yardstick.benchmark.scenarios.compute package¶
Submodules¶
yardstick.benchmark.scenarios.compute.cpuload module¶
Processor statistics and system load.
-
class
yardstick.benchmark.scenarios.compute.cpuload.
CPULoad
(scenario_cfg, context_cfg)[source]¶ Bases:
yardstick.benchmark.scenarios.base.Scenario
Collect processor statistics and system load.
This scenario reads system load averages and CPU usage statistics on a Linux host.
CPU usage statistics are read using the utility ‘mpstat’.
If ‘mpstat’ is not installed on the host usage statistics are instead read directly from ‘/proc/stat’.
Load averages are read from the file ‘/proc/loadavg’ on the Linux host.
Parameters: - - Time interval to measure CPU usage. A value of 0 (interval) – indicates that processors statistics are to be reported for the time since system startup (boot)
- type – [int]
- unit – seconds
- default – 0
-
MPSTAT_FIELD_SIZE
= 10¶
yardstick.benchmark.scenarios.compute.cyclictest module¶
-
class
yardstick.benchmark.scenarios.compute.cyclictest.
Cyclictest
(scenario_cfg, context_cfg)[source]¶ Bases:
yardstick.benchmark.scenarios.base.Scenario
Execute cyclictest benchmark on guest vm
Parameters: - - run thread #N on processor #N, if possible (affinity) – type: int unit: na default: 1
- - base interval of thread (interval) – type: int unit: us default: 1000
- - number of loops, 0 for endless (loops) – type: int unit: na default: 1000
- - priority of highest prio thread (priority) – type: int unit: na default: 99
- - number of threads (threads) – type: int unit: na default: 1
- - dump a latency histogram to stdout after the run (histogram) – here set the max time to be tracked
type: int unit: ms default: 90
- link below for more fio args description (Read) – https://rt.wiki.kernel.org/index.php/Cyclictest
-
REBOOT_CMD_PATTERN
= ';\\s*reboot\\b'¶
-
TARGET_SCRIPT
= 'cyclictest_benchmark.bash'¶
-
WORKSPACE
= '/root/workspace/'¶
yardstick.benchmark.scenarios.compute.lmbench module¶
-
class
yardstick.benchmark.scenarios.compute.lmbench.
Lmbench
(scenario_cfg, context_cfg)[source]¶ Bases:
yardstick.benchmark.scenarios.base.Scenario
Execute lmbench memory read latency or memory bandwidth benchmark in a host
Parameters: - specifies whether to measure memory latency or bandwidth (test_type) – type: string unit: na default: “latency” - Parameters for memory read latency benchmark
- stride - number of locations in memory between starts of array elements
- type: int unit: bytes default: 128
- stop_size - maximum array size to test (minimum value is 0.000512)
- type: float unit: megabytes default: 16.0
Results are accurate to the ~2-5 nanosecond range.
- Parameters for memory bandwidth benchmark
- size - the amount of memory to test
- type: int unit: kilobyte default: 128
benchmark - the name of the memory bandwidth benchmark test to execute. Valid test names are rd, wr, rdwr, cp, frd, fwr, fcp, bzero, bcopy
type: string unit: na default: “rd”- warmup - the number of repetitons to perform before taking measurements
- type: int unit: na default: 0
more info http://manpages.ubuntu.com/manpages/trusty/lmbench.8.html
-
BANDWIDTH_BENCHMARK_SCRIPT
= 'lmbench_bandwidth_benchmark.bash'¶
-
LATENCY_BENCHMARK_SCRIPT
= 'lmbench_latency_benchmark.bash'¶
yardstick.benchmark.scenarios.compute.perf module¶
-
class
yardstick.benchmark.scenarios.compute.perf.
Perf
(scenario_cfg, context_cfg)[source]¶ Bases:
yardstick.benchmark.scenarios.base.Scenario
Execute perf benchmark in a host
Parameters: - - perf tool software, hardware or tracepoint events (events) – type: [str] unit: na default: [‘task-clock’]
- - simulate load on the host by doing IO operations (load) – type: bool unit: na default: false
- more info about perf and perf events see https (For) – //perf.wiki.kernel.org
-
TARGET_SCRIPT
= 'perf_benchmark.bash'¶
yardstick.benchmark.scenarios.compute.unixbench module¶
-
class
yardstick.benchmark.scenarios.compute.unixbench.
Unixbench
(scenario_cfg, context_cfg)[source]¶ Bases:
yardstick.benchmark.scenarios.base.Scenario
Execute Unixbench cpu benchmark in a host The Run script takes a number of options which you can use to customise a test, and you can specify the names of the tests to run. The full usage is:
Run [ -q | -v ] [-i <n> ] [-c <n> [-c <n> ...]] [test ...]
-i <count> Run <count> iterations for each test – slower tests use <count> / 3, but at least 1. Defaults to 10 (3 for slow tests). -c <n> Run <n> copies of each test in parallel. - Parameters for setting unixbench
- run_mode - Run in quiet mode or verbose mode
- type: string unit: None default: None
- test_type - The available tests are organised into categories;
- type: string unit: None default: None
iterations - Run <count> iterations for each test – slower tests use <count> / 3, but at least 1. Defaults to 10 (3 for slow tests).
type: int unit: None default: None- copies - Run <n> copies of each test in parallel.
- type: int unit: None default: None
more info https://github.com/kdlucas/byte-unixbench/blob/master/UnixBench
-
TARGET_SCRIPT
= 'unixbench_benchmark.bash'¶