1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2010-2014 Intel Corporation.
4 VM Power Management Application
5 ===============================
10 Applications running in Virtual Environments have an abstract view of
11 the underlying hardware on the Host, in particular applications cannot see
12 the binding of virtual to physical hardware.
13 When looking at CPU resourcing, the pinning of Virtual CPUs(vCPUs) to
14 Host Physical CPUs(pCPUS) is not apparent to an application
15 and this pinning may change over time.
16 Furthermore, Operating Systems on virtual machines do not have the ability
17 to govern their own power policy; the Machine Specific Registers (MSRs)
18 for enabling P-State transitions are not exposed to Operating Systems
19 running on Virtual Machines(VMs).
21 The Virtual Machine Power Management solution shows an example of
22 how a DPDK application can indicate its processing requirements using VM local
23 only information(vCPU/lcore, etc.) to a Host based Monitor which is responsible
24 for accepting requests for frequency changes for a vCPU, translating the vCPU
25 to a pCPU via libvirt and affecting the change in frequency.
27 The solution is comprised of two high-level components:
29 #. Example Host Application
31 Using a Command Line Interface(CLI) for VM->Host communication channel management
32 allows adding channels to the Monitor, setting and querying the vCPU to pCPU pinning,
33 inspecting and manually changing the frequency for each CPU.
34 The CLI runs on a single lcore while the thread responsible for managing
35 VM requests runs on a second lcore.
37 VM requests arriving on a channel for frequency changes are passed
38 to the librte_power ACPI cpufreq sysfs based library.
39 The Host Application relies on both qemu-kvm and libvirt to function.
41 This monitoring application is responsible for:
43 - Accepting requests from client applications: Client applications can
44 request frequency changes for a vCPU, translating
45 the vCPU to a pCPU via libvirt and affecting the change in frequency.
47 - Accepting policies from client applications: Client application can
48 send a policy to the host application. The
49 host application will then apply the rules of the policy independent
50 of the application. For example, the policy can contain time-of-day
51 information for busy/quiet periods, and the host application can scale
52 up/down the relevant cores when required. See the details of the guest
53 application below for more information on setting the policy values.
55 - Out-of-band monitoring of workloads via cores hardware event counters:
56 The host application can manage power for an application in a virtualised
57 OR non-virtualised environment by looking at the event counters of the
58 cores and taking action based on the branch hit/miss ratio. See the host
59 application '--core-list' command line parameter below.
61 #. librte_power for Virtual Machines
63 Using an alternate implementation for the librte_power API, requests for
64 frequency changes are forwarded to the host monitor rather than
65 the APCI cpufreq sysfs interface used on the host.
67 The l3fwd-power application will use this implementation when deployed on a VM
68 (see :doc:`l3_forward_power_man`).
70 .. _figure_vm_power_mgr_highlevel:
72 .. figure:: img/vm_power_mgr_highlevel.*
80 VM Power Management employs qemu-kvm to provide communications channels
81 between the host and VMs in the form of Virtio-Serial which appears as
82 a paravirtualized serial device on a VM and can be configured to use
83 various backends on the host. For this example each Virtio-Serial endpoint
84 on the host is configured as AF_UNIX file socket, supporting poll/select
85 and epoll for event notification.
86 In this example each channel endpoint on the host is monitored via
87 epoll for EPOLLIN events.
88 Each channel is specified as qemu-kvm arguments or as libvirt XML for each VM,
89 where each VM can have a number of channels up to a maximum of 64 per VM,
90 in this example each DPDK lcore on a VM has exclusive access to a channel.
92 To enable frequency changes from within a VM, a request via the librte_power interface
93 is forwarded via Virtio-Serial to the host, each request contains the vCPU
94 and power command(scale up/down/min/max).
95 The API for host and guest librte_power is consistent across environments,
96 with the selection of VM or Host Implementation determined at automatically
97 at runtime based on the environment.
99 Upon receiving a request, the host translates the vCPU to a pCPU via
100 the libvirt API before forwarding to the host librte_power.
102 .. _figure_vm_power_mgr_vm_request_seq:
104 .. figure:: img/vm_power_mgr_vm_request_seq.*
106 VM request to scale frequency
109 Performance Considerations
110 ~~~~~~~~~~~~~~~~~~~~~~~~~~
112 While Haswell Microarchitecture allows for independent power control for each core,
113 earlier Microarchtectures do not offer such fine grained control.
114 When deployed on pre-Haswell platforms greater care must be taken in selecting
115 which cores are assigned to a VM, for instance a core will not scale down
116 until its sibling is similarly scaled.
124 Enhanced Intel SpeedStepĀ® Technology must be enabled in the platform BIOS
125 if the power management feature of DPDK is to be used.
126 Otherwise, the sys file folder /sys/devices/system/cpu/cpu0/cpufreq will not exist,
127 and the CPU frequency-based power management cannot be used.
128 Consult the relevant BIOS documentation to determine how these settings
131 Host Operating System
132 ~~~~~~~~~~~~~~~~~~~~~
134 The Host OS must also have the *apci_cpufreq* module installed, in some cases
135 the *intel_pstate* driver may be the default Power Management environment.
136 To enable *acpi_cpufreq* and disable *intel_pstate*, add the following
137 to the grub Linux command line:
139 .. code-block:: console
143 Upon rebooting, load the *acpi_cpufreq* module:
145 .. code-block:: console
147 modprobe acpi_cpufreq
149 Hypervisor Channel Configuration
150 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
152 Virtio-Serial channels are configured via libvirt XML:
157 <name>{vm_name}</name>
158 <controller type='virtio-serial' index='0'>
159 <address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x0'/>
161 <channel type='unix'>
162 <source mode='bind' path='/tmp/powermonitor/{vm_name}.{channel_num}'/>
163 <target type='virtio' name='virtio.serial.port.poweragent.{vm_channel_num}'/>
164 <address type='virtio-serial' controller='0' bus='0' port='{N}'/>
168 Where a single controller of type *virtio-serial* is created and up to 32 channels
169 can be associated with a single controller and multiple controllers can be specified.
170 The convention is to use the name of the VM in the host path *{vm_name}* and
171 to increment *{channel_num}* for each channel, likewise the port value *{N}*
172 must be incremented for each channel.
174 Each channel on the host will appear in *path*, the directory */tmp/powermonitor/*
175 must first be created and given qemu permissions
177 .. code-block:: console
179 mkdir /tmp/powermonitor/
180 chown qemu:qemu /tmp/powermonitor
182 Note that files and directories within /tmp are generally removed upon
183 rebooting the host and the above steps may need to be carried out after each reboot.
185 The serial device as it appears on a VM is configured with the *target* element attribute *name*
186 and must be in the form of *virtio.serial.port.poweragent.{vm_channel_num}*,
187 where *vm_channel_num* is typically the lcore channel to be used in DPDK VM applications.
189 Each channel on a VM will be present at */dev/virtio-ports/virtio.serial.port.poweragent.{vm_channel_num}*
191 Compiling and Running the Host Application
192 ------------------------------------------
197 For information on compiling DPDK and the sample applications
198 see :doc:`compiling`.
200 The application is located in the ``vm_power_manager`` sub-directory.
202 To build just the ``vm_power_manager`` application using ``make``:
204 .. code-block:: console
206 export RTE_SDK=/path/to/rte_sdk
207 export RTE_TARGET=build
208 cd ${RTE_SDK}/examples/vm_power_manager/
211 The resulting binary will be ${RTE_SDK}/build/examples/vm_power_manager
213 To build just the ``vm_power_manager`` application using ``meson/ninja``:
215 .. code-block:: console
217 export RTE_SDK=/path/to/rte_sdk
222 meson configure -Dexamples=vm_power_manager
225 The resulting binary will be ${RTE_SDK}/build/examples/dpdk-vm_power_manager
230 The application does not have any specific command line options other than *EAL*:
232 .. code-block:: console
234 ./build/vm_power_mgr [EAL options]
236 The application requires exactly two cores to run, one core is dedicated to the CLI,
237 while the other is dedicated to the channel endpoint monitor, for example to run
238 on cores 0 & 1 on a system with 4 memory channels:
240 .. code-block:: console
242 ./build/vm_power_mgr -l 0-1 -n 4
244 After successful initialization the user is presented with VM Power Manager CLI:
246 .. code-block:: console
250 Virtual Machines can now be added to the VM Power Manager:
252 .. code-block:: console
254 vm_power> add_vm {vm_name}
256 When a {vm_name} is specified with the *add_vm* command a lookup is performed
257 with libvirt to ensure that the VM exists, {vm_name} is used as an unique identifier
258 to associate channels with a particular VM and for executing operations on a VM within the CLI.
259 VMs do not have to be running in order to add them.
261 A number of commands can be issued via the CLI in relation to VMs:
263 Remove a Virtual Machine identified by {vm_name} from the VM Power Manager.
265 .. code-block:: console
269 Add communication channels for the specified VM, the virtio channels must be enabled
270 in the VM configuration(qemu/libvirt) and the associated VM must be active.
271 {list} is a comma-separated list of channel numbers to add, using the keyword 'all'
272 will attempt to add all channels for the VM:
274 .. code-block:: console
276 add_channels {vm_name} {list}|all
278 Enable or disable the communication channels in {list}(comma-separated)
279 for the specified VM, alternatively list can be replaced with keyword 'all'.
280 Disabled channels will still receive packets on the host, however the commands
281 they specify will be ignored. Set status to 'enabled' to begin processing requests again:
283 .. code-block:: console
285 set_channel_status {vm_name} {list}|all enabled|disabled
287 Print to the CLI the information on the specified VM, the information
288 lists the number of vCPUS, the pinning to pCPU(s) as a bit mask, along with
289 any communication channels associated with each VM, along with the status of each channel:
291 .. code-block:: console
295 Set the binding of Virtual CPU on VM with name {vm_name} to the Physical CPU mask:
297 .. code-block:: console
299 set_pcpu_mask {vm_name} {vcpu} {pcpu}
301 Set the binding of Virtual CPU on VM to the Physical CPU:
303 .. code-block:: console
305 set_pcpu {vm_name} {vcpu} {pcpu}
307 Manual control and inspection can also be carried in relation CPU frequency scaling:
309 Get the current frequency for each core specified in the mask:
311 .. code-block:: console
313 show_cpu_freq_mask {mask}
315 Set the current frequency for the cores specified in {core_mask} by scaling each up/down/min/max:
317 .. code-block:: console
319 set_cpu_freq {core_mask} up|down|min|max
321 Get the current frequency for the specified core:
323 .. code-block:: console
325 show_cpu_freq {core_num}
327 Set the current frequency for the specified core by scaling up/down/min/max:
329 .. code-block:: console
331 set_cpu_freq {core_num} up|down|min|max
333 There are also some command line parameters for enabling the out-of-band
334 monitoring of branch ratio on cores doing busy polling via PMDs.
336 .. code-block:: console
338 --core-list {list of cores}
340 When this parameter is used, the list of cores specified will monitor the ratio
341 between branch hits and branch misses. A tightly polling PMD thread will have a
342 very low branch ratio, so the core frequency will be scaled down to the minimim
343 allowed value. When packets are received, the code path will alter, causing the
344 branch ratio to increase. When the ratio goes above the ratio threshold, the
345 core frequency will be scaled up to the maximum allowed value.
347 .. code-block:: console
349 --branch-ratio {ratio}
351 The branch ratio is a floating point number that specifies the threshold at which
352 to scale up or down for the given workload. The default branch ratio is 0.01,
353 and will need to be adjusted for different workloads.
360 In addition to the command line interface for host command and a virtio-serial
361 interface for VM power policies, there is also a JSON interface through which
362 power commands and policies can be sent. This functionality adds a dependency
363 on the Jansson library, and the Jansson development package must be installed
364 on the system before the JSON parsing functionality is included in the app.
367 .. code-block:: javascript
369 apt-get install libjansson-dev
371 The command and package name may be different depending on your operating
372 system. It's worth noting that the app will successfully build without this
373 package present, but a warning is shown during compilation, and the JSON
374 parsing functionality will not be present in the app.
376 Sending a command or policy to the power manager application is achieved by
377 simply opening a fifo file, writing a JSON string to that fifo, and closing
380 The fifo is at /tmp/powermonitor/fifo
382 The jason string can be a policy or instruction, and takes the following
385 .. code-block:: javascript
392 The 'packet_type' header can contain one of two values, depending on
393 whether a policy or power command is being sent. The two possible values are
394 "policy" and "instruction", and the expected name-value pairs is different
395 depending on which type is being sent.
397 The pairs are the format of standard JSON name-value pairs. The value type
398 varies between the different name/value pairs, and may be integers, strings,
399 arrays, etc. Examples of policies follow later in this document. The allowed
400 names and value types are as follows:
404 :Description: Name of the VM or Host. Allows the parser to associate the
405 policy with the relevant VM or Host OS.
407 :Values: any valid string
411 .. code-block:: javascript
416 :Pair Name: "command"
417 :Description: The type of packet we're sending to the power manager. We can be
418 creating or destroying a policy, or sending a direct command to adjust
419 the frequency of a core, similar to the command line interface.
423 :CREATE: used when creating a new policy,
424 :DESTROY: used when removing a policy,
425 :POWER: used when sending an immediate command, max, min, etc.
429 .. code-block:: javascript
434 :Pair Name: "policy_type"
435 :Description: Type of policy to apply. Please see vm_power_manager documentation
436 for more information on the types of policies that may be used.
440 :TIME: Time-of-day policy. Frequencies of the relevant cores are
441 scaled up/down depending on busy and quiet hours.
442 :TRAFFIC: This policy takes statistics from the NIC and scales up
443 and down accordingly.
444 :WORKLOAD: This policy looks at how heavily loaded the cores are,
445 and scales up and down accordingly.
446 :BRANCH_RATIO: This out-of-band policy can look at the ratio between
447 branch hits and misses on a core, and is useful for detecting
448 how much packet processing a core is doing.
449 :Required: only for CREATE/DESTROY command
452 .. code-block:: javascript
454 "policy_type", "TIME"
456 :Pair Name: "busy_hours"
457 :Description: The hours of the day in which we scale up the cores for busy
459 :Type: array of integers
460 :Values: array with list of hour numbers, (0-23)
461 :Required: only for TIME policy
464 .. code-block:: javascript
466 "busy_hours":[ 17, 18, 19, 20, 21, 22, 23 ]
468 :Pair Name: "quiet_hours"
469 :Description: The hours of the day in which we scale down the cores for quiet
471 :Type: array of integers
472 :Values: array with list of hour numbers, (0-23)
473 :Required: only for TIME policy
476 .. code-block:: javascript
478 "quiet_hours":[ 2, 3, 4, 5, 6 ]
480 :Pair Name: "avg_packet_thresh"
481 :Description: Threshold below which the frequency will be set to min for
482 the TRAFFIC policy. If the traffic rate is above this and below max, the
483 frequency will be set to medium.
485 :Values: The number of packets below which the TRAFFIC policy applies the
486 minimum frequency, or medium frequency if between avg and max thresholds.
487 :Required: only for TRAFFIC policy
490 .. code-block:: javascript
492 "avg_packet_thresh": 100000
494 :Pair Name: "max_packet_thresh"
495 :Description: Threshold above which the frequency will be set to max for
498 :Values: The number of packets per interval above which the TRAFFIC policy
499 applies the maximum frequency
500 :Required: only for TRAFFIC policy
503 .. code-block:: javascript
505 "max_packet_thresh": 500000
507 :Pair Name: "core_list"
508 :Description: The cores to which to apply the policy.
509 :Type: array of integers
510 :Values: array with list of virtual CPUs.
511 :Required: only policy CREATE/DESTROY
514 .. code-block:: javascript
516 "core_list":[ 10, 11 ]
518 :Pair Name: "workload"
519 :Description: When our policy is of type WORKLOAD, we need to specify how
520 heavy our workload is.
524 :HIGH: For cores running workloads that require high frequencies
525 :MEDIUM: For cores running workloads that require medium frequencies
526 :LOW: For cores running workloads that require low frequencies
527 :Required: only for WORKLOAD policy types
530 .. code-block:: javascript
534 :Pair Name: "mac_list"
535 :Description: When our policy is of type TRAFFIC, we need to specify the
536 MAC addresses that the host needs to monitor
538 :Values: array with a list of mac address strings.
539 :Required: only for TRAFFIC policy types
542 .. code-block:: javascript
544 "mac_list":[ "de:ad:be:ef:01:01", "de:ad:be:ef:01:02" ]
547 :Description: the type of power operation to apply in the command
551 :SCALE_MAX: Scale frequency of this core to maximum
552 :SCALE_MIN: Scale frequency of this core to minimum
553 :SCALE_UP: Scale up frequency of this core
554 :SCALE_DOWN: Scale down frequency of this core
555 :ENABLE_TURBO: Enable Turbo Boost for this core
556 :DISABLE_TURBO: Disable Turbo Boost for this core
557 :Required: only for POWER instruction
560 .. code-block:: javascript
564 :Pair Name: "resource_id"
565 :Description: The core to which to apply the power command.
567 :Values: valid core id for VM or host OS.
568 :Required: only POWER instruction
571 .. code-block:: javascript
578 Profile create example:
580 .. code-block:: javascript
585 "policy_type": "TIME",
586 "busy_hours":[ 17, 18, 19, 20, 21, 22, 23 ],
587 "quiet_hours":[ 2, 3, 4, 5, 6 ],
591 Profile destroy example:
593 .. code-block:: javascript
597 "command": "destroy",
600 Power command example:
602 .. code-block:: javascript
610 To send a JSON string to the Power Manager application, simply paste the
611 example JSON string into a text file and cat it into the fifo:
613 .. code-block:: console
615 cat file.json >/tmp/powermonitor/fifo
617 The console of the Power Manager application should indicate the command that
618 was just received via the fifo.
620 Compiling and Running the Guest Applications
621 --------------------------------------------
623 l3fwd-power is one sample application that can be used with vm_power_manager.
625 A guest CLI is also provided for validating the setup.
627 For both l3fwd-power and guest CLI, the channels for the VM must be monitored by the
628 host application using the *add_channels* command on the host. This typically uses
629 the following commands in the host application:
631 .. code-block:: console
633 vm_power> add_vm vmname
634 vm_power> add_channels vmname all
635 vm_power> set_channel_status vmname all enabled
636 vm_power> show_vm vmname
642 For information on compiling DPDK and the sample applications
643 see :doc:`compiling`.
645 For compiling and running l3fwd-power, see :doc:`l3_forward_power_man`.
647 The application is located in the ``guest_cli`` sub-directory under ``vm_power_manager``.
649 To build just the ``guest_vm_power_manager`` application using ``make``:
651 .. code-block:: console
653 export RTE_SDK=/path/to/rte_sdk
654 export RTE_TARGET=build
655 cd ${RTE_SDK}/examples/vm_power_manager/guest_cli/
658 The resulting binary will be ${RTE_SDK}/build/examples/guest_cli
661 This sample application conditionally links in the Jansson JSON
662 library, so if you are using a multilib or cross compile environment you
663 may need to set the ``PKG_CONFIG_LIBDIR`` environmental variable to point to
664 the relevant pkgconfig folder so that the correct library is linked in.
666 For example, if you are building for a 32-bit target, you could find the
667 correct directory using the following ``find`` command:
669 .. code-block:: console
671 # find /usr -type d -name pkgconfig
672 /usr/lib/i386-linux-gnu/pkgconfig
673 /usr/lib/x86_64-linux-gnu/pkgconfig
677 .. code-block:: console
679 export PKG_CONFIG_LIBDIR=/usr/lib/i386-linux-gnu/pkgconfig
681 You then use the make command as normal, which should find the 32-bit
682 version of the library, if it installed. If not, the application will
683 be built without the JSON interface functionality.
685 To build just the ``vm_power_manager`` application using ``meson/ninja``:
687 .. code-block:: console
689 export RTE_SDK=/path/to/rte_sdk
694 meson configure -Dexamples=vm_power_manager/guest_cli
697 The resulting binary will be ${RTE_SDK}/build/examples/guest_cli
702 The standard *EAL* command line parameters are required:
704 .. code-block:: console
706 ./build/guest_vm_power_mgr [EAL options] -- [guest options]
708 The guest example uses a channel for each lcore enabled. For example,
709 to run on cores 0,1,2,3:
711 .. code-block:: console
713 ./build/guest_vm_power_mgr -l 0-3
715 Optionally, there is a list of command line parameter should the user wish to send a power
716 policy down to the host application. These parameters are as follows:
718 .. code-block:: console
720 --vm-name {name of guest vm}
722 This parameter allows the user to change the Virtual Machine name passed down to the
723 host application via the power policy. The default is "ubuntu2"
725 .. code-block:: console
727 --vcpu-list {list vm cores}
729 A comma-separated list of cores in the VM that the user wants the host application to
730 monitor. The list of cores in any vm starts at zero, and these are mapped to the
731 physical cores by the host application once the policy is passed down.
732 Valid syntax includes individial cores '2,3,4', or a range of cores '2-4', or a
733 combination of both '1,3,5-7'
735 .. code-block:: console
737 --busy-hours {list of busy hours}
739 A comma-separated list of hours within which to set the core frequency to maximum.
740 Valid syntax includes individial hours '2,3,4', or a range of hours '2-4', or a
741 combination of both '1,3,5-7'. Valid hours are 0 to 23.
743 .. code-block:: console
745 --quiet-hours {list of quiet hours}
747 A comma-separated list of hours within which to set the core frequency to minimum.
748 Valid syntax includes individial hours '2,3,4', or a range of hours '2-4', or a
749 combination of both '1,3,5-7'. Valid hours are 0 to 23.
751 .. code-block:: console
753 --policy {policy type}
755 The type of policy. This can be one of the following values:
756 TRAFFIC - based on incoming traffic rates on the NIC.
757 TIME - busy/quiet hours policy.
758 BRANCH_RATIO - uses branch ratio counters to determine core busyness.
759 Not all parameters are needed for all policy types. For example, BRANCH_RATIO
760 only needs the vcpu-list parameter, not any of the hours.
763 After successful initialization the user is presented with VM Power Manager Guest CLI:
765 .. code-block:: console
769 To change the frequency of a lcore, use the set_cpu_freq command.
770 Where {core_num} is the lcore and channel to change frequency by scaling up/down/min/max.
772 .. code-block:: console
774 set_cpu_freq {core_num} up|down|min|max
776 To start the application and configure the power policy, and send it to the host:
778 .. code-block:: console
780 ./build/guest_vm_power_mgr -l 0-3 -n 4 -- --vm-name=ubuntu --policy=BRANCH_RATIO --vcpu-list=2-4
782 Once the VM Power Manager Guest CLI appears, issuing the 'send_policy now' command
783 will send the policy to the host:
785 .. code-block:: console
789 Once the policy is sent to the host, the host application takes over the power monitoring
790 of the specified cores in the policy.