Skip to content

Virtualization performance test scripts for network and block devices.

License

Notifications You must be signed in to change notification settings

schrht/virt-perf-scripts

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

virt-perf-scripts

The performance test scripts for virtualization.

FIO test

This tool is designed for running an FIO benchmark in guest.

Prepare the guest

  1. Install fio on the guest.

sudo yum install -y libaio fio gnuplot

  1. Install the following Python modules:
  • click
  • pandas
  • numpy
  • scipy
  • yaml

Notes:
You can use ./block/setup.sh for step 1 and 2 on RHEL systems.

  1. Deliver the following scripts to the guest:
  • ./block/RunFioTest.py
  • ./block/GenerateBenchmarkReport.py
  • ./block/GenerateTestReport.py
  • ./virt_perf_scripts.yaml

Run FIO test

The manual page of RunFioTest.py:

$ python3 ./RunFioTest.py --help
Usage: RunFioTest.py [OPTIONS]

  Command line interface.

  Take arguments from CLI, load default parameters from yaml file. Then
  initialize the fio test.

Options:
  --backend TEXT           The backend device where vdisk image is based on.
  --driver TEXT            The driver to power the vdisk..
  --fs TEXT                The filesystem of the disk to be tested, "RAW" for
                           no fs.
  --rounds INTEGER RANGE   How many rounds the fio test will be repeated.
  --filename TEXT          [FIO] The disk(s) or specified file(s) to be tested
                           by fio. You can specify a number of targets by
                           separating the names with a ':' colon.
  --runtime TEXT           [FIO] Terminate a job after the specified period of
                           time.
  --ioengine TEXT          [FIO] Defines how the job issues I/O to the file.
                           Such as: 'libaio', 'io_uring', etc.
  --direct INTEGER RANGE   [FIO] Direct access to the disk.
  --numjobs INTEGER RANGE  [FIO] Create the specified number of clones of the
                           job.
  --rw_list TEXT           [FIO] Type of I/O pattern.
  --bs_list TEXT           [FIO] The block size in bytes used for I/O units.
  --iodepth_list TEXT      [FIO] # of I/O units to keep in flight against the
                           file.
  --log_path TEXT          Where the *.fiolog files will be saved to.
  --plots / --no-plots     Generate bw/iops/lat logs and plots in their
                           lifetime.
  --dryrun                 Print the commands that would be executed, but do
                           not execute them.
  --help                   Show this message and exit.

If you run ./RunFioTest.py without any parameter, it will load default value from ./virt_perf_scripts.yaml. Please check the yaml file for details.

Typically, you should run the following command to provide enough information:

$ python3 ./RunFioTest.py --backend NVME --driver SCSI --fs RAW --filename /dev/sdb --log_path $HOME/workspace/log/ESXi_FIO_RHEL7u6_20180809

This command will create $HOME/workspace/log/ESXi_FIO_RHEL7u6_20180809 and generate *.fiolog file for each subcase to this path.

Generate FIO test report

The manual page of GenerateTestReport.py:

$ python3 ./GenerateTestReport.py --help
Usage: GenerateTestReport.py [OPTIONS]

  Command Line Interface.

Options:
  --result_path PATH  Specify the path where *.fiolog files are stored in.
  --report_csv PATH   Specify the name of CSV file for fio test reports.
  --help              Show this message and exit.

Typically, you should run the following command:

$ python3 ./GenerateTestReport.py --result_path $HOME/workspace/log/ESXi_FIO_RHEL7u6_20180809 --report_csv ESXi_FIO_RHEL7u6_20180809.csv

This command will create a CSV test report with all the performance KPIs in.

Generate FIO benchmark report

The manual page of GenerateBenchmarkReport.py:

$ python3 ./GenerateBenchmarkReport.py --help
Usage: GenerateBenchmarkReport.py [OPTIONS]

  Command Line Interface.

Options:
  --base_csv PATH    Specify the CSV file of the base samples.
  --test_csv PATH    Specify the CSV file of the test samples.
  --report_csv PATH  Specify the CSV file to store the benchmark report.
  --help             Show this message and exit.

Typically, you should run the following command:

$ python3 ./GenerateBenchmarkReport.py --base_csv ./ESXi_FIO_RHEL7u5_20180401.csv --test_csv ./ESXi_FIO_RHEL7u6_20180809.csv --report_csv ESXi_FIO_Benchmark_RHEL7u6_against_RHEL7u5_20180809.csv

This command will create a CSV benchmark report which comparing RHEL7.6 performance KPIs against RHEL7.5.

About the index and conclusion

The conclusion can be the following values in specific situations:

Conclusion              Situation
Data Invalid            The input data is invalid;
Variance Too Large      The %SD beyonds the MAX_PCT_DEV;
No Difference           The %DIFF is zero;
No Significance         The Significance less than CONFIDENCE_THRESHOLD;
Major Improvement       The Significance beyonds CONFIDENCE_THRESHOLD and %DIFF beyonds REGRESSION_THRESHOLD;
Major Regression        The Significance beyonds CONFIDENCE_THRESHOLD and %DIFF beyonds REGRESSION_THRESHOLD;
Minor Improvement       The Significance beyonds CONFIDENCE_THRESHOLD but %DIFF belows REGRESSION_THRESHOLD;
Minor Regression        The Significance beyonds CONFIDENCE_THRESHOLD but %DIFF belows REGRESSION_THRESHOLD;

MAX_PCT_DEV = 10
REGRESSION_THRESHOLD = 5
CONFIDENCE_THRESHOLD = 0.95

Calculation:

AVG = SUM(the performance number of sample 1~5) / 5
%SD = (The Standard Deviation of the 5 samples) / AVG * 100%
%DIFF = (TEST-AVG - BASE-AVG) / BASE-AVG * 100%
Significance = (1 - TTEST(BASE Sample 1~5, TEST Sample 1~5))

Paste the results into Google Speardsheets

You can copy & paste the contents from CSV file into the Template of Google Speardsheets. So that you could check the benchmark results much more conveniently.

About

Virtualization performance test scripts for network and block devices.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published