Summary of SnapVDI Features and Performance …...by American Megatrends, Inc. Performance Lab...

Click here to load reader

  • date post

    18-Jun-2020
  • Category

    Documents

  • view

    2
  • download

    0

Embed Size (px)

Transcript of Summary of SnapVDI Features and Performance …...by American Megatrends, Inc. Performance Lab...

  • by American Megatrends, Inc.

    Performance Lab Report &

    Architecture Overview

    Summary of SnapVDI Features and Performance Testing Using Login VSI

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 2

    Table of Contents Executive Summary ....................................................................................................................................... 4

    1. Introduction .......................................................................................................................................... 5

    2. Desktop Virtualization with SnapVDI® ................................................................................................... 5

    2.1. SnapVDI® Components .................................................................................................................. 5

    2.1.1. SnapVDI® Manager (SVM) ..................................................................................................... 6

    2.1.2. SnapVDI® Spark ...................................................................................................................... 6

    2.1.3. SnapVDI® Zero Client ............................................................................................................. 6

    2.2. SnapVDI® Reference Architecture ................................................................................................. 7

    2.2.1. Overview ............................................................................................................................... 7

    2.2.2. Architecture Components ..................................................................................................... 7

    2.2.3. Infrastructure ........................................................................................................................ 8

    2.2.4. Network Design Considerations ............................................................................................ 8

    3. Test Results Summary ........................................................................................................................... 9

    4. Test Methodology ............................................................................................................................... 10

    4.1. Test Objectives ............................................................................................................................ 10

    4.2. Test Tool ...................................................................................................................................... 10

    4.2.1. Load Generation.................................................................................................................. 10

    4.3. Test Criteria ................................................................................................................................. 10

    4.4. Test Configuration ....................................................................................................................... 10

    5. Test Results and Analysis .................................................................................................................... 11

    5.1. Test Scenarios ............................................................................................................................. 11

    5.1.1. Storage IOPS ........................................................................................................................ 12

    5.1.2. Personal and Pooled Desktop Solution ............................................................................... 12

    5.1.2.1. Login VSI Office Worker Workload User Test ................................................................. 12

    5.1.2.2. Boot Storm I/O ................................................................................................................ 12

    5.1.2.3. Login Storm I/O ............................................................................................................... 12

    5.1.2.4. Steady State I/O Workload for Office Users ................................................................... 12

    5.1.3. Pooled Desktop Solution ..................................................................................................... 12

    5.1.4. Migration of Virtual Desktops During Failover ................................................................... 13

    5.2. Test Results ................................................................................................................................. 13

    5.2.1. Storage IOPS ........................................................................................................................ 13

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 3

    5.2.2. Pooled and Personal Desktop Solution ............................................................................... 13

    5.2.2.1. Login VSI Testing ............................................................................................................. 13

    5.2.2.2. Boot Storm I/O ................................................................................................................ 14

    5.2.2.3. Login Storm I/O ............................................................................................................... 15

    5.2.2.4. Steady State I/O .............................................................................................................. 16

    5.2.3. Pooled Desktop Solution ..................................................................................................... 16

    5.2.3.1. Login VSI Testing ............................................................................................................. 16

    5.2.3.2. Steady State I/O .............................................................................................................. 19

    5.2.4. High Availability (HA) .......................................................................................................... 19

    6. Configuration ...................................................................................................................................... 19

    6.1. Server Configuration ................................................................................................................... 19

    6.2. SnapVDI® Configuration .............................................................................................................. 20

    6.3. Windows 8.1 Virtual Desktop Configuration .............................................................................. 20

    6.4. Login VSI Test Configuration ....................................................................................................... 20

    7. Conclusion ........................................................................................................................................... 21

    8. References .......................................................................................................................................... 22

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 4

    Executive Summary Virtual Desktop Infrastructure (VDI) deployments have become increasingly prevalent among companies

    and organizations due to VDI’s inherent benefits of enhanced security, manageability, mobility, and

    productivity. The turnkey, all-in-one SnapVDI® solution from AMI includes both software and hardware

    and is deployed in a matter of hours – allowing users to resolve the cost, complexity, and performance

    concerns associated with traditional desktop environments.

    SnapVDI® Solution

    Characteristics Result

    Density 120 Mixed Personal/Pooled desktops on High Availability 150 Pooled desktops on High Availability

    Performance Login VSI baseline of 1,473 ms with a VSImax 4.1 average of 2164 ms (VSImax not reached)

    This technical white paper describes the SnapVDI® solution with detailed notes on its hardware

    components and architecture, including patented SnapVDI® Spark Technology, SnapVDI® Manager

    (SVM) and SnapVDI® Zero Client.

    The document also demonstrates how a modular virtual desktop environment can be deployed in a

    SnapVDI® VDI infrastructure leveraging two SnapVDI® servers in high-availability (HA) mode.

    The SnapVDI® solution was validated using Login VSI and Iometer and was able to host 150 pooled

    virtual desktops with no compromise in user experience and performance. The solution includes built-in

    failover that guarantees high availability with near-zero downtime along with unprecedented boot

    storm and login storm performance.

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 5

    1. Introduction Virtual Desktop Infrastructure deployments are replacing traditional PC environments in

    organizations of all sizes to simplify management, increase security, and reduce overall IT costs. A

    centralized installation of client systems through Virtual Machines (VMs) drastically reduces the

    Total Cost of Ownership (TCO), as well as the complexity of infrastructure management that

    administrators face with individual PCs. In addition, organizations seeking to increase security with a

    private cloud can do so with a VDI solution. SnapVDI® is the end-to-end solution from AMI that helps

    to realize the benefits of VDI and avoid the many pitfalls of integrating components from numerous

    vendors.

    2. Desktop Virtualization with SnapVDI® SnapVDI® is a complete VDI solution that scales from tens of users up to thousands. Unlike most VDI

    component-based products, SnapVDI® is shipped as a ‘Ready-to-Deploy’ solution that makes

    implementing VDI quick and simple. SnapVDI® was architected to leverage the latest virtual desktop

    technologies while keeping in mind the Total Cost of Ownership (TCO).

    Most virtual desktop implementations fail due to the complexity and high cost associated with the

    project. Hundreds of users have shared their desire for VDI, but expressed concerns over the

    roadblocks they had experienced during past VDI evaluations. SnapVDI® resolves these common

    concerns by delivering a lower cost per user than traditional desktops while providing unparalleled

    functionality and performance.

    2.1. SnapVDI® Components The SnapVDI® solution includes:

    SnapVDI® Manager (SVM), which allows administrators to manage and monitor the entire VDI

    environment from a single interface

    SnapVDI® Spark™, a highly patented VDI acceleration algorithm that guarantees intelligent

    usage of hardware resources to deliver unrivaled performance

    SnapVDI® Zero Client, the end user device with proprietary SnapVDI® VUE firmware for

    enhanced user experience

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 6

    Figure 1 : SnapVDI® Components

    2.1.1. SnapVDI® Manager (SVM)

    SnapVDI® Manager is the centralized management software that aids in managing the entire VDI

    solution. Using this software, an administrator can monitor, manage, and provision all VDI

    components - including VM pool creation, deletion, modification, provisioning and management

    of users, storage, zero clients and inventory - all from a single local or remote location. In

    addition, SnapVDI® Manager enables the administrator to build a private cloud environment to

    greatly increase the utility and cost-effectiveness of the VDI solution.

    2.1.2. SnapVDI® Spark

    The brain of the SnapVDI® architecture, SnapVDI® Spark™ provides intelligent caching and

    storage with unprecedented high performance. Spark™ also enables a seamless user experience

    with on-the-fly backup and recovery built in. This layer of SnapVDI® empowers Remote Desktop

    Services (RDS), giving users an experience identical to that of a local desktop. SnapVDI® Spark™

    is designed for scalability and performance with minimal downtime. In fact, all Virtual Machines

    (VMs) in the entire VDI environment can be booted up in a matter of seconds.

    2.1.3. SnapVDI® Zero Client

    SnapVDI® Zero Client is the physical device given to each user to access the Virtual Desktop

    Infrastructure. All zero clients, as well as all users, are simply managed through a single point by

    SnapVDI® Manager. The SnapVDI Zero Client is a high performance System-On-Chip (SOC) device

    that supports Microsoft RDS-based VDI solutions and uses only 7 watts of power at the

    maximum. The client can also be integrated on devices to make a portable workstation.

    SnapVDI® Zero Client also supports HD graphics (maximum resolution of 1920x1080) for

    workstations and graphics oriented applications.

    SnapVDI® Zero Client does not require any OS to be installed by the end user as it comes with an

    embedded OS pre-installed.

    Zero Client manageability, including access to desktops, user privileges, and peripheral device

    control, are managed by SVM through user policies. If any unauthorized access to the device is

    attempted, the device will be locked by SVM and it can be unlocked only by the administrator

    through SVM.

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 7

    2.2. SnapVDI® Reference Architecture

    2.2.1. Overview

    This paper is intended to help customers, IT architects, consultants, and administrators involved

    in the early phases of planning, designing, or deploying SnapVDI®-based VDI solutions. The

    purpose is to provide a standard, repeatable, and highly scalable design that can be easily

    adapted to specific environments and customer requirements.

    Some key features that can help an organization get started quickly with a solution that

    integrates easily into existing IT processes and procedures include:

    Standardized, validated, readily available components

    Scalable designs that allow room for future growth

    Validated and tested designs that reduce implementation and operational risks

    Quick implementation, reduced costs, and minimized risk

    2.2.2. Architecture Components

    This reference architecture uses common components to minimize support costs and

    deployment risks.

    The desktop virtualization solution referenced in this whitepaper, the SnapVDI B150 brick,

    combines the best of breed of data center, virtualization, and network technologies. The B150

    employs the below tabulated hardware and software components:

    Component Quantity

    Host Hardware 2U SnapVDI® Server (HA) 2

    2U SnapVDI® Dual Controller Storage Console 1

    Intel® Xeon® Processor E5-2630 v4 2

    32GB DIMM RAM 48

    SAS HBA 2

    1TB SAS HDD 16

    Network 20 x1Gbps Ports, 4 x10Gbps Ports -

    Software Microsoft® Windows Server 2012 R2 Data Center 2 Processors 2

    Microsoft® WINENT Software Assurance Pack OLV License 1 YR 1

    Microsoft® Windows Remote Desktop Service Cal 1 YR Per User

    Microsoft® VDA Single Sub 1 Month Per Device (12 Months) Per User

    SnapVDI® Manager 1

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 8

    SnapVDI® Spark 1

    SnapVDI® User License Per User

    Table 1 : Architecture components

    2.2.3. Infrastructure

    The below reference architectural block diagram represents the logical infrastructure SnapVDI®

    solution with the hardware and software components in place.

    Figure 2 : Reference Architecture

    2.2.4. Network Design Considerations

    The figure below shows the network layout of the two host 2U SnapVDI servers. Out of the two

    10 Gbps network ports, one is connected to the 10 Gbps port of the network switch and the

    other 10 Gbps port is connected in crossover between the two servers assisting in monitoring

    heartbeat for the High Availability (HA).

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 9

    Figure 3 : Network Consideration

    3. Test Results Summary The table below provides the summary of the test results based on testing methods that were used

    to demonstrate the performance of SnapVDI® solution (on high-availability). A more in depth

    analysis is provided in the testing sections to explain how these results were achieved.

    IOMeter Test Summary

    VDI load (4KiB 20% Read and 80% Write with 80% random distribution) test for storage IOPS

    2,612 IOPS

    4KiB 50% Read 50% Write with 100% Sequential distribution test for storage IOPS

    12,514 IOPS

    Login VSI Test Summary

    Test for Personal and Pooled Desktop collection (30 – 70%)

    120 user sessions ran successfully VSImax office worker 4.1 not reached with VSIbase of 1621 ms

    Test for Pooled Desktop 150 user sessions ran successfully VSImax office worker 4.1 not reached with VSIbase of 1576 ms

    High Availability Test Summary

    Failover Test All virtual desktops in both simulations were available and running in approx. 15 minutes

    Table 2: Test Results Summary

    Speed 10 Gbps Speed 10 Gbps for SnapVDI® Heartbeat Speed 1 Gbps

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 10

    4. Test Methodology This section outlines the test objectives along with the test tools and criteria used to determine the

    maximum number of desktops that can be supported in the environment.

    4.1. Test Objectives The test objectives are:

    Determine how many virtual desktops can be deployed in this environment using SnapVDI

    servers with acceptable user experience indicators for an office user workload profile.

    Determine the performance impact on the Spark disk during peak I/O activity such as boot and

    login storms.

    Determine the time taken for the virtual desktops to migrate from one server to another server

    during failover.

    4.2. Test Tool The tests were conducted using Login VSI 4.1 as the workload generator and user experience

    analyzer tool and Iometer as both workload generator and the measurement tool along with

    Windows built-in Performance monitor for IOPS.

    Login VSI is the industry standard load testing solution for centralized virtualized desktop

    environments. (More information can be found at www.loginvsi.com.)

    Iometer is an I/O subsystem measurement and characterization tool for single and clustered

    systems. (More information can be found at http://www.iometer.org/)

    4.2.1. Load Generation

    The “Office worker” workload from Login VSI was used to simulate the user workload. Although

    Login VSI provides other workloads, the office worker workload is selected as it focuses on

    broad Office and Internet Explorer usage and is designed to run within 1vCPU virtual desktops

    which closely resemble the test setup.

    VDI workload (4KiB 20% Read and 80% Write with 80% random distribution) and 4KiB 50% Read

    and 50% Write with 100% sequential distribution workload is generated in Iometer to get the

    IOPS from the disks.

    4.3. Test Criteria The primary focus of the test was to determine the maximum number of desktops that can be

    supported on SnapVDI servers in high-availability mode while using SnapVDI® Spark and SnapVDI®

    Manager. Virtual desktop user experience was also monitored.

    4.4. Test Configuration A SnapVDI B150 brick was used in both test scenarios. The desktop (pooled and personal) collections

    were configured using the SnapVDI® Manager interface. Each pool was built from a Windows 8.1

    base image. The following section explains the desktop pool configuration used for testing.

    http://www.loginvsi.com/http://www.iometer.org/

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 11

    Desktop pool configurations

    - The first configuration was comprised of four Pooled collections with 80 desktops and two

    Personal collections with 40 Desktops for 120 combined desktops. The second configuration

    was comprised of eight Pooled desktop collections for 152 pooled virtual desktops.

    - The disk(s) that contained the collection was de-duplicated for storage efficiency.

    Six launch systems were used for the mixed Pooled and Personal test, while eight launcher

    systems were used in the Pooled only test in Login VSI.

    Setup Pooled – 150VMs Pooled/Personal – 120 VMs

    Storage Pool 2 Pools with 4 Physical Disks mirrored

    Virtual Disk Resiliency

    Mirror, 2 Columns, 256 KB Interleave 2 Virtual Disks for three Sparkdisk

    Sparkdisk Configuration

    Polling Intervals : 0.5ms (write, stream, delete) Processing Window Size : 5000 (write, stream) 1000 (delete)

    Collections Pooled – 6 collections of 25 VMs each Pooled/Personal – 4 collections of 20 VMs (Pooled), 2 collections of 20 VMs (Personal)

    LoginVSI Pooled – Office worker workload, 2880 seconds as launch window with 8 launchers launching 19 VMs each Pooled/Personal - Office worker workload, 2880 seconds as launch window with 6 launchers launching 20 VMs each

    SVM VM Running -- SVM 2027 (RC version)

    Operating System for VM

    Windows 8.1 running Windows Office 2010 and other Login VSI test suite apps.

    VM Details 2 vCPU per user 2GB Dynamic vRAM per user

    Server Details HA Server Setup running Server 2012 R2 OS Server configuration:

    - 10 Core CPU - 768 GB RAM

    5. Test Results and Analysis This section presents the results from the different SnapVDI® characterization tests and the key findings

    from each test. The office worker workload represents a majority of the VDI users in the industry today,

    and the testing was focused on this workload profile.

    5.1. Test Scenarios The following tests were conducted to gather results and analyze the solution:

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 12

    5.1.1. Storage IOPS

    The typical industry standard IOPS recommended for an office worker is 20 IOPS/user.

    Maintaining this limit ensures good user application response times when there are no other

    bottlenecks at the infrastructure layer.

    5.1.2. Personal and Pooled Desktop Solution

    In these tests, the solution is configured with 40 personal desktop collections and with 80

    pooled desktop collections and the below test cases are carried out. The intent of this test was

    to show that the solution can offer the standard VDI solution configuration of 70% pooled and

    30% personal desktop collection. The testing was performed on a SnapVDI B150 brick (two

    SnapVDI Servers and a SnapVDI Storage console).

    5.1.2.1. Login VSI Office Worker Workload User Test

    The Login VSI office worker workload was run for at least for one hour with the standard

    recommended launch window setting of 2880 seconds to simulate the real world scenario of

    users performing their daily tasks. The VSImax (Dynamic) parameter from Login VSI is used

    to evaluate the user experience of a simulated user working on the virtual desktop

    throughout this test.

    5.1.2.2. Boot Storm I/O

    Boot storms represent the worst-case scenario where many virtual desktops are powered

    on at the same time and they all contend for the system resources simultaneously. This test

    was used to evaluate if the storage array hosting the desktops was capable of handling large

    spikes in storage I/O without causing significant impact on user experience and other

    services.

    5.1.2.3. Login Storm I/O

    Login storms also represent a heavy I/O scenario where many users are logging into their

    virtual desktops at the beginning of a work day or a shift at the same time. In this test, all of

    the desktops were pre-booted and left in an idle state for more than 20 minutes to let their

    I/O settle prior to running the Login VSI Office worker workload to simulate users logging

    into their virtual desktops.

    5.1.2.4. Steady State I/O Workload for Office Users

    The test is intended to study the steady state of I/O that is required for the desktops to

    perform optimally during the Login VSI office worker workload.

    5.1.3. Pooled Desktop Solution

    In these tests, the solution is configured with 150 pooled/non-persistent desktop collections

    running on a SnapVDI B150 brick (two SnapVDI Servers and a SnapVDI Storage console). The

    boot storm, login storm, and steady state tests were repeated for this configuration.

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 13

    5.1.4. Migration of Virtual Desktops During Failover

    Highly available systems build redundancy into the application and architecture layers to protect

    against disasters. Maintaining reliable, recoverable, and continuous operable resources with

    minimal downtime ensures a highly redundant system.

    5.2. Test Results

    5.2.1. Storage IOPS

    To study on the storage capacity and capability of the Spark disks, Iometer is made to generate

    different workloads and made to run for 20 minutes. Below are the findings from the Iometer

    test:

    Load Total IOPS

    4KiB 50% Read 50% Write with 100% Sequential distribution 12,514

    4KiB 50% Read 50% Write with 100% Random distribution 3,414

    4KiB 50% Read 50% Write with 50% Sequential and 50% Random distribution 3,948

    VDI Load (4KiB 20% Read and 80% Write with 80% random distribution) 2,612

    Table 3 : Test results for Storage IOPS

    5.2.2. Pooled and Personal Desktop Solution

    5.2.2.1. Login VSI Testing

    To test the SnapVDI® solution in a real-world scenario, Login VSI was configured with office

    user workloads and was made to launch and test 120 VMs for about an hour with launch

    window of 2880 seconds across six launchers. Below is the Login VSI user experience graph

    and it can be observed that VSI v4 max was not reached – indicating it’s likely that more

    desktops are able to run on this system.

    (Login VSI is the industry standard load testing solution for centralized virtualized environments.

    Enterprise IT departments use Login VSI to test the performance and scalability of VMware Horizon

    View, Citrix XenDesktop and XenApp, Microsoft Remote Desktop Services (Terminal Services) or any

    other Windows based virtual desktop solution. Login VSI benefits IT organizations in all phases of their

    deployment—from capacity planning, to load testing, to change impact prediction--for more

    predicable performance, higher availability and a more consistent end user experience.

    For more information, download a trial at www.loginvsi.com.)

    http://www.loginvsi.com/

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 14

    Figure 4 : Login VSI Test result for Pooled and Personal desktop solution

    5.2.2.2. Boot Storm I/O

    To simulate a boot storm, the 120 virtual desktops were started simultaneously from the

    management interface. The below figure shows the storage characteristics during the boot

    storm – the Spark disk delivered 4,000 IOPS under the peak load during this test at an

    average of 18-20 IOPS per VM. All 120 of the virtual desktops were booted up and were

    available in about 148 seconds, approximately 1.2 seconds per VM.

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 15

    (The mark on the graph represents the time when all 120 VMs booted up.)

    Figure 5 : Boot Storm IOPS for Pooled and Personal desktop Solution

    5.2.2.3. Login Storm I/O

    Login VSI was programmed to launch 120 virtual desktops over a period of about 5 minutes

    after pre-booting the virtual desktops. The peak IOPS observed during the login storm was

    about 3,350 IOPS at an average of 12-15 IOPS per VM.

    0

    500

    1000

    1500

    2000

    2500

    3000

    3500

    4000

    4500IO

    PS

    Duration (secs)

    Boot Storm IOPS

    Total IOPS Read IOPS Write IOPS

    0

    500

    1000

    1500

    2000

    2500

    3000

    3500

    4000

    IOP

    S

    Duration (secs)

    Login Storm IOPS

    Total IOPS Read IOPS Write IOPS

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 16

    Figure 6 : Login Storm IOPS for Pooled and Personal desktop Solution

    5.2.2.4. Steady State I/O

    The average IOPS on the Spark disks during the steady state with all 120 users logged in was

    around 1,100 (8-10 IOPS per VM). Of this, the read IOPS accounted for about 330

    (approximately 30 % of the total I/O load) and the remaining 770 IOPS or 70% were write

    IOPS.

    Figure 7 : Steady State IOPS for Pooled and Personal desktop Solution

    5.2.3. Pooled Desktop Solution

    5.2.3.1. Login VSI Testing

    Login VSI testing was carried out with office user workloads and was made to launch and

    test 150 VMs for about an hour with launch window of 2880 seconds across eight launchers.

    Below is a chart listing the hardware and software components utilized in the test, as well as

    a Login VSI user experience graph. It can be observed from the graph that VSI v4 max was

    not reached; therefore it’s likely that more desktops are able to run on this system as well.

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 17

    Setup 150VMs

    Storage Pool 2 Pools with 4 Physical Disks mirrored

    Virtual Disk Resiliency

    Mirror, 2 Columns, 256 KB Interleave 2 Virtual Disks for three Sparkdisk Pool1 -> vDisk1 Pool2 -> vDisk2 Pool2 -> vDisk1

    Sparkdisk 3 disks of 140 GB with write-through and mirroring enabled for Pooled Desktop Collection

    Sparkdisk Configuration

    Threshold : 495K-500K Polling Intervals : 0.5ms (write, stream, delete) Processing Window Size : 5000 (write, stream) 1000 (delete)

    Collections Pooled-6 collections of 25VMs each(2 on each RAM Disk)

    Write-through mode

    Sparkdisk 1 -> vDisk1 Sparkdisk 2 -> vDisk2 Sparkdisk 3 -> vDisk1

    LoginVSI Office worker (1vCPU), 2880 seconds as launch window with 8 launchers launching 19 VMs each

    SVM VM Running -- SVM 2027 (RC version)

    Operating System for VM

    Windows 8.1 running Windows Office 2010 and other Login VSI test suite apps.

    VM Details Pooled – 2 vCPU and 2GB Dynamic vRAM per user Personal – 2 vCPU and 4GB Dynamic vRAM per user

    Server Details HA Server Setup running Server 2012 R2 OS Server configuration:

    - 10 Core CPU per server (2x) - 768 GB RAM per server (2x)

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 18

    Figure 8 : Login VSI Test result for Pooled desktop solution

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 19

    5.2.3.2. Steady State I/O

    The total IOPS on the Spark disks averaged during the steady state with all the 150 users

    logged in was approx. 2,100 (14-15 IOPS per VM). Of this, the read IOPS accounted for about

    630 (approximately 30 % of the total I/O load) and the remaining 1,470 IOPS or 70% were

    write IOPS.

    Figure 9 : Steady State IOPS for Pooled desktop Solution

    5.2.4. High Availability (HA)

    The high availability of the solution is determined by simulating failover. The active server was

    shut down properly, keeping all 120 virtual desktops actively running in the mixed user test, as

    well as all 150 users in the pooled user test. After the changing of nodes (i.e. the passive node

    becoming active), all virtual desktops were available up and running in approximately 15

    minutes in both simulations.

    6. Configuration

    6.1. Server Configuration The 2U SnapVDI® Servers are configured as below:

    The server was equipped with two Intel® Xeon® processor E5-2630 v4, 10 cores @ 2.2 GHz,

    48x32GB DIMM RAM, two SAS HDDs and one LSI SAS HBA controller.

    The SnapVDI® Storage console is populated with 12 SAS HDD and is connected to the server

    via the LSI SAS HBA controller.

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 20

    The HDDs for the server is RAID1 configured for resiliency and high availability with

    Microsoft Windows server 2012 R2 64 bit OS is installed on it.

    Remote Desktop Services, Hyper-V and Cluster Manager Roles and Features are installed

    alongside SnapVDI® Spark for hosting virtual desktops.

    The servers are clustered and the RDS is configured in high availability mode with SnapVDI®

    Spark disk mirrored making the solution resilient to disaster.

    The media storage in the SnapVDI® Storage Console is advertised storage spaces for the

    clustered servers and is primarily used for backing up the contents of the SnapVDI® Spark

    disk.

    6.2. SnapVDI® Configuration The SnapVDI® Manager is installed on the centralized desktop and is configured to manage the

    entire solution from a single point. The Create Desktop collection in the SnapVDI® Manager was

    used to create the virtual desktop collection.

    6.3. Windows 8.1 Virtual Desktop Configuration Following the guideline from Login VSI, the Windows 8.1 base image was generated based on

    the below generic template with the following properties:

    SnapVDI® Host Agent 2.0

    Two virtual CPU

    Dynamic RAM, 4GB for Personal virtual desktop and 2 GB for Pooled virtual desktop

    30 GB virtual hard drive

    One virtual NIC connected to the VDI network

    Windows 8.1 64 bit OS with the latest guest additions installed

    6.4. Login VSI Test Configuration Login VSI is an industry-standard solution that simulates typical user behavior in centralized

    virtualized desktop environments. When used for benchmarking, the software measures the total

    response time of several specific user operations being performed within a desktop workload in a

    scripted loop. The baseline is the measurement of the response time of specific operations

    performed in the desktop workload, which is measured in milliseconds (ms).

    There are two values in particular that are important to note: VSIbase and VSImax.

    1. VSIbase

    A score reflecting the response time of specific operations performed in the desktop workload

    when there is little or no stress on the system. A low baseline indicates a better user experience,

    resulting in applications responding faster in the environment

    2. VSImax

    The maximum number of desktop sessions attainable on the host before experiencing

    degradation in host and desktop performance

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 21

    Based on the Login VSI guideline, this reference architecture uses the below logical infrastructure for

    using Login VSI and testing the desktops using Login VSI.

    Figure 10 : Login VSI test configuration

    7. Conclusion The SnapVDI® solution serves as an entry point for SMBs to migrate from the conventional desktop

    architecture to a more affordable and manageable virtual desktop infrastructure.

    The results in the document demonstrate how a modular office worker virtual desktop environment can

    be deployed using SnapVDI® Manager. Extensive workload, operations, and resiliency testing shows that

    the SnapVDI® solution delivers high levels of performance, a great end-user experience, and solid

    system resiliency, all at a low price point with failover built-in. Organizations can leverage this modular

    VDI solution to start small and then grow as needed with zero disruption.

  • Copyright © 2016 | American Megatrends Inc. | Technical White Paper 22

    8. References 1. Login VSI documentation - http://www.loginvsi.com/documentation/Login_VSI

    2. IOMETER - http://www.iometer.org/

    3. Performance Data collection and IOPS collection through Microsoft counters -

    http://blogs.technet.com/b/askpfeplat/archive/2013/09/23/using-powershell-to-gather-

    performance-data.aspx and https://technet.microsoft.com/en-us/library/hh849685.aspx

    4. Super Micro Servers datasheet and reference -

    http://www.supermicro.com/products/chassis/2U/829/SC829BTQ-R920W.cfm

    5. VDI workload - https://community.atlantiscomputing.com/blog/Atlantis/August-2013/How-to-use-

    Iometer-to-Simulate-a-Desktop-Workload.aspx

    Login VSI, Inc. delivers industry-standard testing solutions for virtualized desktop and server environments. The

    world’s leading virtualization vendors use the flagship product, Login VSI, to benchmark the performance and

    scalability of their solutions. Enterprise IT departments use Login VSI in all phases of their virtual desktop

    deployment—from capacity planning, to load testing, to change impact prediction—for more predictable

    performance, higher availability and a more consistent end user experience. With minimal configuration, Login VSI

    works in VMware Horizon View, Citrix XenDesktop and XenApp, Microsoft Remote Desktop Services (Terminal

    Services) and any other Windows-based virtual desktop solution.

    For more information, download a trial at www.loginvsi.com.

    Login VSI bears no responsibility for this publication in any way and cannot be held liable for any damages following from or related to any information in this publication or any conclusions that may be drawn from it.

    http://www.loginvsi.com/documentation/Login_VSIhttp://www.iometer.org/http://blogs.technet.com/b/askpfeplat/archive/2013/09/23/using-powershell-to-gather-performance-data.aspxhttp://blogs.technet.com/b/askpfeplat/archive/2013/09/23/using-powershell-to-gather-performance-data.aspxhttps://technet.microsoft.com/en-us/library/hh849685.aspxhttp://www.supermicro.com/products/chassis/2U/829/SC829BTQ-R920W.cfmhttps://community.atlantiscomputing.com/blog/Atlantis/August-2013/How-to-use-Iometer-to-Simulate-a-Desktop-Workload.aspxhttps://community.atlantiscomputing.com/blog/Atlantis/August-2013/How-to-use-Iometer-to-Simulate-a-Desktop-Workload.aspxhttp://www.loginvsi.com/