test suite reviews and discussions
 help / color / mirror / Atom feed
* [dts] [PATCH V1 0/3] power_empty_poll: upload automation script and test plan
@ 2019-10-21  6:11 yufengmx
  2019-10-21  6:11 ` [dts] [PATCH V1 1/3] tests/power_empty_poll: upload automation script yufengmx
                   ` (2 more replies)
  0 siblings, 3 replies; 9+ messages in thread
From: yufengmx @ 2019-10-21  6:11 UTC (permalink / raw)
  To: dts, lei.a.yao; +Cc: yufengmx

 For packet processing workloads such as DPDK polling is continuous. 
 This means CPU cores always show 100% busy independent of how much work 
 those cores are doing. It is critical to accurately determine how busy 
 a core is hugely important for the following reasons. 

yufengmx (3):
  tests/power_empty_poll: upload automation script
  test_plans/power_empty_poll: add test plan index
  test_plans/power_empty_poll: upload test plan

 test_plans/index.rst                      |   1 +
 test_plans/power_empty_poll_test_plan.rst | 138 +++++++++
 tests/TestSuite_power_empty_poll.py       | 359 ++++++++++++++++++++++
 3 files changed, 498 insertions(+)
 create mode 100644 test_plans/power_empty_poll_test_plan.rst
 create mode 100644 tests/TestSuite_power_empty_poll.py

-- 
2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* [dts] [PATCH V1 1/3] tests/power_empty_poll: upload automation script
  2019-10-21  6:11 [dts] [PATCH V1 0/3] power_empty_poll: upload automation script and test plan yufengmx
@ 2019-10-21  6:11 ` yufengmx
  2019-10-21  7:44   ` Yao, Lei A
  2019-10-21  6:11 ` [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index yufengmx
  2019-10-21  6:11 ` [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan yufengmx
  2 siblings, 1 reply; 9+ messages in thread
From: yufengmx @ 2019-10-21  6:11 UTC (permalink / raw)
  To: dts, lei.a.yao; +Cc: yufengmx


For packet processing workloads such as DPDK polling is continuous.
This means CPU cores always show 100% busy independent of how much work
those cores are doing. It is critical to accurately determine how busy
a core is hugely important for the following reasons.

Signed-off-by: yufengmx <yufengx.mo@intel.com>
---
 tests/TestSuite_power_empty_poll.py | 359 ++++++++++++++++++++++++++++
 1 file changed, 359 insertions(+)
 create mode 100644 tests/TestSuite_power_empty_poll.py

diff --git a/tests/TestSuite_power_empty_poll.py b/tests/TestSuite_power_empty_poll.py
new file mode 100644
index 0000000..3556e74
--- /dev/null
+++ b/tests/TestSuite_power_empty_poll.py
@@ -0,0 +1,359 @@
+# BSD LICENSE
+#
+# Copyright(c) 2010-2019 Intel Corporation. All rights reserved.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#
+#   * Redistributions of source code must retain the above copyright
+#     notice, this list of conditions and the following disclaimer.
+#   * Redistributions in binary form must reproduce the above copyright
+#     notice, this list of conditions and the following disclaimer in
+#     the documentation and/or other materials provided with the
+#     distribution.
+#   * Neither the name of Intel Corporation nor the names of its
+#     contributors may be used to endorse or promote products derived
+#     from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+"""
+DPDK Test suite.
+l3fwd-power power management test suite.
+"""
+import os
+import time
+import textwrap
+import traceback
+from copy import deepcopy
+
+from utils import create_mask as dts_create_mask
+from test_case import TestCase
+
+from packet import Packet
+from pktgen import TRANSMIT_CONT
+
+
+class TestPowerEmptPoll(TestCase):
+    TRAIN = 'train'
+    NOTRAIN = 'no-train'
+    MED = 'med_threshold'
+    HIGH = 'high_threshold'
+    query_min_freq = '/tmp/cpu_min.log'
+    query_max_freq = '/tmp/cpu_max.log'
+    output_path = '/tmp'
+
+    @property
+    def target_dir(self):
+        # get absolute directory of target source code
+        target_dir = '/root' + self.dut.base_dir[1:] \
+                     if self.dut.base_dir.startswith('~') else \
+                     self.dut.base_dir
+        return target_dir
+
+    def d_con(self, cmd):
+        _cmd = [cmd, '# ', 10] if isinstance(cmd, (str, unicode)) else cmd
+        return self.dut.send_expect(*_cmd)
+
+    def d_a_con(self, cmd):
+        _cmd = [cmd, '# ', 10] if isinstance(cmd, (str, unicode)) else cmd
+        return self.dut.alt_session.send_expect(*_cmd)
+
+    def prepare_binary(self, name):
+        example_dir = "examples/" + name
+        out = self.dut.build_dpdk_apps('./' + example_dir)
+        self.verify("Error" not in out, "Compilation error")
+        self.verify("No such" not in out, "Compilation error")
+        binary_dir = os.path.join(self.target_dir, example_dir, 'build')
+        cmd = ["ls -F {0} | grep '*'".format(binary_dir), '# ', 5]
+        exec_file = self.d_a_con(cmd)
+        binary_file = os.path.join(binary_dir, exec_file[:-1])
+        return binary_file
+
+    def get_cores_mask(self, cores_list):
+        return dts_create_mask(cores_list)
+
+    def set_pktgen_stream(self, txport, rxport, send_pkts, option):
+        stream_ids = []
+        cnt = 0
+        for pkt in send_pkts:
+            _option = deepcopy(option)
+            _option['pcap'] = pkt
+            stream_id = self.tester.pktgen.add_stream(txport, rxport, pkt)
+            self.tester.pktgen.config_stream(stream_id, _option)
+            stream_ids.append(stream_id)
+            # rxport -> txport
+            stream_id = self.tester.pktgen.add_stream(rxport, txport, pkt)
+            self.tester.pktgen.config_stream(stream_id, _option)
+            stream_ids.append(stream_id)
+            cnt += 1
+        return stream_ids
+
+    def run_traffic(self, option):
+        txport = self.tester.get_local_port(self.dut_ports[0])
+        rxport = self.tester.get_local_port(self.dut_ports[1])
+        stm_type = option.get('stm_types')
+        rate_percent = option.get('rate', float(100))
+        duration = option.get('duration', 10)
+        send_pkts = self.set_stream(stm_type)
+        # set stream into pktgen
+        option = {
+            'stream_config': {
+                'txmode': {},
+                'transmit_mode': TRANSMIT_CONT,
+                'rate': rate_percent, }
+        }
+        stream_ids = self.set_pktgen_stream(txport, rxport, send_pkts, option)
+        # run traffic options
+        traffic_opt = {
+            'method': 'throughput',
+            'duration': duration, }
+        # run pktgen(ixia/trex) traffic
+        result = self.tester.pktgen.measure(stream_ids, traffic_opt)
+
+        return result
+
+    def set_stream(self, stm_names=None):
+        # set streams for traffic
+        pkt_configs = {
+            'UDP_1': {
+                'type': 'UDP',
+                'pkt_layers': {'ipv4': {'dst': '1.1.1.1'}, }},
+        }
+        # create packet instance for send
+        streams = []
+        for stm_name in stm_names:
+            if stm_name not in pkt_configs.keys():
+                continue
+            values = pkt_configs[stm_name]
+            pkt_type = values.get('type')
+            pkt_layers = values.get('pkt_layers')
+            pkt = Packet(pkt_type=pkt_type)
+            for layer in pkt_layers.keys():
+                pkt.config_layer(layer, pkt_layers[layer])
+            streams.append(pkt.pktgen.pkt)
+
+        return streams
+
+    @property
+    def empty_poll_options(self):
+        table = {
+            'train': '1,0,0,',
+            'no-train': '0,350000,500000', }
+        return table
+
+    def init_l3fwd_power(self):
+        self.l3fwd_power = self.prepare_binary('l3fwd-power')
+
+    def start_l3fwd_power(self, core):
+        train_mode = self.empty_poll_options.get(self.train_mode)
+        option = (' '
+                  '-c {core_mask} '
+                  '-n {mem_channel} '
+                  '-- '
+                  '-p 0x1 '
+                  '-P '
+                  '--config="(0,0,2)" '
+                  '-l 10 -m 6 -h 1'
+                  '--empty-poll="{empty-poll}" '
+                  ).format(**{
+                      'core_mask': self.get_cores_mask(core),
+                      'mem_channel': self.dut.get_memory_channels(),
+                      'empty-poll': train_mode, })
+        prompt = 'L3FWD_POWER: entering main loop on lcore'
+        cmd = [' '.join([self.l3fwd_power, option]), prompt, 60]
+        self.d_con(cmd)
+
+    def close_l3fwd_power(self):
+        cmd = 'killall l3fwd-power'
+        self.d_a_con(cmd)
+
+    def init_query_script(self):
+        script_content = textwrap.dedent("""
+            # $1: delay time before traffic start
+            # $2: core number
+            sleep 5
+            while :
+            do
+                sleep 1
+                cat /sys/devices/system/cpu/cpu$1/cpufreq/scaling_min_freq >> {0}
+                cat /sys/devices/system/cpu/cpu$1/cpufreq/scaling_max_freq >> {1}
+            done
+        """).format(self.query_min_freq, self.query_max_freq)
+        fileName = 'vm_power_core.sh'
+        query_script = os.path.join(self.output_path, fileName)
+        with open(query_script, 'wb') as fp:
+            fp.write('#! /bin/sh' + os.linesep + script_content)
+        self.dut.session.copy_file_to(query_script, self.target_dir)
+        self.query_tool = ';'.join([
+            'cd {}'.format(self.target_dir),
+            'chmod 777 {}'.format(fileName),
+            './' + fileName])
+
+    def start_query(self, core):
+        cmd = self.query_tool + ' {0} > /dev/null 2>&1 &'.format(core)
+        self.d_a_con(cmd)
+
+    def stop_query(self):
+        cmd = 'pkill {}'.format(os.path.basename(self.query_tool))
+        self.d_a_con(cmd)
+        self.dut.session.copy_file_from(self.query_min_freq, self.output_path)
+        self.dut.session.copy_file_from(self.query_max_freq, self.output_path)
+
+    def get_sys_power_driver(self):
+        drv_file = "/sys/devices/system/cpu/cpu0/cpufreq/scaling_driver"
+        output = self.d_a_con('cat ' + drv_file)
+        if not output:
+            msg = 'unknown power driver'
+            self.verify(False, msg)
+        drv_name = output.splitlines()[0].strip()
+        return drv_name
+
+    def get_no_turbo_max(self):
+        cmd = 'rdmsr -p 1 0x0CE -f 15:8 -d'
+        output = self.d_a_con(cmd)
+        freq = output.strip() + '00000'
+        return int(freq)
+
+    def check_core_freq_in_traffic(self, core_index, mode):
+        '''
+        check the cores frequency when running traffic
+             highest frequency[no_turbo_max]: cur_min=cur_max=no_turbo_max
+        '''
+        self.stop_query()
+        freq = self.get_no_turbo_max()
+        expected_freq = str(freq if mode == self.HIGH else (freq - 500000))
+        query_max_freq = os.path.join(
+            self.output_path, os.path.basename(self.query_max_freq))
+        with open(query_max_freq, 'rb') as fp:
+            content = fp.read()
+        msg = 'max freq are not the same as highest frequency <{0}>'
+        self.verify(expected_freq in content, msg.format(expected_freq))
+        query_min_freq = os.path.join(
+            self.output_path, os.path.basename(self.query_min_freq))
+        with open(query_min_freq, 'rb') as fp:
+            content = fp.read()
+        msg = 'min freq are not the same as highest frequency <{0}>'
+        self.verify(expected_freq in content, msg.format(expected_freq))
+        msg = 'core <{0}>: max freq/min_freq/expected freq<{1}> are the same'
+        self.logger.info(msg.format(core_index, expected_freq))
+
+    def check_no_train(self):
+        output = self.dut.get_session_output(timeout=2)
+        msg = 'training steps should not be executed'
+        self.verify('POWER: Training is Complete' not in output, msg)
+
+    def verify_train_mode(self):
+        except_content = None
+        # begin run vm power policy testing
+        try:
+            self.start_l3fwd_power(self.check_core)
+            if self.train_mode == self.NOTRAIN:
+                self.check_no_train()
+            else:
+                # Injected Rate(64B, dst_ip=1.1.1.1):
+                #      10G -> 0.1G -> 10G -> 0.1G -> 10G -> 0.1G
+                check_item = [
+                    [100, self.HIGH],
+                    [1, self.MED],
+                    [100, self.HIGH],
+                    [1, self.MED],
+                    [100, self.HIGH],
+                    [1, self.MED], ]
+                msg = '{0} begin test mode <{1}> with traffic rate percent {2}%'
+                for rate, mode in check_item:
+                    self.logger.info(msg.format(self.train_mode, mode, rate))
+                    self.start_query(self.check_core[1])
+                    info = {
+                        'stm_types': ['UDP_1'],
+                        'rate': rate}
+                    # run traffic
+                    self.run_traffic(info)
+                    time.sleep(2)
+                    self.stop_query()
+                    # check test result
+                    self.check_core_freq_in_traffic(self.check_core[1], mode)
+        except Exception as e:
+            self.logger.error(traceback.format_exc())
+            except_content = e
+        finally:
+            self.close_l3fwd_power()
+
+        # check verify result
+        if except_content:
+            raise Exception(except_content)
+        else:
+            msg = "test <{0}> successful !!!".format(self.train_mode)
+            self.logger.info(msg)
+
+    def verify_power_driver(self):
+        expected_drv = 'intel_pstate'
+        power_drv = self.get_sys_power_driver()
+        msg = "{0} should work with {1} driver on DUT".format(
+            self.suite_name, expected_drv)
+        self.verify(power_drv == expected_drv, msg)
+
+    def preset_test_environment(self):
+        self.check_core = [1, 2]
+        # modprobe msr module to let the application can get the CPU HW info
+        self.d_a_con('modprobe msr')
+        # init binary
+        self.init_l3fwd_power()
+        self.init_query_script()
+    #
+    # Test cases.
+    #
+
+    def set_up_all(self):
+        """
+        Run at the start of each test suite.
+        """
+        self.verify_power_driver()
+        self.dut_ports = self.dut.get_ports(self.nic)
+        self.verify(len(self.dut_ports) >= 2, "Not enough ports")
+        # prepare testing environment
+        self.preset_test_environment()
+
+    def tear_down_all(self):
+        """
+        Run after each test suite.
+        """
+        pass
+
+    def set_up(self):
+        """
+        Run before each test case.
+        """
+        pass
+
+    def tear_down(self):
+        """
+        Run after each test case.
+        """
+        self.dut.kill_all()
+
+    def test_perf_basic_train_mode(self):
+        """
+        Set Branch-Ratio Rate by User
+        """
+        self.train_mode = self.TRAIN
+        self.verify_train_mode()
+
+    def test_perf_no_training_mode(self):
+        """
+        Set Branch-Ratio Rate by User
+        """
+        self.train_mode = self.NOTRAIN
+        self.verify_train_mode()
-- 
2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index
  2019-10-21  6:11 [dts] [PATCH V1 0/3] power_empty_poll: upload automation script and test plan yufengmx
  2019-10-21  6:11 ` [dts] [PATCH V1 1/3] tests/power_empty_poll: upload automation script yufengmx
@ 2019-10-21  6:11 ` yufengmx
  2019-10-21  7:45   ` Yao, Lei A
  2019-10-23 10:03   ` Tu, Lijuan
  2019-10-21  6:11 ` [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan yufengmx
  2 siblings, 2 replies; 9+ messages in thread
From: yufengmx @ 2019-10-21  6:11 UTC (permalink / raw)
  To: dts, lei.a.yao; +Cc: yufengmx


For packet processing workloads such as DPDK polling is continuous.
This means CPU cores always show 100% busy independent of how much work
those cores are doing. It is critical to accurately determine how busy
a core is hugely important for the following reasons.

Signed-off-by: yufengmx <yufengx.mo@intel.com>
---
 test_plans/index.rst | 1 +
 1 file changed, 1 insertion(+)

diff --git a/test_plans/index.rst b/test_plans/index.rst
index a10d171..a8a9543 100644
--- a/test_plans/index.rst
+++ b/test_plans/index.rst
@@ -154,6 +154,7 @@ The following are the test plans for the DPDK DTS automated test system.
     vhost_qemu_mtu_test_plan
     vhost_user_live_migration_test_plan
     vm_power_manager_test_plan
+    power_empty_poll_test_plan
     vmdq_test_plan
     vf_l3fwd_test_plan
     softnic_test_plan
-- 
2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan
  2019-10-21  6:11 [dts] [PATCH V1 0/3] power_empty_poll: upload automation script and test plan yufengmx
  2019-10-21  6:11 ` [dts] [PATCH V1 1/3] tests/power_empty_poll: upload automation script yufengmx
  2019-10-21  6:11 ` [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index yufengmx
@ 2019-10-21  6:11 ` yufengmx
  2019-10-21  7:45   ` Yao, Lei A
  2019-10-23 10:03   ` Tu, Lijuan
  2 siblings, 2 replies; 9+ messages in thread
From: yufengmx @ 2019-10-21  6:11 UTC (permalink / raw)
  To: dts, lei.a.yao; +Cc: yufengmx

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=a, Size: 7180 bytes --]


For packet processing workloads such as DPDK polling is continuous.
This means CPU cores always show 100% busy independent of how much work
those cores are doing. It is critical to accurately determine how busy
a core is hugely important for the following reasons.

Signed-off-by: yufengmx <yufengx.mo@intel.com>
---
 test_plans/power_empty_poll_test_plan.rst | 138 ++++++++++++++++++++++
 1 file changed, 138 insertions(+)
 create mode 100644 test_plans/power_empty_poll_test_plan.rst

diff --git a/test_plans/power_empty_poll_test_plan.rst b/test_plans/power_empty_poll_test_plan.rst
new file mode 100644
index 0000000..01c0ebd
--- /dev/null
+++ b/test_plans/power_empty_poll_test_plan.rst
@@ -0,0 +1,138 @@
+.. Copyright (c) <2010-2019>, Intel Corporation
+   All rights reserved.
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions
+   are met:
+
+   - Redistributions of source code must retain the above copyright
+     notice, this list of conditions and the following disclaimer.
+
+   - Redistributions in binary form must reproduce the above copyright
+     notice, this list of conditions and the following disclaimer in
+     the documentation and/or other materials provided with the
+     distribution.
+
+   - Neither the name of Intel Corporation nor the names of its
+     contributors may be used to endorse or promote products derived
+     from this software without specific prior written permission.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+   FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+   COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+   INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+   (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+   SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+   HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+   STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+   ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+   OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=========================
+Power Lib Empty Poll Test
+=========================
+
+Inband Policy Control
+=====================
+
+For packet processing workloads such as DPDK polling is continuous. This means
+CPU cores always show 100% busy independent of how much work those cores are
+doing. It is critical to accurately determine how busy a core is hugely
+important for the following reasons:
+
+   * No indication of overload conditions
+
+   * User do not know how much real load is on a system meaning resulted in
+     wasted energy as no power management is utilized
+
+Tried and failed schemes include calculating the cycles required from the load
+on the core, in other words the busyness. For example, how many cycles it costs
+to handle each packet and determining the frequency cost per core. Due to the
+varying nature of traffic, types of frames and cost in cycles to process, this
+mechanism becomes complex quickly where a simple scheme is required to solve
+the problems.
+
+For all polling mechanism, the proposed solution focus on how many times empty
+poll executed instead of calculating how many cycles it cost to handle each
+packet. The less empty poll number means current core is busy with processing
+workload, therefore,  the higher frequency is needed. The high empty poll
+number indicate current core has lots spare time, therefore, we can lower the
+frequency.
+
+2.1 Power state definition:
+
+LOW:  the frequency is used for purge mode.
+
+MED:  the frequency is used to process modest traffic workload.
+
+HIGH: the frequency is used to process busy traffic workload.
+
+2.2 There are two phases to establish the power management system:
+
+a.Initialization/Training phase. There is no traffic pass-through, the system
+will test average empty poll numbers  with LOW/MED/HIGH  power state. Those
+average empty poll numbers will be the baseline for the normal phase. The
+system will collect all core's counter every 100ms. The Training phase will
+take 5 seconds.
+
+b.Normal phase. When the real traffic pass-though, the system will compare
+run-time empty poll moving average value with base line then make decision to
+move to HIGH power state of MED  power state. The system will collect all
+core's counter every 10ms.
+
+``training_flag`` : optional, enable/disable training mode. Default value is 0.
+ If the training_flag is set as 1(true), then the application will start in
+ training mode and print out the trained threshold values. If the training_flag
+ is set as 0(false), the application will start in normal mode, and will use
+ either the default thresholds or those supplied on the command line. The
+ trained threshold values are specific to the user’s system, may give a better
+ power profile when compared to the default threshold values.
+
+``med_threshold`` : optional, sets the empty poll threshold of a modestly busy
+system state. If this is not supplied, the application will apply the default
+value of 350000.
+
+``high_threshold`` : optional, sets the empty poll threshold of a busy system
+state. If this is not supplied, the application will apply the default value of
+580000.
+
+
+Preparation Work for Settings
+=============================
+1. Turn on Speedstep option in BIOS
+2. Turn on Turbo in BIOS
+3. Use intel_pstate driver for CPU frequency control
+4. modprobe msr
+
+sys_min=/sys/devices/system/cpu/cpu{}/cpufreq/cpuinfo_min_freq
+sys_max=/sys/devices/system/cpu/cpu{}/cpufreq/cpuinfo_max_freq 
+no_turbo_max=$(rdmsr -p 1 0x0CE -f 15:8 -d)00000
+
+cur_min=/sys/devices/system/cpu/cpu{}/cpufreq/scaling_min_freq
+cur_max=/sys/devices/system/cpu/cpu{}/cpufreq/scaling_max_freq
+
+
+Test Case1 : Basic Training mode test based on one NIC with l3fwd-power
+=======================================================================
+Step 1. Bind One NIC to DPDK driver, launch l3fwd-power with empty-poll enabled
+
+    ./l3fwd-power -l 1-2 -n 4 -- -p 0x1 -P --config="(0,0,2)" --empty-poll="1,0,0" -l 10 -m 6 -h 1
+
+Step 2. Check the log also when changing the inject packet rate as following:
+
+    Injected Rate(64B, dst_ip=1.1.1.1): 10G -> 0.1G -> 10G -> 0.1G -> 10G ->
+    0.1G The frequency will be set to MED when we inject 0.1G and return to HGH
+    when inject 10G Rate, check the frequency of the forwarding core(core 2)
+    When traffic is 10G:  cur_min=cur_max=no_turbo_max
+    When traffic is 0.1G: cur_min=cur_max=[no_turbo_max-500000]
+
+
+Test Case2: No-Training mode test based on one NIC with l3fwd-power
+===================================================================
+Step 1. Bind One NIC to DPDK driver, launch l3fwd-power with empty-poll enabled
+
+   ./l3fwd-power -l 1-2 -n 4  -- -p 0x1 -P --config="(0,0,2)" --empty-poll="0,350000,500000" -l 10 -m 6 -h 1
+
+Step 2. Check no training steps are executed in sample's launch log.
-- 
2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [dts] [PATCH V1 1/3] tests/power_empty_poll: upload automation script
  2019-10-21  6:11 ` [dts] [PATCH V1 1/3] tests/power_empty_poll: upload automation script yufengmx
@ 2019-10-21  7:44   ` Yao, Lei A
  0 siblings, 0 replies; 9+ messages in thread
From: Yao, Lei A @ 2019-10-21  7:44 UTC (permalink / raw)
  To: Mo, YufengX, dts



> -----Original Message-----
> From: Mo, YufengX <yufengx.mo@intel.com>
> Sent: Monday, October 21, 2019 2:11 PM
> To: dts@dpdk.org; Yao, Lei A <lei.a.yao@intel.com>
> Cc: Mo, YufengX <yufengx.mo@intel.com>
> Subject: [dts][PATCH V1 1/3] tests/power_empty_poll: upload automation
> script
> 
> 
> For packet processing workloads such as DPDK polling is continuous.
> This means CPU cores always show 100% busy independent of how much
> work those cores are doing. It is critical to accurately determine how busy a
> core is hugely important for the following reasons.
> 
> Signed-off-by: yufengmx <yufengx.mo@intel.com>
Acked-by: lei yao<lei.a.yao@intel.com>
> ---
>  tests/TestSuite_power_empty_poll.py | 359
> ++++++++++++++++++++++++++++
>  1 file changed, 359 insertions(+)
>  create mode 100644 tests/TestSuite_power_empty_poll.py
> 
> diff --git a/tests/TestSuite_power_empty_poll.py
> b/tests/TestSuite_power_empty_poll.py
> new file mode 100644
> index 0000000..3556e74
> --- /dev/null
> +++ b/tests/TestSuite_power_empty_poll.py
> @@ -0,0 +1,359 @@
> +# BSD LICENSE
> +#
> +# Copyright(c) 2010-2019 Intel Corporation. All rights reserved.
> +# All rights reserved.
> +#
> +# Redistribution and use in source and binary forms, with or without #
> +modification, are permitted provided that the following conditions #
> +are met:
> +#
> +#   * Redistributions of source code must retain the above copyright
> +#     notice, this list of conditions and the following disclaimer.
> +#   * Redistributions in binary form must reproduce the above copyright
> +#     notice, this list of conditions and the following disclaimer in
> +#     the documentation and/or other materials provided with the
> +#     distribution.
> +#   * Neither the name of Intel Corporation nor the names of its
> +#     contributors may be used to endorse or promote products derived
> +#     from this software without specific prior written permission.
> +#
> +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> CONTRIBUTORS #
> +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
> #
> +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
> FOR #
> +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> COPYRIGHT #
> +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> INCIDENTAL, #
> +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
> NOT #
> +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
> USE, #
> +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
> ON ANY #
> +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT #
> +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
> THE USE #
> +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> DAMAGE.
> +
> +"""
> +DPDK Test suite.
> +l3fwd-power power management test suite.
> +"""
> +import os
> +import time
> +import textwrap
> +import traceback
> +from copy import deepcopy
> +
> +from utils import create_mask as dts_create_mask from test_case import
> +TestCase
> +
> +from packet import Packet
> +from pktgen import TRANSMIT_CONT
> +
> +
> +class TestPowerEmptPoll(TestCase):
> +    TRAIN = 'train'
> +    NOTRAIN = 'no-train'
> +    MED = 'med_threshold'
> +    HIGH = 'high_threshold'
> +    query_min_freq = '/tmp/cpu_min.log'
> +    query_max_freq = '/tmp/cpu_max.log'
> +    output_path = '/tmp'
> +
> +    @property
> +    def target_dir(self):
> +        # get absolute directory of target source code
> +        target_dir = '/root' + self.dut.base_dir[1:] \
> +                     if self.dut.base_dir.startswith('~') else \
> +                     self.dut.base_dir
> +        return target_dir
> +
> +    def d_con(self, cmd):
> +        _cmd = [cmd, '# ', 10] if isinstance(cmd, (str, unicode)) else cmd
> +        return self.dut.send_expect(*_cmd)
> +
> +    def d_a_con(self, cmd):
> +        _cmd = [cmd, '# ', 10] if isinstance(cmd, (str, unicode)) else cmd
> +        return self.dut.alt_session.send_expect(*_cmd)
> +
> +    def prepare_binary(self, name):
> +        example_dir = "examples/" + name
> +        out = self.dut.build_dpdk_apps('./' + example_dir)
> +        self.verify("Error" not in out, "Compilation error")
> +        self.verify("No such" not in out, "Compilation error")
> +        binary_dir = os.path.join(self.target_dir, example_dir, 'build')
> +        cmd = ["ls -F {0} | grep '*'".format(binary_dir), '# ', 5]
> +        exec_file = self.d_a_con(cmd)
> +        binary_file = os.path.join(binary_dir, exec_file[:-1])
> +        return binary_file
> +
> +    def get_cores_mask(self, cores_list):
> +        return dts_create_mask(cores_list)
> +
> +    def set_pktgen_stream(self, txport, rxport, send_pkts, option):
> +        stream_ids = []
> +        cnt = 0
> +        for pkt in send_pkts:
> +            _option = deepcopy(option)
> +            _option['pcap'] = pkt
> +            stream_id = self.tester.pktgen.add_stream(txport, rxport, pkt)
> +            self.tester.pktgen.config_stream(stream_id, _option)
> +            stream_ids.append(stream_id)
> +            # rxport -> txport
> +            stream_id = self.tester.pktgen.add_stream(rxport, txport, pkt)
> +            self.tester.pktgen.config_stream(stream_id, _option)
> +            stream_ids.append(stream_id)
> +            cnt += 1
> +        return stream_ids
> +
> +    def run_traffic(self, option):
> +        txport = self.tester.get_local_port(self.dut_ports[0])
> +        rxport = self.tester.get_local_port(self.dut_ports[1])
> +        stm_type = option.get('stm_types')
> +        rate_percent = option.get('rate', float(100))
> +        duration = option.get('duration', 10)
> +        send_pkts = self.set_stream(stm_type)
> +        # set stream into pktgen
> +        option = {
> +            'stream_config': {
> +                'txmode': {},
> +                'transmit_mode': TRANSMIT_CONT,
> +                'rate': rate_percent, }
> +        }
> +        stream_ids = self.set_pktgen_stream(txport, rxport, send_pkts, option)
> +        # run traffic options
> +        traffic_opt = {
> +            'method': 'throughput',
> +            'duration': duration, }
> +        # run pktgen(ixia/trex) traffic
> +        result = self.tester.pktgen.measure(stream_ids, traffic_opt)
> +
> +        return result
> +
> +    def set_stream(self, stm_names=None):
> +        # set streams for traffic
> +        pkt_configs = {
> +            'UDP_1': {
> +                'type': 'UDP',
> +                'pkt_layers': {'ipv4': {'dst': '1.1.1.1'}, }},
> +        }
> +        # create packet instance for send
> +        streams = []
> +        for stm_name in stm_names:
> +            if stm_name not in pkt_configs.keys():
> +                continue
> +            values = pkt_configs[stm_name]
> +            pkt_type = values.get('type')
> +            pkt_layers = values.get('pkt_layers')
> +            pkt = Packet(pkt_type=pkt_type)
> +            for layer in pkt_layers.keys():
> +                pkt.config_layer(layer, pkt_layers[layer])
> +            streams.append(pkt.pktgen.pkt)
> +
> +        return streams
> +
> +    @property
> +    def empty_poll_options(self):
> +        table = {
> +            'train': '1,0,0,',
> +            'no-train': '0,350000,500000', }
> +        return table
> +
> +    def init_l3fwd_power(self):
> +        self.l3fwd_power = self.prepare_binary('l3fwd-power')
> +
> +    def start_l3fwd_power(self, core):
> +        train_mode = self.empty_poll_options.get(self.train_mode)
> +        option = (' '
> +                  '-c {core_mask} '
> +                  '-n {mem_channel} '
> +                  '-- '
> +                  '-p 0x1 '
> +                  '-P '
> +                  '--config="(0,0,2)" '
> +                  '-l 10 -m 6 -h 1'
> +                  '--empty-poll="{empty-poll}" '
> +                  ).format(**{
> +                      'core_mask': self.get_cores_mask(core),
> +                      'mem_channel': self.dut.get_memory_channels(),
> +                      'empty-poll': train_mode, })
> +        prompt = 'L3FWD_POWER: entering main loop on lcore'
> +        cmd = [' '.join([self.l3fwd_power, option]), prompt, 60]
> +        self.d_con(cmd)
> +
> +    def close_l3fwd_power(self):
> +        cmd = 'killall l3fwd-power'
> +        self.d_a_con(cmd)
> +
> +    def init_query_script(self):
> +        script_content = textwrap.dedent("""
> +            # $1: delay time before traffic start
> +            # $2: core number
> +            sleep 5
> +            while :
> +            do
> +                sleep 1
> +                cat /sys/devices/system/cpu/cpu$1/cpufreq/scaling_min_freq >>
> {0}
> +                cat /sys/devices/system/cpu/cpu$1/cpufreq/scaling_max_freq >>
> {1}
> +            done
> +        """).format(self.query_min_freq, self.query_max_freq)
> +        fileName = 'vm_power_core.sh'
> +        query_script = os.path.join(self.output_path, fileName)
> +        with open(query_script, 'wb') as fp:
> +            fp.write('#! /bin/sh' + os.linesep + script_content)
> +        self.dut.session.copy_file_to(query_script, self.target_dir)
> +        self.query_tool = ';'.join([
> +            'cd {}'.format(self.target_dir),
> +            'chmod 777 {}'.format(fileName),
> +            './' + fileName])
> +
> +    def start_query(self, core):
> +        cmd = self.query_tool + ' {0} > /dev/null 2>&1 &'.format(core)
> +        self.d_a_con(cmd)
> +
> +    def stop_query(self):
> +        cmd = 'pkill {}'.format(os.path.basename(self.query_tool))
> +        self.d_a_con(cmd)
> +        self.dut.session.copy_file_from(self.query_min_freq, self.output_path)
> +        self.dut.session.copy_file_from(self.query_max_freq,
> + self.output_path)
> +
> +    def get_sys_power_driver(self):
> +        drv_file = "/sys/devices/system/cpu/cpu0/cpufreq/scaling_driver"
> +        output = self.d_a_con('cat ' + drv_file)
> +        if not output:
> +            msg = 'unknown power driver'
> +            self.verify(False, msg)
> +        drv_name = output.splitlines()[0].strip()
> +        return drv_name
> +
> +    def get_no_turbo_max(self):
> +        cmd = 'rdmsr -p 1 0x0CE -f 15:8 -d'
> +        output = self.d_a_con(cmd)
> +        freq = output.strip() + '00000'
> +        return int(freq)
> +
> +    def check_core_freq_in_traffic(self, core_index, mode):
> +        '''
> +        check the cores frequency when running traffic
> +             highest frequency[no_turbo_max]:
> cur_min=cur_max=no_turbo_max
> +        '''
> +        self.stop_query()
> +        freq = self.get_no_turbo_max()
> +        expected_freq = str(freq if mode == self.HIGH else (freq - 500000))
> +        query_max_freq = os.path.join(
> +            self.output_path, os.path.basename(self.query_max_freq))
> +        with open(query_max_freq, 'rb') as fp:
> +            content = fp.read()
> +        msg = 'max freq are not the same as highest frequency <{0}>'
> +        self.verify(expected_freq in content, msg.format(expected_freq))
> +        query_min_freq = os.path.join(
> +            self.output_path, os.path.basename(self.query_min_freq))
> +        with open(query_min_freq, 'rb') as fp:
> +            content = fp.read()
> +        msg = 'min freq are not the same as highest frequency <{0}>'
> +        self.verify(expected_freq in content, msg.format(expected_freq))
> +        msg = 'core <{0}>: max freq/min_freq/expected freq<{1}> are the
> same'
> +        self.logger.info(msg.format(core_index, expected_freq))
> +
> +    def check_no_train(self):
> +        output = self.dut.get_session_output(timeout=2)
> +        msg = 'training steps should not be executed'
> +        self.verify('POWER: Training is Complete' not in output, msg)
> +
> +    def verify_train_mode(self):
> +        except_content = None
> +        # begin run vm power policy testing
> +        try:
> +            self.start_l3fwd_power(self.check_core)
> +            if self.train_mode == self.NOTRAIN:
> +                self.check_no_train()
> +            else:
> +                # Injected Rate(64B, dst_ip=1.1.1.1):
> +                #      10G -> 0.1G -> 10G -> 0.1G -> 10G -> 0.1G
> +                check_item = [
> +                    [100, self.HIGH],
> +                    [1, self.MED],
> +                    [100, self.HIGH],
> +                    [1, self.MED],
> +                    [100, self.HIGH],
> +                    [1, self.MED], ]
> +                msg = '{0} begin test mode <{1}> with traffic rate percent {2}%'
> +                for rate, mode in check_item:
> +                    self.logger.info(msg.format(self.train_mode, mode, rate))
> +                    self.start_query(self.check_core[1])
> +                    info = {
> +                        'stm_types': ['UDP_1'],
> +                        'rate': rate}
> +                    # run traffic
> +                    self.run_traffic(info)
> +                    time.sleep(2)
> +                    self.stop_query()
> +                    # check test result
> +                    self.check_core_freq_in_traffic(self.check_core[1], mode)
> +        except Exception as e:
> +            self.logger.error(traceback.format_exc())
> +            except_content = e
> +        finally:
> +            self.close_l3fwd_power()
> +
> +        # check verify result
> +        if except_content:
> +            raise Exception(except_content)
> +        else:
> +            msg = "test <{0}> successful !!!".format(self.train_mode)
> +            self.logger.info(msg)
> +
> +    def verify_power_driver(self):
> +        expected_drv = 'intel_pstate'
> +        power_drv = self.get_sys_power_driver()
> +        msg = "{0} should work with {1} driver on DUT".format(
> +            self.suite_name, expected_drv)
> +        self.verify(power_drv == expected_drv, msg)
> +
> +    def preset_test_environment(self):
> +        self.check_core = [1, 2]
> +        # modprobe msr module to let the application can get the CPU HW info
> +        self.d_a_con('modprobe msr')
> +        # init binary
> +        self.init_l3fwd_power()
> +        self.init_query_script()
> +    #
> +    # Test cases.
> +    #
> +
> +    def set_up_all(self):
> +        """
> +        Run at the start of each test suite.
> +        """
> +        self.verify_power_driver()
> +        self.dut_ports = self.dut.get_ports(self.nic)
> +        self.verify(len(self.dut_ports) >= 2, "Not enough ports")
> +        # prepare testing environment
> +        self.preset_test_environment()
> +
> +    def tear_down_all(self):
> +        """
> +        Run after each test suite.
> +        """
> +        pass
> +
> +    def set_up(self):
> +        """
> +        Run before each test case.
> +        """
> +        pass
> +
> +    def tear_down(self):
> +        """
> +        Run after each test case.
> +        """
> +        self.dut.kill_all()
> +
> +    def test_perf_basic_train_mode(self):
> +        """
> +        Set Branch-Ratio Rate by User
> +        """
> +        self.train_mode = self.TRAIN
> +        self.verify_train_mode()
> +
> +    def test_perf_no_training_mode(self):
> +        """
> +        Set Branch-Ratio Rate by User
> +        """
> +        self.train_mode = self.NOTRAIN
> +        self.verify_train_mode()
> --
> 2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index
  2019-10-21  6:11 ` [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index yufengmx
@ 2019-10-21  7:45   ` Yao, Lei A
  2019-10-23 10:03   ` Tu, Lijuan
  1 sibling, 0 replies; 9+ messages in thread
From: Yao, Lei A @ 2019-10-21  7:45 UTC (permalink / raw)
  To: Mo, YufengX, dts



> -----Original Message-----
> From: Mo, YufengX <yufengx.mo@intel.com>
> Sent: Monday, October 21, 2019 2:11 PM
> To: dts@dpdk.org; Yao, Lei A <lei.a.yao@intel.com>
> Cc: Mo, YufengX <yufengx.mo@intel.com>
> Subject: [dts][PATCH V1 2/3] test_plans/power_empty_poll: add test plan
> index
> 
> 
> For packet processing workloads such as DPDK polling is continuous.
> This means CPU cores always show 100% busy independent of how much
> work those cores are doing. It is critical to accurately determine how busy a
> core is hugely important for the following reasons.
> 
> Signed-off-by: yufengmx <yufengx.mo@intel.com>
Acked-by: lei yao<lei.a.yao@intel.com>
> ---
>  test_plans/index.rst | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/test_plans/index.rst b/test_plans/index.rst index
> a10d171..a8a9543 100644
> --- a/test_plans/index.rst
> +++ b/test_plans/index.rst
> @@ -154,6 +154,7 @@ The following are the test plans for the DPDK DTS
> automated test system.
>      vhost_qemu_mtu_test_plan
>      vhost_user_live_migration_test_plan
>      vm_power_manager_test_plan
> +    power_empty_poll_test_plan
>      vmdq_test_plan
>      vf_l3fwd_test_plan
>      softnic_test_plan
> --
> 2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan
  2019-10-21  6:11 ` [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan yufengmx
@ 2019-10-21  7:45   ` Yao, Lei A
  2019-10-23 10:03   ` Tu, Lijuan
  1 sibling, 0 replies; 9+ messages in thread
From: Yao, Lei A @ 2019-10-21  7:45 UTC (permalink / raw)
  To: Mo, YufengX, dts



> -----Original Message-----
> From: Mo, YufengX <yufengx.mo@intel.com>
> Sent: Monday, October 21, 2019 2:11 PM
> To: dts@dpdk.org; Yao, Lei A <lei.a.yao@intel.com>
> Cc: Mo, YufengX <yufengx.mo@intel.com>
> Subject: [dts][PATCH V1 3/3] test_plans/power_empty_poll: upload test
> plan
> 
> 
> For packet processing workloads such as DPDK polling is continuous.
> This means CPU cores always show 100% busy independent of how much
> work those cores are doing. It is critical to accurately determine how busy a
> core is hugely important for the following reasons.
> 
> Signed-off-by: yufengmx <yufengx.mo@intel.com>
Acked: lei yao<lei.a.yao@intel.com>
> ---
>  test_plans/power_empty_poll_test_plan.rst | 138
> ++++++++++++++++++++++
>  1 file changed, 138 insertions(+)
>  create mode 100644 test_plans/power_empty_poll_test_plan.rst
> 
> diff --git a/test_plans/power_empty_poll_test_plan.rst
> b/test_plans/power_empty_poll_test_plan.rst
> new file mode 100644
> index 0000000..01c0ebd
> --- /dev/null
> +++ b/test_plans/power_empty_poll_test_plan.rst
> @@ -0,0 +1,138 @@
> +.. Copyright (c) <2010-2019>, Intel Corporation
> +   All rights reserved.
> +
> +   Redistribution and use in source and binary forms, with or without
> +   modification, are permitted provided that the following conditions
> +   are met:
> +
> +   - Redistributions of source code must retain the above copyright
> +     notice, this list of conditions and the following disclaimer.
> +
> +   - Redistributions in binary form must reproduce the above copyright
> +     notice, this list of conditions and the following disclaimer in
> +     the documentation and/or other materials provided with the
> +     distribution.
> +
> +   - Neither the name of Intel Corporation nor the names of its
> +     contributors may be used to endorse or promote products derived
> +     from this software without specific prior written permission.
> +
> +   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> CONTRIBUTORS
> +   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT
> NOT
> +   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> FITNESS
> +   FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> +   COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> INDIRECT,
> +   INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> +   (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
> GOODS OR
> +   SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
> +   HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
> CONTRACT,
> +   STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
> +   ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
> ADVISED
> +   OF THE POSSIBILITY OF SUCH DAMAGE.
> +
> +=========================
> +Power Lib Empty Poll Test
> +=========================
> +
> +Inband Policy Control
> +=====================
> +
> +For packet processing workloads such as DPDK polling is continuous.
> +This means CPU cores always show 100% busy independent of how much
> work
> +those cores are doing. It is critical to accurately determine how busy
> +a core is hugely important for the following reasons:
> +
> +   * No indication of overload conditions
> +
> +   * User do not know how much real load is on a system meaning resulted
> in
> +     wasted energy as no power management is utilized
> +
> +Tried and failed schemes include calculating the cycles required from
> +the load on the core, in other words the busyness. For example, how
> +many cycles it costs to handle each packet and determining the
> +frequency cost per core. Due to the varying nature of traffic, types of
> +frames and cost in cycles to process, this mechanism becomes complex
> +quickly where a simple scheme is required to solve the problems.
> +
> +For all polling mechanism, the proposed solution focus on how many
> +times empty poll executed instead of calculating how many cycles it
> +cost to handle each packet. The less empty poll number means current
> +core is busy with processing workload, therefore,  the higher frequency
> +is needed. The high empty poll number indicate current core has lots
> +spare time, therefore, we can lower the frequency.
> +
> +2.1 Power state definition:
> +
> +LOW:  the frequency is used for purge mode.
> +
> +MED:  the frequency is used to process modest traffic workload.
> +
> +HIGH: the frequency is used to process busy traffic workload.
> +
> +2.2 There are two phases to establish the power management system:
> +
> +a.Initialization/Training phase. There is no traffic pass-through, the
> +system will test average empty poll numbers  with LOW/MED/HIGH  power
> +state. Those average empty poll numbers will be the baseline for the
> +normal phase. The system will collect all core's counter every 100ms.
> +The Training phase will take 5 seconds.
> +
> +b.Normal phase. When the real traffic pass-though, the system will
> +compare run-time empty poll moving average value with base line then
> +make decision to move to HIGH power state of MED  power state. The
> +system will collect all core's counter every 10ms.
> +
> +``training_flag`` : optional, enable/disable training mode. Default value is 0.
> + If the training_flag is set as 1(true), then the application will
> +start in  training mode and print out the trained threshold values. If
> +the training_flag  is set as 0(false), the application will start in
> +normal mode, and will use  either the default thresholds or those
> +supplied on the command line. The  trained threshold values are
> +specific to the user’s system, may give a better  power profile when
> compared to the default threshold values.
> +
> +``med_threshold`` : optional, sets the empty poll threshold of a
> +modestly busy system state. If this is not supplied, the application
> +will apply the default value of 350000.
> +
> +``high_threshold`` : optional, sets the empty poll threshold of a busy
> +system state. If this is not supplied, the application will apply the
> +default value of 580000.
> +
> +
> +Preparation Work for Settings
> +=============================
> +1. Turn on Speedstep option in BIOS
> +2. Turn on Turbo in BIOS
> +3. Use intel_pstate driver for CPU frequency control 4. modprobe msr
> +
> +sys_min=/sys/devices/system/cpu/cpu{}/cpufreq/cpuinfo_min_freq
> +sys_max=/sys/devices/system/cpu/cpu{}/cpufreq/cpuinfo_max_freq
> +no_turbo_max=$(rdmsr -p 1 0x0CE -f 15:8 -d)00000
> +
> +cur_min=/sys/devices/system/cpu/cpu{}/cpufreq/scaling_min_freq
> +cur_max=/sys/devices/system/cpu/cpu{}/cpufreq/scaling_max_freq
> +
> +
> +Test Case1 : Basic Training mode test based on one NIC with l3fwd-power
> +=========================================================
> ==============
> +Step 1. Bind One NIC to DPDK driver, launch l3fwd-power with empty-poll
> +enabled
> +
> +    ./l3fwd-power -l 1-2 -n 4 -- -p 0x1 -P --config="(0,0,2)"
> + --empty-poll="1,0,0" -l 10 -m 6 -h 1
> +
> +Step 2. Check the log also when changing the inject packet rate as following:
> +
> +    Injected Rate(64B, dst_ip=1.1.1.1): 10G -> 0.1G -> 10G -> 0.1G -> 10G ->
> +    0.1G The frequency will be set to MED when we inject 0.1G and return to
> HGH
> +    when inject 10G Rate, check the frequency of the forwarding core(core 2)
> +    When traffic is 10G:  cur_min=cur_max=no_turbo_max
> +    When traffic is 0.1G: cur_min=cur_max=[no_turbo_max-500000]
> +
> +
> +Test Case2: No-Training mode test based on one NIC with l3fwd-power
> +=========================================================
> ==========
> +Step 1. Bind One NIC to DPDK driver, launch l3fwd-power with empty-poll
> +enabled
> +
> +   ./l3fwd-power -l 1-2 -n 4  -- -p 0x1 -P --config="(0,0,2)"
> + --empty-poll="0,350000,500000" -l 10 -m 6 -h 1
> +
> +Step 2. Check no training steps are executed in sample's launch log.
> --
> 2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan
  2019-10-21  6:11 ` [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan yufengmx
  2019-10-21  7:45   ` Yao, Lei A
@ 2019-10-23 10:03   ` Tu, Lijuan
  1 sibling, 0 replies; 9+ messages in thread
From: Tu, Lijuan @ 2019-10-23 10:03 UTC (permalink / raw)
  To: Mo, YufengX, dts, Yao, Lei A; +Cc: Mo, YufengX

Please fix warning , thanks

Applying: test_plans/power_empty_poll: upload test plan
.git/rebase-apply/patch:121: trailing whitespace.
sys_max=/sys/devices/system/cpu/cpu{}/cpufreq/cpuinfo_max_freq
warning: 1 line adds whitespace errors

> -----Original Message-----
> From: dts [mailto:dts-bounces@dpdk.org] On Behalf Of yufengmx
> Sent: Monday, October 21, 2019 2:11 PM
> To: dts@dpdk.org; Yao, Lei A <lei.a.yao@intel.com>
> Cc: Mo, YufengX <yufengx.mo@intel.com>
> Subject: [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan
> 
> 
> For packet processing workloads such as DPDK polling is continuous.
> This means CPU cores always show 100% busy independent of how much
> work those cores are doing. It is critical to accurately determine how busy a
> core is hugely important for the following reasons.
> 
> Signed-off-by: yufengmx <yufengx.mo@intel.com>
> ---
>  test_plans/power_empty_poll_test_plan.rst | 138
> ++++++++++++++++++++++
>  1 file changed, 138 insertions(+)
>  create mode 100644 test_plans/power_empty_poll_test_plan.rst
> 
> diff --git a/test_plans/power_empty_poll_test_plan.rst
> b/test_plans/power_empty_poll_test_plan.rst
> new file mode 100644
> index 0000000..01c0ebd
> --- /dev/null
> +++ b/test_plans/power_empty_poll_test_plan.rst
> @@ -0,0 +1,138 @@
> +.. Copyright (c) <2010-2019>, Intel Corporation
> +   All rights reserved.
> +
> +   Redistribution and use in source and binary forms, with or without
> +   modification, are permitted provided that the following conditions
> +   are met:
> +
> +   - Redistributions of source code must retain the above copyright
> +     notice, this list of conditions and the following disclaimer.
> +
> +   - Redistributions in binary form must reproduce the above copyright
> +     notice, this list of conditions and the following disclaimer in
> +     the documentation and/or other materials provided with the
> +     distribution.
> +
> +   - Neither the name of Intel Corporation nor the names of its
> +     contributors may be used to endorse or promote products derived
> +     from this software without specific prior written permission.
> +
> +   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> CONTRIBUTORS
> +   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
> +   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> FITNESS
> +   FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> +   COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> INDIRECT,
> +   INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> +   (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
> GOODS OR
> +   SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
> +   HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
> CONTRACT,
> +   STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
> +   ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
> ADVISED
> +   OF THE POSSIBILITY OF SUCH DAMAGE.
> +
> +=========================
> +Power Lib Empty Poll Test
> +=========================
> +
> +Inband Policy Control
> +=====================
> +
> +For packet processing workloads such as DPDK polling is continuous.
> +This means CPU cores always show 100% busy independent of how much
> work
> +those cores are doing. It is critical to accurately determine how busy
> +a core is hugely important for the following reasons:
> +
> +   * No indication of overload conditions
> +
> +   * User do not know how much real load is on a system meaning resulted
> in
> +     wasted energy as no power management is utilized
> +
> +Tried and failed schemes include calculating the cycles required from
> +the load on the core, in other words the busyness. For example, how
> +many cycles it costs to handle each packet and determining the
> +frequency cost per core. Due to the varying nature of traffic, types of
> +frames and cost in cycles to process, this mechanism becomes complex
> +quickly where a simple scheme is required to solve the problems.
> +
> +For all polling mechanism, the proposed solution focus on how many
> +times empty poll executed instead of calculating how many cycles it
> +cost to handle each packet. The less empty poll number means current
> +core is busy with processing workload, therefore,  the higher frequency
> +is needed. The high empty poll number indicate current core has lots
> +spare time, therefore, we can lower the frequency.
> +
> +2.1 Power state definition:
> +
> +LOW:  the frequency is used for purge mode.
> +
> +MED:  the frequency is used to process modest traffic workload.
> +
> +HIGH: the frequency is used to process busy traffic workload.
> +
> +2.2 There are two phases to establish the power management system:
> +
> +a.Initialization/Training phase. There is no traffic pass-through, the
> +system will test average empty poll numbers  with LOW/MED/HIGH  power
> +state. Those average empty poll numbers will be the baseline for the
> +normal phase. The system will collect all core's counter every 100ms.
> +The Training phase will take 5 seconds.
> +
> +b.Normal phase. When the real traffic pass-though, the system will
> +compare run-time empty poll moving average value with base line then
> +make decision to move to HIGH power state of MED  power state. The
> +system will collect all core's counter every 10ms.
> +
> +``training_flag`` : optional, enable/disable training mode. Default value is 0.
> + If the training_flag is set as 1(true), then the application will
> +start in  training mode and print out the trained threshold values. If
> +the training_flag  is set as 0(false), the application will start in
> +normal mode, and will use  either the default thresholds or those
> +supplied on the command line. The  trained threshold values are
> +specific to the user’s system, may give a better  power profile when
> compared to the default threshold values.
> +
> +``med_threshold`` : optional, sets the empty poll threshold of a
> +modestly busy system state. If this is not supplied, the application
> +will apply the default value of 350000.
> +
> +``high_threshold`` : optional, sets the empty poll threshold of a busy
> +system state. If this is not supplied, the application will apply the
> +default value of 580000.
> +
> +
> +Preparation Work for Settings
> +=============================
> +1. Turn on Speedstep option in BIOS
> +2. Turn on Turbo in BIOS
> +3. Use intel_pstate driver for CPU frequency control 4. modprobe msr
> +
> +sys_min=/sys/devices/system/cpu/cpu{}/cpufreq/cpuinfo_min_freq
> +sys_max=/sys/devices/system/cpu/cpu{}/cpufreq/cpuinfo_max_freq
> +no_turbo_max=$(rdmsr -p 1 0x0CE -f 15:8 -d)00000
> +
> +cur_min=/sys/devices/system/cpu/cpu{}/cpufreq/scaling_min_freq
> +cur_max=/sys/devices/system/cpu/cpu{}/cpufreq/scaling_max_freq
> +
> +
> +Test Case1 : Basic Training mode test based on one NIC with l3fwd-power
> +===============================================================
> ========
> +Step 1. Bind One NIC to DPDK driver, launch l3fwd-power with empty-poll
> +enabled
> +
> +    ./l3fwd-power -l 1-2 -n 4 -- -p 0x1 -P --config="(0,0,2)"
> + --empty-poll="1,0,0" -l 10 -m 6 -h 1
> +
> +Step 2. Check the log also when changing the inject packet rate as following:
> +
> +    Injected Rate(64B, dst_ip=1.1.1.1): 10G -> 0.1G -> 10G -> 0.1G -> 10G ->
> +    0.1G The frequency will be set to MED when we inject 0.1G and return to
> HGH
> +    when inject 10G Rate, check the frequency of the forwarding core(core 2)
> +    When traffic is 10G:  cur_min=cur_max=no_turbo_max
> +    When traffic is 0.1G: cur_min=cur_max=[no_turbo_max-500000]
> +
> +
> +Test Case2: No-Training mode test based on one NIC with l3fwd-power
> +===============================================================
> ====
> +Step 1. Bind One NIC to DPDK driver, launch l3fwd-power with empty-poll
> +enabled
> +
> +   ./l3fwd-power -l 1-2 -n 4  -- -p 0x1 -P --config="(0,0,2)"
> + --empty-poll="0,350000,500000" -l 10 -m 6 -h 1
> +
> +Step 2. Check no training steps are executed in sample's launch log.
> --
> 2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index
  2019-10-21  6:11 ` [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index yufengmx
  2019-10-21  7:45   ` Yao, Lei A
@ 2019-10-23 10:03   ` Tu, Lijuan
  1 sibling, 0 replies; 9+ messages in thread
From: Tu, Lijuan @ 2019-10-23 10:03 UTC (permalink / raw)
  To: Mo, YufengX, dts, Yao, Lei A; +Cc: Mo, YufengX

Code base changed, please rework

> -----Original Message-----
> From: dts [mailto:dts-bounces@dpdk.org] On Behalf Of yufengmx
> Sent: Monday, October 21, 2019 2:11 PM
> To: dts@dpdk.org; Yao, Lei A <lei.a.yao@intel.com>
> Cc: Mo, YufengX <yufengx.mo@intel.com>
> Subject: [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan
> index
> 
> 
> For packet processing workloads such as DPDK polling is continuous.
> This means CPU cores always show 100% busy independent of how much
> work those cores are doing. It is critical to accurately determine how busy a
> core is hugely important for the following reasons.
> 
> Signed-off-by: yufengmx <yufengx.mo@intel.com>
> ---
>  test_plans/index.rst | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/test_plans/index.rst b/test_plans/index.rst index
> a10d171..a8a9543 100644
> --- a/test_plans/index.rst
> +++ b/test_plans/index.rst
> @@ -154,6 +154,7 @@ The following are the test plans for the DPDK DTS
> automated test system.
>      vhost_qemu_mtu_test_plan
>      vhost_user_live_migration_test_plan
>      vm_power_manager_test_plan
> +    power_empty_poll_test_plan
>      vmdq_test_plan
>      vf_l3fwd_test_plan
>      softnic_test_plan
> --
> 2.21.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2019-10-23 10:03 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-10-21  6:11 [dts] [PATCH V1 0/3] power_empty_poll: upload automation script and test plan yufengmx
2019-10-21  6:11 ` [dts] [PATCH V1 1/3] tests/power_empty_poll: upload automation script yufengmx
2019-10-21  7:44   ` Yao, Lei A
2019-10-21  6:11 ` [dts] [PATCH V1 2/3] test_plans/power_empty_poll: add test plan index yufengmx
2019-10-21  7:45   ` Yao, Lei A
2019-10-23 10:03   ` Tu, Lijuan
2019-10-21  6:11 ` [dts] [PATCH V1 3/3] test_plans/power_empty_poll: upload test plan yufengmx
2019-10-21  7:45   ` Yao, Lei A
2019-10-23 10:03   ` Tu, Lijuan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).