HP and Citrix VDI-in-a-Box: Part 1 - Reference Architecture Scaling

HP and Citrix VDI-in-a-Box: Part 1 - Reference Architecture Scaling -

Coupled with HP ProLiant servers, Citrix VDI-in-a-Box creates a low cost cookie-cutter approach to rapidly deploy VDI environments. Together, these products form a key-like device in hand which makes the deployment of VDI remarkably easy -. The challenge is simply to determine the number and type of HP ProLiant servers needed to match the ability of users and the demands of the workload

To take the guesswork out of capacity planning and design, I work with the client HP Virtualization Architect Kirk Manzer to validate specific HP server and Citrix VDI-in-a-Box configurations. We wanted to find scalability guidelines to help customers predict capacity and identify the number of users that can be supported with acceptable response times under a defined workload. The white paper, "HP Client Virtualization SMB Reference Architecture for Citrix VDI-in-a-Box" contains details of the testing and data that I collected.

Test Environment Overview

The test environment included these basic components:

• Citrix VDI-in-a-Box Citrix VDI-in-a-Box is a virtual appliance that runs on a hypervisor. -enabled server to provide centrally managed virtual desktops. It is a complete solution designed to VDI scale by simply adding additional units and features integrated connection brokering and load balancing across multiple servers. An N + 1 configuration provides a highly available solution - if a server fails, Citrix VDI-in-a-Box switches to the remaining server to host virtual desktops. Citrix VDI-in-a-Box takes advantage of Citrix Receiver and HDX protocol to support a variety of types of devices, including smartphones, tablets, thin clients, laptops, and workstations.

• HP ProLiant Gen8 Server DL380p. In testing, I used two different configurations of HP ProLiant servers, from design to the 50 capacity and support 100 users. The test validated server configurations with these capabilities. The local storage configuration (based on HP Smart Array PCIe 6 Gb / SAS 3.0 controller and eight 10,000 RPM SAS drives configured as RAID 0 + 1 volume) was the same for both servers; Only the configurations of compute and memory differed. To support 50 users, I configured a server with one Intel Xeon E5-2680 CPU @ 2.70GHz (8 cores, 16 threads) and 96 GB of memory, while the 100-user configuration has twice the CPU resources - Intel Xeon E5 CPU two -2680 - and 128 GB of memory. (For a production environment 100 users, Citrix recommends 192GB RAM to ensure sufficient memory resources.)

• Microsoft Server 08 R2 SP1 Hyper-V . VDI-in-a-Box runs on Microsoft Hyper-V, Citrix XenServer and VMware vSphere hypervisors. I chose Microsoft Hyper-V for testing.

The test harness I used included Login VSI ( www.loginvsi.com ), a load generation tool for benchmarking VDI that simulates user workload production. For this test, I chose the default average workload to simulate desktop activity of a typical knowledge worker. Login VSI generates an office productivity workload that includes Office 2010 with Microsoft Outlook, Word, PowerPoint and Excel, Internet Explorer with Flash video applet Java application and Adobe Acrobat Reader.

Configuration Shades

to other performance tests I've done, it is clear that maintaining a balance calculation, memory and I / O resources is essential to scale the number of effectively workstations. In these trials, having twice the compute and memory resources to handle twice the capacity of the user seemed to be a reasonable starting point for acceptable performance. In the design of production configurations, the type of application workload is a significant impact, so be sure to use the results here as a general guideline rather than a

in the strict test configurations rule of thumb. Citrix VDI-in-a-Box takes cheap local storage advantage rather than shared SAN or NAS complex configurations. A variable in the design of a production environment is how much storage is needed, which depends on the number and size of desktop images, set the swap space, and additional space for the growth. VDI-in-a-Box supports a one-to-many imaging approach that conserves disk space - workstations share a gold master image and the software follows the individual differences of this master. You can estimate the disk space requirements depending on the number and size of your golden images and the number of jobs that will be created from each.

Although scalability (rather than availability) was the center of my tests, it is important to emphasize that the integral features of the solution allow for highly available desktop services. In a deployment with N + 1 servers, Citrix VDI-in-a-Box uses an architecture without sharing so that all servers are also replaceable and function as peers to each other. As long as you pad the deployment of an additional server capacity, if a server fails, another server can continue to provide virtual desktops for your environment.

Test Methodology

For each test, I follow this sequence of steps:

1 ) Using the management console VDI-in-a-Box, I verified that all positions were fueled and in a state of rest "Hold".

2) I started the VSI launchers and verified that they were ready for the test.

3) I started a script that invoked PerfMon scripts to capture complete system of performance indicators.

4) with workstations on and left idle, I initiated the portion of the test workload simulation using Login VSI. According to the test, 50 or 100 desktop sessions were launched and Login VSI simulated user connections on each.

Once all users have logged on, the part to the stable state of the trial began in which the Login VSI followed statistics applying perfomance, loop through specific operations and measure response time at regular intervals. Response times are used to determine Login VSIMax, the maximum number of users that the test environment can support before performance is constantly deteriorating. In the test, I define success criteria that the response time of applications lower than 4000ms (based) with exceptions that occur less than six times. In these series of initial tests of 50 and 100 users, VSIMax scores were not achieved.

5) Although Login VSI loop through its workload, I followed function test progress using the VDI-in-a-Box Manager and consoles Login VSI. (Office report launched in assets should not fall behind more than 2 or 3 workstations for single-server test.) After a specified amount of time elapsed from the equilibrium state, Login VSI began to disconnect desktop sessions.

6) After all the sessions were disconnected, I stopped performance monitoring scripts.

7) Finally, I treat Login VSI logs using VSI Analyzer and PerfMon CSV using PAL to analyze the test results .

test results

VDI configurations I tested showed linear scalabilityfor 50 or 100 users in an average workload Login VSI. Given the calculation to scale and memory resources in the HP ProLiant server, I could linear scale from 50 to 100 users, while maintaining acceptable response times without depleting subsystem resources. Detailed results for the test series are given below.

Results 50-user

The following graphs show VSIMax Connection, CPU usage, memory consumption, network traffic (incoming and outgoing), and calculated read and write IOPS , respectively, for this test configuration. The graphs show that the only Xeon E5-2680 CPU with 8 physical cores / 16 threads and 96 GB of memory provided sufficient resources to maintain acceptable response times throughout the 50-user test.

The test configuration

50 users follow these general parameters:

· physical basic office report: 6.25 to 1

· office computer memory: 1GB per office was tested, VSI average workload consumes ~ 800MB per workstation peak period (minimally 1.5GB per desktop is recommended with an additional allowance for overhead for the operating system hypervisor)

· storage read / write ratio

o average 47/53 read / write

o Max: 58/42 read / write

· IOPS Storage

o By average office: ~ 10.04

o by Max office: ~ 39.18

It is important to note that the above data storage reflects all test phases (logons, steady state and disconnections). While you can use these results as general guidelines for sizing, remember that you have to adapt all configurations to match specific types of workloads and user capabilities.

Login VSI Data (no VSIMax)

The graph below shows that the configuration setting in easily scale to support 50 users while maintaining acceptable response times. The response time never exceeded 4000 ms more than six times, so the 50-user configuration never reached Login VSIMax.

Log VSIMax, the maximum capacity of the system tested, expressed in number of sessions Login VSI, is derived from the analysis tool Login VSI. Within each workload test loop, the response time of seven specific operations are assessed at regular intervals, six times in each loop. The times of these seven response operations are used to establish VSIMax.

CPU (16 single logical cores)

This metric use of physical processors in the host computer records. "Logical Processor Hyper-V Hypervisor (*) % Total Run Time" performance counter is more accurate than using "% Processor Time" counter of the system because the "% Processor Time" against only measures the host processor time. "logical processor Hyper-V Hypervisor (*) % total Run time" performance counter is the best counter to be used to analyze the use of the overall processor Hyper-V server.

memory (96GB total)

MBytes available is the amount of memory in megabytes, immediately available for allocation to a process or for system use. it is equal to the sum of memory assigned to the standby (cached), free and zero page lists. If this counter is low, then the computer is low on RAM physics.

IOPS Storage

IOPS calculated statistic is the rate combined reading and writing on C :. Disc

Network

total network interface bytes / sec is the combined rate at which bytes are sent and received on each card network, including framing characters.

Results 100 users

The following charts show VSIMax connection, use of the processor, memory usage, network traffic (incoming and outgoing), calculated read and write IOPS, respectively, for the 100- user configuration. The graphs show that the HP ProLiant server with two Xeon E5 2680 processors (16 physical cores / 32 son in total) and 128 GB of memory provided sufficient resources to maintain acceptable response times throughout the test.

In rule in design, testing 100 generated by users of these parameters:

· office to cores report 6.25 to 1

· Memory the workplace: 1GB per office was tested, VSI load average work consumes ~ 800MB per desktop during the peak (minimally 1.5GB per workstation is recommended with an additional allowance for overhead for the operating system hypervisor)

· Storage reading / writing report

o average 42/58 read / write

o Max: 55/45 read / write

· IOPS Storage

o By average office: ~ 9.2

o By Max office: ~ 23.54

It is important to note that the above data storage reflects all test phases (logons, the state of balance and disconnections). Remember that these results are general guidelines and configurations must match specific types of workloads and user capabilities.

VSI data connection (not VSIMax)

The graph below shows the configuration scaling easily to support 100 users while maintaining acceptable response times in an average workload Login VSI. Again, the response time does not exceed 4000 ms more than six times, so that the user 100 configuration has not reached Log VSIMax, the maximum capacity expressed in number of sessions Login VSI. To establish VSIMax, the Log Analysis Tool VSI saves time seven response operations within each workload test loop.

CPU (Dual, 32 logical cores)

This metric records overall CPU usage Hyper-V server.

memory (128GB total)

This measure indicates the available amount of memory in megabytes.

IOPS Storage

Calculated IOPS shows the rate of reading and writing.

Network

This measure is the rate at which bytes are sent and received on each network adapter including framing characters.

Summary

While you should take into account all relevant factors configuration - type of user workload, availability requirements, etc. - When designing a production deployment, my tests validated reference architecture 50- and 100-user configurations. Results showed linear scalability of the HP ProLiant and Citrix VDI-in-a-Box solution under average workload Login VSI. As CPU resources and memory scaled, user population changes linearly without any significant degradation in response time. For more details on the tests and results, see the reference architecture report.

Log VSIMax Since (the point where performance begins to degrade constantly) has never been reached or within 50 or 100 user tests run, I wondered how the performance would begin to deteriorate . To better understand the capacity limitations, I fell a further series of tests that you will be able to read in parts 2 and 3.

References

• Virtualization HP SMB Client Reference architecture for Citrix VDI-in-a-Box (PDF)

• HP and Citrix VDI-in-a-Box: Part 2 - Pushing the limits (Blog)

• HP and Citrix VDI-in-a-Box: Part 3 - user Experience testing (Blog)

Tweet