1.. SPDX-License-Identifier: BSD-3-Clause 2 Copyright(c) 2010-2014 Intel Corporation. 3 4L3 Forwarding with Power Management Sample Application 5====================================================== 6 7Introduction 8------------ 9 10The L3 Forwarding with Power Management application is an example of power-aware packet processing using the DPDK. 11The application is based on existing L3 Forwarding sample application, 12with the power management algorithms to control the P-states and 13C-states of the Intel processor via a power management library. 14 15Overview 16-------- 17 18The application demonstrates the use of the Power libraries in the DPDK to implement packet forwarding. 19The initialization and run-time paths are very similar to those of the :doc:`l3_forward`. 20The main difference from the L3 Forwarding sample application is that this application introduces power-aware optimization algorithms 21by leveraging the Power library to control P-state and C-state of processor based on packet load. 22 23The DPDK includes poll-mode drivers to configure Intel NIC devices and their receive (Rx) and transmit (Tx) queues. 24The design principle of this PMD is to access the Rx and Tx descriptors directly without any interrupts to quickly receive, 25process and deliver packets in the user space. 26 27In general, the DPDK executes an endless packet processing loop on dedicated IA cores that include the following steps: 28 29* Retrieve input packets through the PMD to poll Rx queue 30 31* Process each received packet or provide received packets to other processing cores through software queues 32 33* Send pending output packets to Tx queue through the PMD 34 35In this way, the PMD achieves better performance than a traditional interrupt-mode driver, 36at the cost of keeping cores active and running at the highest frequency, 37hence consuming the maximum power all the time. 38However, during the period of processing light network traffic, 39which happens regularly in communication infrastructure systems due to well-known "tidal effect", 40the PMD is still busy waiting for network packets, which wastes a lot of power. 41 42Processor performance states (P-states) are the capability of an Intel processor 43to switch between different supported operating frequencies and voltages. 44If configured correctly, according to system workload, this feature provides power savings. 45CPUFreq is the infrastructure provided by the Linux* kernel to control the processor performance state capability. 46CPUFreq supports a user space governor that enables setting frequency via manipulating the virtual file device from a user space application. 47The Power library in the DPDK provides a set of APIs for manipulating a virtual file device to allow user space application 48to set the CPUFreq governor and set the frequency of specific cores. 49 50This application includes a P-state power management algorithm to generate a frequency hint to be sent to CPUFreq. 51The algorithm uses the number of received and available Rx packets on recent polls to make a heuristic decision to scale frequency up/down. 52Specifically, some thresholds are checked to see whether a specific core running a DPDK polling thread needs to increase frequency 53a step up based on the near to full trend of polled Rx queues. 54Also, it decreases frequency a step if packet processed per loop is far less than the expected threshold 55or the thread's sleeping time exceeds a threshold. 56 57C-States are also known as sleep states. 58They allow software to put an Intel core into a low power idle state from which it is possible to exit via an event, such as an interrupt. 59However, there is a tradeoff between the power consumed in the idle state and the time required to wake up from the idle state (exit latency). 60Therefore, as you go into deeper C-states, the power consumed is lower but the exit latency is increased. Each C-state has a target residency. 61It is essential that when entering into a C-state, the core remains in this C-state for at least as long as the target residency in order 62to fully realize the benefits of entering the C-state. 63CPUIdle is the infrastructure provide by the Linux kernel to control the processor C-state capability. 64Unlike CPUFreq, CPUIdle does not provide a mechanism that allows the application to change C-state. 65It actually has its own heuristic algorithms in kernel space to select target C-state to enter by executing privileged instructions like HLT and MWAIT, 66based on the speculative sleep duration of the core. 67In this application, we introduce a heuristic algorithm that allows packet processing cores to sleep for a short period 68if there is no Rx packet received on recent polls. 69In this way, CPUIdle automatically forces the corresponding cores to enter deeper C-states 70instead of always running to the C0 state waiting for packets. 71But user can set the CPU resume latency to control C-state selection. 72Setting the CPU resume latency to 0 73can limit the CPU just to enter C0-state to improve performance, 74which may increase power consumption of platform. 75 76.. note:: 77 78 To fully demonstrate the power saving capability of using C-states, 79 it is recommended to enable deeper C3 and C6 states in the BIOS during system boot up. 80 81Compiling the Application 82------------------------- 83 84To compile the sample application see :doc:`compiling`. 85 86The application is located in the ``l3fwd-power`` sub-directory. 87 88Running the Application 89----------------------- 90 91The application has a number of command line options: 92 93.. code-block:: console 94 95 ./<build_dir>/examples/dpdk-l3fwd_power [EAL options] -- -p PORTMASK [-P] --config(port,queue,lcore)[,(port,queue,lcore)] [--max-pkt-len PKTLEN] [--no-numa] 96 97where, 98 99* -p PORTMASK: Hexadecimal bitmask of ports to configure 100 101* -P: Sets all ports to promiscuous mode so that packets are accepted regardless of the packet's Ethernet MAC destination address. 102 Without this option, only packets with the Ethernet MAC destination address set to the Ethernet address of the port are accepted. 103 104* -u: optional, sets uncore min/max frequency to minimum value. 105 106* -U: optional, sets uncore min/max frequency to maximum value. 107 108* -i (frequency index): optional, sets uncore frequency to frequency index value, by setting min and max values to be the same. 109 110* --config (port,queue,lcore)[,(port,queue,lcore)]: determines which queues from which ports are mapped to which cores. 111 112* --cpu-resume-latency LATENCY: set CPU resume latency to control C-state selection, 0 : just allow to enter C0-state. 113 114* --max-pkt-len: optional, maximum packet length in decimal (64-9600) 115 116* --no-numa: optional, disables numa awareness 117 118* --telemetry: Telemetry mode. 119 120* --pmd-mgmt: PMD power management mode. 121 122* --max-empty-polls : Number of empty polls to wait before entering sleep state. Applies to --pmd-mgmt mode only. 123 124* --pause-duration: Set the duration of the pause callback (microseconds). Applies to --pmd-mgmt mode only. 125 126* --scale-freq-min: Set minimum frequency for scaling. Applies to --pmd-mgmt mode only. 127 128* --scale-freq-max: Set maximum frequency for scaling. Applies to --pmd-mgmt mode only. 129 130See :doc:`l3_forward` for details. 131The L3fwd-power example reuses the L3fwd command line options. 132 133Explanation 134----------- 135 136The following sections provide some explanation of the sample application code. 137As mentioned in the overview section, 138the initialization and run-time paths are identical to those of the L3 forwarding application. 139The following sections describe aspects that are specific to the L3 Forwarding with Power Management sample application. 140 141Power Library Initialization 142~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 143 144The Power library is initialized in the main routine. 145It changes the P-state governor to userspace for specific cores that are under control. 146The Timer library is also initialized and several timers are created later on, 147responsible for checking if it needs to scale down frequency at run time by checking CPU utilization statistics. 148 149.. note:: 150 151 Only the power management related initialization is shown. 152 153.. literalinclude:: ../../../examples/l3fwd-power/main.c 154 :language: c 155 :start-after: Power library initialized in the main routine. 8< 156 :end-before: >8 End of power library initialization. 157 158Monitoring Loads of Rx Queues 159~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 160 161In general, the polling nature of the DPDK prevents the OS power management subsystem from knowing 162if the network load is actually heavy or light. 163In this sample, sampling network load work is done by monitoring received and 164available descriptors on NIC Rx queues in recent polls. 165Based on the number of returned and available Rx descriptors, 166this example implements algorithms to generate frequency scaling hints and speculative sleep duration, 167and use them to control P-state and C-state of processors via the power management library. 168Frequency (P-state) control and sleep state (C-state) control work individually for each logical core, 169and the combination of them contributes to a power efficient packet processing solution when serving light network loads. 170 171The rte_eth_rx_burst() function and the newly-added rte_eth_rx_queue_count() function are used in the endless packet processing loop 172to return the number of received and available Rx descriptors. 173And those numbers of specific queue are passed to P-state and C-state heuristic algorithms 174to generate hints based on recent network load trends. 175 176.. note:: 177 178 Only power control related code is shown. 179 180.. literalinclude:: ../../../examples/l3fwd-power/main.c 181 :language: c 182 :start-after: Main processing loop. 8< 183 :end-before: >8 End of main processing loop. 184 185P-State Heuristic Algorithm 186~~~~~~~~~~~~~~~~~~~~~~~~~~~ 187 188The power_freq_scaleup_heuristic() function is responsible for generating a frequency hint for the specified logical core 189according to available descriptor number returned from rte_eth_rx_queue_count(). 190On every poll for new packets, the length of available descriptor on an Rx queue is evaluated, 191and the algorithm used for frequency hinting is as follows: 192 193* If the size of available descriptors exceeds 96, the maximum frequency is hinted. 194 195* If the size of available descriptors exceeds 64, a trend counter is incremented by 100. 196 197* If the length of the ring exceeds 32, the trend counter is incremented by 1. 198 199* When the trend counter reached 10000 the frequency hint is changed to the next higher frequency. 200 201.. note:: 202 203 The assumption is that the Rx queue size is 128 and the thresholds specified above 204 must be adjusted accordingly based on actual hardware Rx queue size, 205 which are configured via the rte_eth_rx_queue_setup() function. 206 207In general, a thread needs to poll packets from multiple Rx queues. 208Most likely, different queue have different load, so they would return different frequency hints. 209The algorithm evaluates all the hints and then scales up frequency in an aggressive manner 210by scaling up to highest frequency as long as one Rx queue requires. 211In this way, we can minimize any negative performance impact. 212 213On the other hand, frequency scaling down is controlled in the timer callback function. 214Specifically, if the sleep times of a logical core indicate that it is sleeping more than 25% of the sampling period, 215or if the average packet per iteration is less than expectation, the frequency is decreased by one step. 216 217C-State Heuristic Algorithm 218~~~~~~~~~~~~~~~~~~~~~~~~~~~ 219 220Whenever recent rte_eth_rx_burst() polls return 5 consecutive zero packets, 221an idle counter begins incrementing for each successive zero poll. 222At the same time, the function power_idle_heuristic() is called to generate speculative sleep duration 223in order to force logical to enter deeper sleeping C-state. 224There is no way to control C- state directly, and the CPUIdle subsystem in OS is intelligent enough 225to select C-state to enter based on actual sleep period time of giving logical core. 226The algorithm has the following sleeping behavior depending on the idle counter: 227 228* If idle count less than 100, the counter value is used as a microsecond sleep value through rte_delay_us() 229 which execute pause instructions to avoid costly context switch but saving power at the same time. 230 231* If idle count is between 100 and 999, a fixed sleep interval of 100 μs is used. 232 A 100 μs sleep interval allows the core to enter the C1 state while keeping a fast response time in case new traffic arrives. 233 234* If idle count is greater than 1000, a fixed sleep value of 1 ms is used until the next timer expiration is used. 235 This allows the core to enter the C3/C6 states. 236 237.. note:: 238 239 The thresholds specified above need to be adjusted for different Intel processors and traffic profiles. 240 241If a thread polls multiple Rx queues and different queue returns different sleep duration values, 242the algorithm controls the sleep time in a conservative manner by sleeping for the least possible time 243in order to avoid a potential performance impact. 244 245Telemetry Mode 246-------------- 247 248The telemetry mode support for ``l3fwd-power`` is a standalone mode, in this mode 249``l3fwd-power`` does simple l3fwding along with calculating empty polls, full polls, 250and busy percentage for each forwarding core. The aggregation of these 251values of all cores is reported as application level telemetry to metric 252library for every 500ms from the main core. 253 254The busy percentage is calculated by recording the poll_count 255and when the count reaches a defined value the total 256cycles it took is measured and compared with minimum and maximum 257reference cycles and accordingly busy rate is set to either 0% or 25850% or 100%. 259 260.. code-block:: console 261 262 ./<build_dir>/examples/dpdk-l3fwd-power --telemetry -l 1-3 -- -p 0x0f --config="(0,0,2),(0,1,3)" --telemetry 263 264The new stats ``empty_poll`` , ``full_poll`` and ``busy_percent`` can be viewed by running the script 265``/usertools/dpdk-telemetry-client.py`` and selecting the menu option ``Send for global Metrics``. 266 267PMD power management Mode 268------------------------- 269 270The PMD power management mode support for ``l3fwd-power`` is a standalone mode. 271In this mode, ``l3fwd-power`` does simple l3fwding 272along with enabling the power saving scheme on specific port/queue/lcore. 273Main purpose for this mode is to demonstrate 274how to use the PMD power management API. 275 276.. code-block:: console 277 278 ./build/examples/dpdk-l3fwd-power -l 1-3 -- --pmd-mgmt -p 0x0f --config="(0,0,2),(0,1,3)" 279 280PMD Power Management Mode 281------------------------- 282 283There is also a traffic-aware operating mode that, 284instead of using explicit power management, 285will use automatic PMD power management. 286This mode is limited to one queue per core, 287and has three available power management schemes: 288 289``baseline`` 290 This mode will not enable any power saving features. 291 292``monitor`` 293 This will use ``rte_power_monitor()`` function to enter 294 a power-optimized state (subject to platform support). 295 296``pause`` 297 This will use ``rte_power_pause()`` or ``rte_pause()`` 298 to avoid busy looping when there is no traffic. 299 300``scale`` 301 This will use frequency scaling routines 302 available in the ``librte_power`` library. 303 The reaction time of the scale mode is longer 304 than the pause and monitor mode. 305 306See :doc:`Power Management<../prog_guide/power_man>` chapter 307in the DPDK Programmer's Guide for more details on PMD power management. 308 309.. code-block:: console 310 311 ./<build_dir>/examples/dpdk-l3fwd-power -l 1-3 -- -p 0x0f --config="(0,0,2),(0,1,3)" --pmd-mgmt=scale 312 313Setting Uncore Values 314--------------------- 315 316Uncore frequency can be adjusted through manipulating related sysfs entries 317to adjust the minimum and maximum uncore values. 318This will be set for each package and die on the SKU. 319The driver for enabling this is available from kernel version 5.6 and above. 320Three options are available for setting uncore frequency: 321 322``-u`` 323 This will set uncore minimum and maximum frequencies to minimum possible value. 324 325``-U`` 326 This will set uncore minimum and maximum frequencies to maximum possible value. 327 328``-i`` 329 This will allow you to set the specific uncore frequency index that you want, 330 by setting the uncore frequency to a frequency pointed by index. 331 Frequency index's are set 100MHz apart from maximum to minimum. 332 Frequency index values are in descending order, 333 i.e., index 0 is maximum frequency index. 334 335.. code-block:: console 336 337 dpdk-l3fwd-power -l 1-3 -- -p 0x0f --config="(0,0,2),(0,1,3)" -i 1 338