStore Run Results
SharePoint Access
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 4 9
Nimble employees across HR, Engineering, QA, Product Management, Marketing, IT and Sales
all had access to “MyNimble” (Nimble’s intranet backed by SharePoint 2013 with SQL 2012 back-
end). All team members were able to access various intranet pages, upload and edit shared
documents, while Exchange Loadgen and DVDStore workloads were running on the SmartStack.
Figure 55 - Example SharePoint Page 1
Figure 56 - Example SharePoint Page 2
Resource Utilization
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 0
Figure 57 - ESXi Server 1 Resource Utilization
Figure 58 - ESXi Server 2 Resource Utilization
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 1
Figure 59 - Exchange DAG node 1 Resource Utilization
Figure 60 - Exchange DAG node 2 Resource Utilization
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 2
Figure 61 - SQL DB serving DVDStore and SharePoint databases Resource Utilization
Figure 62 - Nimble Storage CS220G Array IOPS and Latency chart
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 3
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 4
Appendix C: Storage Performance
Boot Storms
Approximately 15 minutes were needed to boot and log in 500 desktops.
Figure 63 - Boot Storm Storage Performance
Steady State
Approximately 8 hours were monitored. This included (as described earlier):
rollout of the test work load on 500 desktops and the cycling through approximately 8 of
the 10 iterations,
Exchange LoadGen workload,
SQL Server DVD-Store load test.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 5
Figure 64 – Combined Working Day Storage Performance
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 6
Appendix D: Horizon View Pool Details
The VMware Horizon View pools were configured as automated, linked-clone, and floating pools.
To test 500 desktops, two VMware Horizon View pools of 300 & 200 desktop virtual machines
were created in each pool.
Figure 65 - VMware Horizon View Pool Configuration
Each pool had separate datastores provisioned to it from the Nimble Storage volumes presented
to the VMware ESXi hosts. Each datastore was targeted to hold 100 desktop VMs. The replica
and OS disks were placed in the same datastore primarily for simplicity of management.
Figure 66 - View Pool Storage Configuration – Pool 1
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 7
Figure 67 - View Pool Storage Configuration - Pool 2
All desktops were provisioned at the start and created during a separate provisioning period
before the testing was performed. Disposable disks were not selected for the particular test
cases.
For this configuration of basic testing, the VMware Horizon View VSA (View Storage Accelerator)
function was not selected.
Figure 68 - VMware Horizon View Advanced Storage Settings
The VM baseline image was updated to the latest version (vmx-09) to allow compatibility with SE
Sparse functionality. Storage reclamation was not configured for the baseline testing.
All other VMware Horizon View pool settings were default values or those specified in the VMware
View Planner Installation and Users Guide.
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 8
Appendix E: VDI Planning and Sizing Questions
You must understand user groups, applications, and data requirements to properly size
computing, networking, and storage resources.
Some general project questions should be addressed at the outset:
Has a VDI pilot plan been created based on the business analysis of the desktop groups,
applications, and data?
Is the infrastructure and budget in place to run the pilot program?
Are the skill sets needed to implement the VDI project available? Can these be acquired
through hiring or contract?
Have end-user experience performance metrics been identified for each desktop
subgroup?
How will success or failure be measured?
The following is a short, non-exhaustive list of sizing questions that should be addressed for each
user subgroup:
What is the desktop OS planned? Microsoft Windows 7 or Windows XP, 32-bit or 64-bit
desktop OS?
How many virtual desktops will be deployed in the pilot? How many are in production?
Will all desktops use Microsoft Windows 7?
How much memory is needed per target desktop group desktop?
Are there any multimedia, Adobe Flash, or graphics-intensive workloads?
What is the endpoint graphics processing capability?
Are any VMware ThinApp applications planned? Will they be packaged or installed?
What is the storage configuration in the existing environment?
Are sufficient I/O operations per second (IOPS) available for the write-intensive VDI
workload?
Will storage be dedicated and tuned for VDI service?
Does the desktop have a voice component?
Is antivirus software a part of the image?
Is user profile management (non-roaming profile based) part of the solution?
What are the fault-tolerance, failover, and disaster recovery plans?
Are there any additional desktop-subgroup-specific questions that need to be addressed?
N I M B L E S T O R A G E S M A R T S T A C K R E F E R E N C E A R C H I T E C T U R E S E R I E S 5 9
Nimble Storage, Inc.
211 River Oaks Parkway, San Jose, CA 95134
Tel: 877-364-6253; 408-432-9600 | www. nimblestorage. com | info@nimblestorage. com
© 2014 Nimble Storage, Inc. . Nimble Storage, InfoSight, CASL, SmartStack, and NimbleConnect are trademarks or registered trademarks of Nimble Storage, Inc. All other trademarks are the property of their respective owners. SS-DSV-REF-0214