In my previous blog, I am describing the scalability tests conducted to confirm a Citrix VDI in-a -Box and HP ProLiant reference architecture. The reference architecture defines a turnkey appliance as it simplifies VDI deployment task. With this reference architecture as the basis for the design, it is relatively easy to determine the number and type of devices you'll need to adapt your ability to use and requirements of the workload.
In the previous essay, I validated configurations and 50- to 100 users using scripts monitoring performance and VSI connection, VDI testing tool that captures the response times while simulating user workload. Under an average workload Login VSI (representing a typical knowledge worker), my results showed linear scalability without any significant degradation of response times.
Since these 50- and 100-user runs tests failed connection VSIMax (the maximum use of capacity at which the performance begins to degrade), I decided to repeat the test series with greater use of capacities to push the limits of the reference architecture configurations. This blog describes my second series of tests using the same test configurations and users 70 and 0 users capabilities. I also ran two additional stress tests to examine the performance impacts the number of users to scale. Overall, my second set of tests included four tests:
- 70 configuration test user. Since VSIMax connection has not been reached in the original 50-user test, I wanted to see the point where the degraded performance consistently for the 50-user configuration.
- configuration testing 0-user. Using the 100 user configuration (with twice the resources of the processor and memory configuration of 50 users), I repeat 100-user test with 0 users to see when Connection VSIMax happen.
- office restart test. In this test, 75-user, I have assessed the impact of four restart workstations along the way during a test.
- impact test Storage. I booted 50 workstations in a period of 10 minutes (simulating a boot storm) to examine the impact on the storage subsystem.
Test Results
In general, these tests used the same test environment and methodology that the series of initial tests and 50- 100 user (see my first blog on scalability testing for details). Besides Login VSI, I used a set of scripts to capture performance metrics during each test. This way, I could see the impact of the tests on system resources (CPU, memory, storage and network I / O). Detailed results for the four test tracks follow.
70 Results of the user (61 VSIMax)
The following graphs show VSIMax Connection, CPU usage, memory consumption, calculated in read and write IOPS, and traffic network (sent and received) to the test 70 user configuration. With 70 users in total, the 50-user configuration reached VSIMax to 61 users, as shown in the graph below. The server configuration is an HP ProLiant Gen8 server DL380p with a single Intel Xeon E5-2680 CPU @ 2.70GHz (8 cores, 16 threads) and 128 GB of RAM (instead of 96 GB, which I used in the test initial 50-user).
Login VSI data (61 VSIMax 70 sessions started)
To establish connection VSIMax, the script workload Login VSI saves time seven operations response within each workload test loop. The analysis tool uses one of two methods of calculating connection VSIMax, classic or dynamic (see http://www.loginvsi.com/analyzing-results). In this test, the dynamic method (in which the average response time must always exceed a threshold dynamically calculated) has been used since this approach is more appropriate to analyze virtualization applications. The formula for determining the threshold is explained on the website Login VSI. VSIMax connection indicates Login VSI active users connected to the system when the consistent response times exceed this threshold.
CPU (Single, 16 logical cores)
This metric use of physical processors in the host computer records. The "logical processor Hyper-V (*) % Total Run Time" performance counter indicates the use of the overall processor Hyper-V server. This is more accurate than using "% processor time" counter system that as time host processor.
The processor of the graph shows that the only Xeon E5-2680 CPU sustained 100% utilization for a period of more than 4 minutes, indicating that CPU resources become exhausted when the support of more than 50 users in an average workload Login VSI.
memory (128GB total)
This measure is the amount of RAM 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.
IOPS Storage
Calculated IOPS represents the combined rate of literacy on C :. Disc
network
Total / sec Bytes network interface is the combined rate of bytes sent and received over each network adapter including framing characters.
Results 0 users (113 VSIMax)
Using the 100 user configuration, I conducted a trial with 0 users to determine when connection is VSIMax happen. The following graphs show VSIMax Connection, CPU usage, memory consumption, calculated in read and write IOPS, and network traffic (sent and received) to test 0-user configuration.
Login VSI data (113 VSIMax, launched 0 sessions)
As shown in the graph below, the 100 users reached VSIMax configuration at 113 users.
CPU (Dual, 32 logical cores)
chart shows CPU usage for the Hyper-V server. It indicates that server, hosted on two Xeon E5-2680 processors, supported nearly 100% utilization for more than 10 minutes, indicating exhaustion of CPU resources with 0 users in an average workload.
memory (128GB total)
This measure shows the amount of physical RAM in megabytes.
IOPS Storage
Calculated IOPS represents the combined rate of literacy on C :. Disc
network
This measure is the combined rate at which bytes are sent and received on each network adapter, including framing characters.
Restart office test
using an HP ProLiant server configuration with 2 processors and 128GB of RAM, I conducted a stress test of 75 users to study the impact of restarting four work stations along the way. As in other tests, user workstations were launched and Login VSI simulated user workloads. At the halfway (when the workload for the user 37 started), I restarted four desktop sessions. Restarting the four workstations had no significant impact on response time and VSIMax was not reached. performance monitoring scripts indicated that sub-system resources, including CPU cycles on the server, were always readily available.
The following graphs show data Login VSI, CPU usage, memory consumption, calculated read and write IOPS, and network traffic (sent and received), respectively, for the office test restart.
VSI data connection (not VSIMax)
As shown in the graph below, Connection VSIMax has not been reached during this test.
CPU (Dual, 32 logical cores)
The chart shows CPU usage for the Hyper-V server.
memory (128GB total)
This measure indicates the amount of physical RAM in megabytes.
IOPS Storage
Calculated IOPS represents the combined rate of literacy on C :. Disc
network
This measure is the combined rate at which bytes are sent and received on each network adapter, including framing characters.
impact test Storage
in a production Citrix VDI-in-a-Box configuration, VDI-in-a-Box Office Manager falters power-ups, launching an office every 30 seconds. My first test scalability for 50 users, it took 25 minutes to power on all 50 virtual machines. In the stress test, I wanted to evaluate the ability of I / O subsystem storage, so I created an artificial boot storm by turning on 50 workstations in a period of 10 minutes. Although this stress test does not represent a true generation VDI-in-a-Box deployment, it provides valuable information on the performance of the storage subsystem.
The graph below shows the calculated IOPS read and write for storage. impact test
the length of the average disk queue is the average number of read and write requests combined queued for C: during sample interval. The warning range indicates that there are more than two I / O waiting on the physical disk.
Summary
While you should take into account all relevant configuration factors - type user workload, the availability requirements, etc. - When designing an actual production deployment, this test has strengthened the validity of the initial configurations 50- and 100 by the user in an average workload Login VSI. The results of these stress tests indicated that the reference architecture configurations had the ability of users to save, and are not negatively affected by the restart workstations along the way. The storage configuration could also support applications I / O of the artificial boot storm. For more details on configurations, the tests and results, see the report reference architecture .
In my next blog, I will describe the testing of the user experience that I performed for this reference architecture, which served to corroborate the results of my tests of scalability and capacity.
References
- Virtualization HP SMB Client Reference Architecture for Citrix VDI-in-a-Box (PDF)
- HP and Citrix VDI-in-a-Box: Part 1 - reference architecture scaling (Blog)
- HP and Citrix VDI-in -a Box: Part 3 - User Experience Testing (Blog)
0 Komentar