WO2013127151A1 - Power consumption capping control method, device and system - Google Patents

Power consumption capping control method, device and system Download PDF

Info

Publication number
WO2013127151A1
WO2013127151A1 PCT/CN2012/079107 CN2012079107W WO2013127151A1 WO 2013127151 A1 WO2013127151 A1 WO 2013127151A1 CN 2012079107 W CN2012079107 W CN 2012079107W WO 2013127151 A1 WO2013127151 A1 WO 2013127151A1
Authority
WO
WIPO (PCT)
Prior art keywords
power consumption
server
capping
value
power
Prior art date
Application number
PCT/CN2012/079107
Other languages
French (fr)
Chinese (zh)
Inventor
王江涛
李延松
梁伟宁
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2013127151A1 publication Critical patent/WO2013127151A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00Error detection; Error correction; Monitoring
    • G06F11/30Monitoring
    • G06F11/3058Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations
    • G06F11/3062Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations where the monitored property is the power consumption
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/26Power supply means, e.g. regulation thereof
    • G06F1/32Means for saving power
    • G06F1/3203Power management, i.e. event-based initiation of a power-saving mode
    • G06F1/3206Monitoring of events, devices or parameters that trigger a change in power modality

Definitions

  • the present invention relates to communication technologies, and in particular, to a power consumption capping control method, device and system. Background technique
  • IDC Internet Data Center
  • FIG. 1 is a schematic diagram of a resource usage situation in the prior art. As shown in FIG. 1 , by analyzing resource usage of 5000 servers on the existing network, the idle rate of server resources is more than 50%, and if When the 50% resource caused by power consumption is idle, the idle rate of the entire machine is more than 75%.
  • the power capping technology can solve the above problems.
  • the power capping scheme in the prior art includes three main parts: setting a capping value, monitoring running power consumption, and performing a capping action.
  • the power cap value of each rack server is a fixed value.
  • the out-of-band management system monitors the power consumption of the whole machine. If the power consumption exceeds the capping value, the capping action is performed.
  • resource allocation and use between servers are independent of each other, and the capping function is not flexible, and resources cannot be used reasonably. Summary of the invention
  • the embodiment of the invention provides a control method, device and system for power consumption capping, which are used to solve the problem that the capping function existing in the prior art is inflexible and the resource cannot be used reasonably.
  • An aspect of the embodiments of the present invention provides a power consumption capping control method, including: monitoring a power consumption of a whole frame, and obtaining a whole frame monitoring result, where the monitoring result of the whole frame is a resource pool overflow or a resource pool does not overflow.
  • the resource pool is a power supply resource shared by multiple servers inserted in the same chassis; when the monitoring result of the entire frame is that the resource pool does not overflow, power consumption is not capped for each server; When the resource pool overflows, the power capping control command is sent to each server, so that the servers receive the power capping control command and then perform power capping control.
  • a further aspect of the embodiments of the present invention provides a power consumption capping control device, including: a monitoring module, configured to monitor a power consumption of a whole frame, and obtain a whole frame monitoring result, where the whole frame monitoring result is a resource pool overflow Or the resource pool does not overflow, and the resource pool is a power supply resource shared by multiple servers inserted in the same chassis;
  • the first top control module is configured to: when the monitoring result of the whole frame is that the resource pool does not overflow, do not perform power capping on each server; when the monitoring result of the whole frame is a resource pool overflow, send power consumption capping to each server. And controlling the instructions to enable the each server to perform power consumption capping control after receiving the power capping control command.
  • a power consumption capping control system including a management board, a power supply unit, and a plurality of blade servers, wherein the management board includes the power consumption capping control device, and the blade server includes a single Board Management Control Unit, Basic Input Output System BIOS and Central Processing Unit CPU.
  • the technical effect of the embodiment of the present invention is: by monitoring the power consumption of the entire frame, when the obtained monitoring result of the whole frame is that the resource pool does not overflow, the power consumption capping operation is not performed on each server, and the obtained whole frame monitoring result is obtained.
  • the power capping control command is sent to each server, so that the servers receive the power capping control command and then perform power capping control.
  • FIG. 1 is a schematic diagram of a curve of resource usage in the prior art
  • FIG. 2 is a flowchart of Embodiment 1 of a method for controlling power capping according to the present invention
  • FIG. 3 is a flowchart of Embodiment 2 of a method for controlling power capping according to the present invention
  • FIG. 4 is a schematic diagram of a system architecture in a second embodiment of a power consumption capping control method according to the present invention
  • FIG. 5 is a schematic diagram of a server power consumption change in a second embodiment of a power capping control method according to the present invention
  • FIG. 6 is a schematic diagram of a power consumption change of a whole frame in Embodiment 2 of a method for controlling power capping according to the present invention
  • Embodiment 7 is a schematic structural diagram of Embodiment 1 of a power consumption capping control device according to the present invention.
  • FIG. 8 is a schematic structural diagram of Embodiment 2 of a power capping control device according to the present invention. detailed description
  • Step 201 Monitor the power consumption of the entire frame to obtain the entire frame monitoring result.
  • the control method of the power consumption capping provided in this embodiment is mainly directed to the power consumption capping process of the blade server or the multi-node server with the management board, that is, the server in this embodiment may be specifically a blade server or a multi-node server with a management board.
  • the management board provides management functions for each server and chassis, and is responsible for the power consumption of the chassis.
  • the management board When the management board starts to perform power capping control, that is, when the power capping level switch is started, the management board periodically monitors the power consumption of the entire frame, and obtains the entire frame monitoring result in real time.
  • the power consumption of the whole frame monitored by the management board is the power consumption of all the servers and components in the running state of the entire chassis. As the server operates and the external environment changes continuously, the power consumption of the entire frame is also constantly changing. Therefore, in this embodiment, the current value of the entire frame power consumption can be obtained in real time by monitoring.
  • the monitoring result of the entire frame may be the utilization of the resource pool. For example, if the resource pool overflows, the obtained whole frame monitoring result may be a resource pool overflow or the resource pool does not overflow.
  • the resource pool here is a power resource shared by multiple servers inserted in the same chassis.
  • multiple servers inserted in the same chassis share one resource pool, that is, multiple servers share power modules.
  • the management board does not perform power consumption capping operation on each server, but continues to execute the above-mentioned full frame power consumption. Monitoring operation.
  • the power consumption capping operation of the server is not performed, and the server is allowed to run without limit, even if the power consumption of some servers far exceeds the set server power consumption cap value.
  • the server is also not controlled, so that the speed of some servers can be greatly improved without affecting other servers.
  • Step 203 When the whole frame monitoring result is that the resource pool overflows, send a power consumption capping control instruction to each server, so that each server receives the power consumption capping control instruction and performs power consumption capping control.
  • the management board sends a power capping control command to each server, and each server performs power capping control after receiving the power capping control command, which is equivalent to turning on each server.
  • the level capping switch allows the server power consumption of each server after power capping to not exceed the server power capping value.
  • the power capping control in this embodiment is combined with real When monitoring the entire frame power consumption of the entire chassis and the server power consumption of each server, each server shares a resource pool, and resource allocation and use between them are not independent of each other. Therefore, this embodiment is relative to the present The technical capping function is more flexible and can maximize the rational use of resources.
  • This embodiment provides a control method for power capping. By monitoring the power consumption of the entire frame, when the obtained monitoring result of the entire frame is that the resource pool is not overflowed, the power capping operation is not performed on each server. When the resource pool overflows, the power consumption capping control command is sent to each server, so that the servers receive the power consumption capping control command and then perform power capping control.
  • a capping switch is respectively disposed on each server.
  • the capping switch of each server When the management board obtains the resource pool overflow by monitoring the power consumption of the entire frame, the capping switch of each server is turned on, where the capping switch is specifically a power cap. Secondary switch.
  • the board management controller (BMC) unit in the server periodically monitors the server's server power consumption to obtain real-time access. Current server power consumption to the server.
  • the BMC unit in the server is a power consumption capping secondary switch of the server, and the power consumption capping secondary switch is started to start the BMC unit for power capping control.
  • the BMC unit performs power capping control according to the monitored server power consumption and the server power cap value.
  • the server power consumption cap value in this step is obtained by real-time updating of the entire frame power consumption obtained by the management board according to the monitoring, and specifically, in the process of monitoring the power consumption of the entire frame by the management board, the management board obtains the whole according to the monitoring.
  • the box power consumption updates the server power cap value in real time.
  • the BMC unit in the server combines the server power consumption obtained by real-time monitoring with the server power consumption cap value obtained by real-time monitoring when performing power capping control. Therefore, the power capping control process of the embodiment is flexible. , can maximize the rational use of resources. FIG.
  • FIG. 3 is a flowchart of a second embodiment of a method for controlling power capping according to the present invention.
  • the embodiment provides a control method for power capping, which may specifically include the following steps:
  • 4 is a schematic diagram of a system architecture in a second embodiment of a power consumption capping control method according to the present invention. As shown in FIG. 4, it is assumed that N blade servers are inserted in a chassis of the embodiment, that is, a blade server and a blade server. 2, ... blade server N.
  • the management board in the figure is the chassis management board of the blade server.
  • the management board obtains the power consumption of the entire frame of the chassis in real time through the power supply unit (Power Supply Unit; the following package: PSU).
  • the power consumption of the entire frame is inserted in the chassis.
  • the total power consumption of all blade servers and components during operation; the management board also communicates with each blade server through the "Management Communication" channel to deliver the server power cap value of each blade server in real time.
  • the PSU is used to power each blade server and report the power consumption of the entire frame to the management board in real time through the "management signal line".
  • Each of the blade servers may be composed of a BMC unit, a BIOS, a power consumption detecting unit, and a CPU.
  • the BMC unit is an out-of-band management unit of the blade server, and cooperates with the BIOS of the blade server to implement power capping. Consumption of the top switch.
  • the BIOS is used to receive the control commands of the BMC unit, so as to the operating frequency state of the CPU (the performance state; the following cylinder: P-state) and the clock duty state (Throttle state; the following cylinder: T-state), the memory
  • the P-state and T-state and other components are adjusted to implement the capping action of the blade server.
  • the power consumption detection unit is configured to detect the server power consumption of the entire blade server, and report the detection data to the BMC unit in real time. In this step, before the power management capping control is started on the management board, the management board first sets the power consumption capping value of the entire frame according to the machine rejection and service load.
  • the management board can be specifically based on the machine rejection power supply requirement and the blade server is running normally.
  • the actual power consumption and service pressure requirements are used to configure the power consumption capping value of the entire frame, that is, the machine power rejection requirements are met, and the maximum power and average value of the actual power consumption of the blade server can be combined.
  • the service pressure requirement such as the service load situation
  • the entire frame power consumption capping value is configured.
  • the specific configuration method may be a method well known to those skilled in the art in the prior art, and details are not described herein again. It is assumed here that the configured full frame power consumption cap value is P. , ie P. An initial value for the full frame power consumption cap value.
  • Step 302 The management board calculates a server power consumption capping value of each blade server according to the entire frame power consumption capping value.
  • the management board After the management board sets the power consumption cap value of the entire frame, it can be based on the power consumption cap value of the entire frame.
  • the server power capping value is P m .
  • the calculation method of the server power consumption capping value in this embodiment is to use the entire frame power consumption capping value P.
  • the components on the chassis other than the blade server may include, for example, a chassis fan, a power supply, a management board, a switchboard, and the like.
  • Step 303 The management board monitors the power consumption of the entire frame according to the preset whole frame monitoring period, and updates the server power consumption capping value of each server according to the monitored full frame power consumption.
  • the management board After completing the setting of the whole frame power cap value and the server power cap value, the management board can be started to control the power capping. After the management board is enabled to control the power capping process, the management board monitors the power consumption of the entire frame according to the preset frame monitoring period. It can be assumed that the entire frame power consumption period can be based on the actual situation. Specific settings, for example, can be set to monitor the power consumption of the entire frame 10 times per second. At the same time, the management board updates the server power consumption cap value of each server according to the power consumption of the whole frame obtained by the monitoring. Specifically, the management board can refresh the server power consumption cap value of each blade server in real time according to the current frame power consumption of the chassis.
  • the management board may The server power capping value is refreshed every second, that is, the preset time period is 1 second, and the value of 10 full-frame power consumption and the total power consumption of 10 servers can be separately monitored in one second; The average power consumption of the server and the average power consumption in one second, the two average values are subtracted to obtain the total power consumption of all components except the blade server P_other; The box power cap value P. Subtract P_other to get a difference, and then divide the difference into each in-position blade server, that is, get the updated server power capping value P m . Specifically, the following formula can be used.
  • the average power consumption of the entire frame is the average value of the power consumption of the entire frame, which is the average value of the total power consumption of the server, and N is the number of servers.
  • the preset formula is used.
  • the server power consumption cap value of each server is updated.
  • the power consumption of the entire frame is different in different time periods.
  • the power consumption capping value of the computing server is refreshed in real time by monitoring the power consumption of the entire frame, so that the server power consumption is obtained.
  • the capping value can be closer to the current server operation, which can greatly reduce the capping error caused by the inaccuracy of the server power capping value, and improve the accuracy of the capping action in the subsequent steps.
  • Step 304 The management board determines whether the power consumption of the entire frame obtained by the monitoring is less than the product of the preset upper limit coefficient and the entire frame power consumption cap value. If yes, step 305 is performed; otherwise, step 306 is performed.
  • the management board determines whether the power consumption of the entire frame is less than a preset upper limit coefficient and a full frame power consumption cap value P.
  • the product, where the upper limit coefficient of the interval can be set or modified according to the actual situation, for example, set to 0.7, then this step is specifically for the management panel to determine the power consumption of the whole frame obtained by monitoring! ⁇ Is it less than ⁇ . '0.7, if yes, go to step 305, otherwise go to step 306.
  • the product of the interval upper limit coefficient and the entire frame power consumption cap value in this embodiment is the upper threshold of the resource pool shared by the blade server, and the interval lower limit coefficient, the interval lower limit coefficient and the entire frame power consumption cap value are also involved in the subsequent steps.
  • the product is the lower threshold of the resource pool, and the range between the upper threshold and the lower threshold constitutes the anti-shock interval of the resource pool.
  • Step 305 The management board obtains the whole frame monitoring result that the resource pool does not overflow, and returns to the execution step.
  • the power consumption of the entire frame obtained by the management board is less than the accumulation of the preset upper limit coefficient and the entire frame power consumption cap value. If the resource pool is not overflowed, the management board obtains the monitoring result of the entire frame. The management board does not perform any further operations. The management board can perform the above steps 303 and continue to monitor the power consumption of the entire frame. . Step 306, the management board obtains the whole box monitoring result as a resource pool overflow. When the power consumption of the entire frame obtained by the management board is greater than or equal to the preset upper limit coefficient of the interval and the power consumption cap value of the entire frame. If the resource pool is overflowed, the management board obtains the entire box monitoring result as the resource pool overflow, and performs the following step 307.
  • the power capping control of the blade server is turned on.
  • the management board closes the resource pool and sends a power capping control command to each server.
  • Step 308 The BMC unit in the blade server monitors the server power consumption of the blade server, and determines whether the monitored server power consumption is less than the server power consumption cap value. If yes, continue to perform step 308, otherwise step 309 is performed.
  • Step 309 The BMC unit sends a cap execution notification to the basic input/output system (Basic Input Output System; BIOS) of the blade server according to the difference between the server power consumption and the server power consumption cap value.
  • BIOS Basic Input Output System
  • BMC unit When the resulting BMC unit monitoring server power P n is greater than or equal to said power capping step 307 the server the updated value P m, power consumption of the blade server indicating that the server has reached the cap value server power consumption, BMC unit begins execution function The topping action is consumed. Specifically, the BMC sends a cap execution notification to the blade server's BIOS based on the difference between the server power consumption and the server power cap value. Step 310, the BIOS adjusts the central processor of the blade server according to the capping execution notification. (Central Processing Unit; The following cylinders are called: CPU) operating frequency state and clock duty cycle state, as well as the operating frequency state of the memory and the clock duty cycle state.
  • CPU Central Processing Unit
  • the BIOS of the blade server After receiving the capping execution notification, the BIOS of the blade server adjusts the operating frequency state P-state of the CPU of the blade server and the clock duty state T-state according to the capping execution notification, and the memory of the blade server. -state and T-state and the working state of other components of the blade server.
  • Step 311 The BMC unit determines whether the current monitored server power consumption is less than the server power consumption cap value. If yes, step 312 is performed; otherwise, the process returns to step 310.
  • the monitoring operation of the power consumption of the server by the BMC unit is not stopped due to the power capping execution action, that is, in the process of performing the above steps 309-311, the BMC unit still monitors the server power consumption of the blade server at the same time. .
  • the BMC unit compares the current server power consumption P n with the server power cap value P m to determine whether the current server power consumption P n is less than the server power consumption. cap value P m, if yes, step 312 is performed, otherwise step 310 is performed, power capping operation continues.
  • Step 312 The BMC unit sends a capping execution stop notification to the BIOS of the blade server.
  • the BMC power unit P n recognizes server updates the server power less than the current capping value P m, power capping operation has been previously indicated that the power consumption of the blade server in the server power control Below the cap value, the BMC unit sends a cap execution stop notification to the blade server's BIOS.
  • Step 313 The BIOS stops performing the power consumption capping related operation according to the capping execution stop notification. After receiving the notification of the capping stop execution, the BIOS of the blade server stops the execution of the power capping according to the capping stop execution notification.
  • Step 314 The management board determines whether the current frame power consumption obtained by the current monitoring is less than or equal to a product of the preset interval lower limit coefficient and the whole frame power consumption capping value. If yes, step 315 is performed; otherwise, the process returns to step 308.
  • the monitoring operation of the management board on the power consumption of the entire frame is not stopped due to the server power consumption monitoring of the BMC unit and the execution of the power consumption capping action, that is, in the process of performing the above steps 307-313, the management board
  • the entire frame power consumption is still monitored at the same time.
  • the management board determines whether the power consumption of the entire frame obtained by the current monitoring is less than or equal to the preset interval lower limit coefficient and the entire frame power consumption capping value P. The product, if yes, proceeds to step 315, otherwise returns to step 308.
  • the product of the interval lower limit coefficient and the entire frame power consumption capping value in this embodiment is the lower threshold of the resource pool shared by the blade server.
  • the interval lower limit coefficient may be set to 0.6
  • the interval upper limit coefficient and the entire frame power consumption cap value are the upper threshold of the resource pool shared by the blade server, and the range between the upper threshold and the lower threshold constitutes the anti-shock interval of the resource pool.
  • the anti-shock interval in this embodiment can be modified according to actual conditions. If the frequency of opening and closing the resource pool is high, the lower threshold can be appropriately lowered to increase the anti-shock interval.
  • Step 315 The management board starts the resource pool, and sends a power consumption capping stop instruction to each server, so that each server stops the power consumption capping control after receiving the power consumption capping stop instruction.
  • the management board re-opens the resource pool and sends a power capping stop command to each server, so that each server stops the power capping control after receiving the power capping stop command, and completes one round of capping control.
  • the entire frame power consumption capping value P configured in the above step 301 is assumed. For 3500 watts, there are 10 in-plane blade servers in the chassis. The total power consumption of the fan, switch board, management board, power supply, etc. in the chassis is 500 watts, and the server power of each blade server is calculated.
  • the consumption capping value P m is 300 watts.
  • the power consumption capping control method of the power consumption capping method of the present embodiment is used, and as the load is continuously increased, the power consumption variation result in the resource pool as shown in FIG. 5 can be obtained.
  • FIG. 5 and FIG. 6 are respectively schematic diagrams showing changes in server power consumption and overall frame power consumption in the second embodiment of the power capping control method of the present invention. As can be seen from FIG. 5 and FIG. 6, the embodiment can implement resources. Maximize use.
  • the control method of the power consumption capping provided in this embodiment may be applied to a blade server or a node server having a management board, and the status of the power consumption of the blade server is determined by the management board monitoring the running power of the entire frame in real time.
  • the BMC unit of each blade server acts as the secondary switch of the blade server itself, and cooperates with the BIOS to perform the power capping action of the blade server itself, realizes the power capping flexible monitoring design, and limits the rated power consumption of the blade server to the power cap by power capping.
  • the actual power consumption of the blade server can not only avoid the actual operation of the blade server.
  • the waste of resources caused by 50% of rated power consumption can also greatly increase the density of equipment in the equipment room and machine rejection, and increase the utilization rate of space resources.
  • a shared resource pool is constructed by using a two-stage power capping switch, and the power capping is controlled by combining the whole frame power consumption and the server power consumption in the power capping process, and ensuring high in the case that the resource pool does not overflow.
  • the load blade is running excessively, and the service performance is not affected.
  • overflow of the resource pool ensure that the power consumption of the entire frame does not exceed the preset power consumption capping value of the entire frame, thereby maximizing the power supply on the basis of ensuring power supply security.
  • the power supply resource can be utilized to the maximum extent.
  • the anti-shock interval of the resource pool is set by the upper and lower thresholds of the resource pool. The capping switch caused by the fluctuation of the business load is frequently turned on or off, which improves the stability of the power cap.
  • the management board monitors the power consumption of the fan, the power supply, the management board, and the switch board in real time, and adopts a scheme of real-time refreshing the server power consumption cap value of each blade server, thereby minimizing the power consumption capping. error.
  • the aforementioned program can be stored in a computer readable storage medium. When the program is executed, the steps including the foregoing method embodiments are performed; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.
  • FIG. 7 is a schematic structural diagram of Embodiment 1 of a power consumption capping control device according to the present invention.
  • the present embodiment provides a power consumption capping control device, which may specifically perform the steps in the first embodiment of the foregoing method. , will not repeat them here.
  • the power consumption capping control device provided in this embodiment may be specifically a management board, which may specifically include a monitoring module 701 and a first capping control module 702.
  • the monitoring module 701 is configured to monitor the power consumption of the entire frame and obtain the monitoring result of the entire frame.
  • the monitoring result of the whole frame is that the resource pool overflows or the resource pool does not overflow, and the resource pool is inserted in the same chassis. Power resources shared by servers.
  • the first top control module 702 is configured to: when the monitoring result of the whole frame is that the resource pool does not overflow, do not perform power consumption capping on each server; when the monitoring result of the whole frame is a resource pool overflow, send power consumption capping to each server. And controlling the instructions to enable the each server to perform power consumption capping control after receiving the power capping control command.
  • FIG. 8 is a schematic structural diagram of Embodiment 2 of a power consumption capping control device according to the present invention. As shown in FIG. 8 , the present embodiment provides a power consumption capping control device, which may specifically perform the steps in the second embodiment of the foregoing method. , will not repeat them here.
  • the power consumption capping control device provided in this embodiment is as described above On the basis of the data shown in FIG.
  • the monitoring module 701 may specifically include a monitoring unit 711, a first obtaining unit 721, and a second acquiring unit 731.
  • the monitoring unit 711 is configured to monitor the power consumption of the entire frame according to a preset entire frame monitoring period.
  • the first obtaining unit 721 is configured to: when the power consumption of the whole frame obtained by the monitoring is less than a product of a preset upper limit coefficient and a full-frame power cap value, obtain a whole frame monitoring result that the resource pool does not overflow.
  • the second obtaining unit 731 is configured to obtain the whole box monitoring result as a resource pool overflow when the power consumption of the whole frame obtained by the monitoring is greater than or equal to a product of the preset upper limit coefficient and the full frame power capping value.
  • the power capping control device provided in this embodiment may further include a second capping control module 801.
  • the second capping control module 801 is configured to: when the power consumption capping command is sent to each server, when the monitored full frame power consumption is less than or equal to a product of a preset interval lower limit coefficient and the entire frame power capping value And sending, to the servers, a power consumption capping stop instruction, so that each server stops the power consumption capping control after receiving the power consumption capping stop command, wherein the interval upper limit index is greater than the interval lower limit index.
  • the power capping control device provided in this embodiment may further include a first calculating module 802 and an updating module 803.
  • the first calculating module 802 is configured to calculate an average value of the power consumption of the whole frame and a total power consumption of the server according to values of multiple frame power consumptions monitored by the preset time period and values of total server power consumption. average value.
  • the updating module 803 is configured to update the server power consumption cap value of each server according to the average value of the whole frame power consumption, the average value of the total power consumption of the server, and the entire frame power consumption capping value by using the above formula (2). And in the unit of the preset time period, the server power consumption cap value of each server is updated by using the above formula.
  • the power capping control device provided in this embodiment may further include a setting module 804 and a second calculating module 805.
  • the setting module 804 is configured to set the whole frame power consumption capping value according to the machine rejection power distribution and the service load condition before the power consumption of the entire frame is monitored.
  • the second calculating module 805 is configured to calculate the server power consumption of each server by using the above formula (1) according to the whole frame power consumption capping value and the power consumption value of other components except the servers on the chassis.
  • the cap value The embodiment provides a power consumption capping control device. When the power consumption of the entire frame is monitored, when the obtained monitoring result of the entire frame is not overflowed by the resource pool, the power capping operation is not performed on each server. When the resource pool overflows, the power consumption capping control command is sent to each server, so that the servers receive the power consumption capping control command and then perform power capping control.
  • the power consumption capping control system may further include a management board, a power supply unit PSU, and a plurality of blade servers, as shown in FIG. 4 above.
  • the management board may specifically include the power consumption capping control device shown in FIG. 7 or FIG. 8 above.
  • the blade server may specifically include a board management control unit, a basic input/output system BIOS, and a central processing unit CPU.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Quality & Reliability (AREA)
  • Power Sources (AREA)

Abstract

Provided are a power consumption capping control method, device and system. The method comprises: monitoring whole-frame power consumption, obtaining a whole-frame monitoring result, the whole-frame monitoring result being that a resource pool overflows or does not overflow, the resource pool being a power supply resource shared by a plurality of servers inserted onto a same machine frame; when the whole-frame monitoring result is that the resource pool does not overflow, not performing power consumption capping on each server; and when the whole-frame monitoring result is that the resource pool overflows, sending a power consumption capping control instruction to each server to enable same to perform power consumption capping control after receiving the power consumption capping control instruction. Also provided are a power consumption capping control device and system. The embodiments of the present invention achieve more flexible capping functions, reasonably use resources to a maximum extent, and greatly reduce waste of resources.

Description

功耗封顶的控制方法、 设备和系统  Power capping control method, device and system
技术领域 Technical field
本发明涉及通信技术, 尤其涉及一种功耗封顶的控制方法、 设备和系 统。 背景技术  The present invention relates to communication technologies, and in particular, to a power consumption capping control method, device and system. Background technique
随着互联网数据的爆炸式增长和云计算时代的到来, IT领域对服务器 设备的需求不断增长, 数据中心机房的 IT设备的快速扩容, 给数据中心 的供电、 散热、 空间容量等带来巨大挑战。 一方面是数据中心供电与资源 的紧缺, 另一方面是机房设备的用电效率低、 机拒密度低, 因此在很大程 度上存在资源浪费。 目前互联网数据中心( Internet Data Center; 以下筒称: IDC )机房的机拒配电都有限额, 机拒耗电超过额定值时会导致空开跳闸, 在部署服务器时需要非常谨慎, 服务器数量的配置要按照额定最大功耗来 计算, 而在实际使用中服务器运行时的功耗出现接近额定最大功耗的概率 极小。 图 1为现有技术中资源使用情况的曲线示意图, 如图 1所示, 通过 对现网 5000 台服务器的资源使用情况进行分析, 服务器资源的空闲率达 到 50%以上, 若再考虑按照额定最大功耗配电所导致的 50%资源空闲, 则 整个机拒配电的空闲率达到 75%以上。 而功耗封顶技术可以解决目前存在 的上述问题。 现有技术中的功耗封顶方案包括三个主要部分: 设置封顶值、 监视运 行功耗、 执行封顶动作。 即先根据机拒配电要求、 服务器正常运行的实际 功耗、 业务压力需求等设置各台机架服务器的功耗封顶值, 然后将该封顶 值写入带外管理系统作为服务器运行的上限功耗, 该功耗封顶值为一个固 定值。 在服务器运行过程中, 带外管理系统监测整机功耗, 如果发现功耗 超过封顶值, 则执行封顶动作。 然而, 现有技术中服务器之间的资源分配和使用是相互独立的, 封顶 功能不灵活, 不能最大化地合理使用资源。 发明内容 With the explosive growth of Internet data and the advent of the cloud computing era, the demand for server equipment in the IT field is growing, and the rapid expansion of IT equipment in the data center equipment room brings great challenges to the power supply, heat dissipation, and space capacity of the data center. . On the one hand, there is a shortage of power and resources in the data center. On the other hand, the power consumption of the equipment in the equipment room is low, and the density of the machine is low. Therefore, there is a great waste of resources. At present, the Internet Data Center (hereinafter referred to as IDC) has a limit on the machine rejection of the machine. If the machine refuses to exceed the rated value, it will cause an open trip. When deploying the server, you need to be very cautious. The configuration is calculated according to the rated maximum power consumption, and in actual use, the probability of the power consumption of the server running close to the rated maximum power consumption is extremely small. FIG. 1 is a schematic diagram of a resource usage situation in the prior art. As shown in FIG. 1 , by analyzing resource usage of 5000 servers on the existing network, the idle rate of server resources is more than 50%, and if When the 50% resource caused by power consumption is idle, the idle rate of the entire machine is more than 75%. The power capping technology can solve the above problems. The power capping scheme in the prior art includes three main parts: setting a capping value, monitoring running power consumption, and performing a capping action. That is, first set the power cap value of each rack server according to the machine rejection power supply requirement, the actual power consumption of the normal operation of the server, the service pressure demand, etc., and then write the cap value to the out-of-band management system as the upper limit operation of the server operation. Consumption, the power cap value is a fixed value. During the running of the server, the out-of-band management system monitors the power consumption of the whole machine. If the power consumption exceeds the capping value, the capping action is performed. However, in the prior art, resource allocation and use between servers are independent of each other, and the capping function is not flexible, and resources cannot be used reasonably. Summary of the invention
本发明实施例提供一种功耗封顶的控制方法、 设备和系统, 用于解决 现有技术存在着的封顶功能不灵活, 不能最大化地合理使用资源的问题。 本发明实施例的一个方面是提供一种功耗封顶的控制方法, 包括: 对整框功耗进行监控, 获取整框监控结果, 所述整框监控结果为资源 池溢出或者资源池不溢出, 所述资源池为插在同一机框上的多个服务器所 共享的供电资源; 当所述整框监控结果为资源池不溢出时, 不对各服务器进行功耗封 顶; 当所述整框监控结果为资源池溢出时, 向各服务器发送功耗封顶控制 指令, 以使所述各服务器收到所述功耗封顶控制指令后进行功耗封顶控 制。 本发明实施例的又一个方面是提供一种功耗封顶的控制设备, 包括: 监控模块, 用于对整框功耗进行监控, 获取整框监控结果, 所述整框 监控结果为资源池溢出或者资源池不溢出, 所述资源池为插在同一机框上 的多个服务器所共享的供电资源;  The embodiment of the invention provides a control method, device and system for power consumption capping, which are used to solve the problem that the capping function existing in the prior art is inflexible and the resource cannot be used reasonably. An aspect of the embodiments of the present invention provides a power consumption capping control method, including: monitoring a power consumption of a whole frame, and obtaining a whole frame monitoring result, where the monitoring result of the whole frame is a resource pool overflow or a resource pool does not overflow. The resource pool is a power supply resource shared by multiple servers inserted in the same chassis; when the monitoring result of the entire frame is that the resource pool does not overflow, power consumption is not capped for each server; When the resource pool overflows, the power capping control command is sent to each server, so that the servers receive the power capping control command and then perform power capping control. A further aspect of the embodiments of the present invention provides a power consumption capping control device, including: a monitoring module, configured to monitor a power consumption of a whole frame, and obtain a whole frame monitoring result, where the whole frame monitoring result is a resource pool overflow Or the resource pool does not overflow, and the resource pool is a power supply resource shared by multiple servers inserted in the same chassis;
第一封顶控制模块, 用于当所述整框监控结果为资源池不溢出时, 不 对各服务器进行功耗封顶; 当所述整框监控结果为资源池溢出时, 向各服 务器发送功耗封顶控制指令, 以使所述各服务器收到所述功耗封顶控制指 令后进行功耗封顶控制。 本发明实施例的又一个方面是提供一种功耗封顶的控制系统, 包括管 理板、 供电单元和多个刀片服务器, 所述管理板包括上述功耗封顶的控制 设备, 所述刀片服务器包括单板管理控制单元、 基本输入输出系统 BIOS 和中央处理器 CPU。  The first top control module is configured to: when the monitoring result of the whole frame is that the resource pool does not overflow, do not perform power capping on each server; when the monitoring result of the whole frame is a resource pool overflow, send power consumption capping to each server. And controlling the instructions to enable the each server to perform power consumption capping control after receiving the power capping control command. A further aspect of the embodiments of the present invention provides a power consumption capping control system, including a management board, a power supply unit, and a plurality of blade servers, wherein the management board includes the power consumption capping control device, and the blade server includes a single Board Management Control Unit, Basic Input Output System BIOS and Central Processing Unit CPU.
本发明实施例的技术效果是: 通过对整框功耗进行监控, 当获取到的 整框监控结果为资源池未溢出时, 不对各服务器进行功耗封顶操作, 当获 取到的整框监控结果为资源池溢出时, 向各服务器发送功耗封顶控制指 令, 以使所述各服务器收到所述功耗封顶控制指令后进行功耗封顶控制。 本实施例实现了更加灵活的封顶功能, 最大化地合理使用资源, 大大降低 了资源的浪费。 附图说明 The technical effect of the embodiment of the present invention is: by monitoring the power consumption of the entire frame, when the obtained monitoring result of the whole frame is that the resource pool does not overflow, the power consumption capping operation is not performed on each server, and the obtained whole frame monitoring result is obtained. When the resource pool overflows, the power capping control command is sent to each server, so that the servers receive the power capping control command and then perform power capping control. This embodiment implements a more flexible capping function, maximizes the rational use of resources, and greatly reduces the waste of resources. DRAWINGS
为了更清楚地说明本发明实施例或现有技术中的技术方案, 下面将对 实施例或现有技术描述中所需要使用的附图作一筒单地介绍, 显而易见 地, 下面描述中的附图是本发明的一些实施例, 对于本领域普通技术人员 来讲, 在不付出创造性劳动性的前提下, 还可以根据这些附图获得其他的 附图。  In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings to be used in the embodiments or the description of the prior art will be briefly described below, and obviously, the attached in the following description The drawings are some embodiments of the present invention, and other drawings may be obtained from those of ordinary skill in the art without departing from the scope of the invention.
图 1为现有技术中资源使用情况的曲线示意图;  1 is a schematic diagram of a curve of resource usage in the prior art;
图 2为本发明功耗封顶的控制方法实施例一的流程图; 图 3为本发明功耗封顶的控制方法实施例二的流程图;  2 is a flowchart of Embodiment 1 of a method for controlling power capping according to the present invention; FIG. 3 is a flowchart of Embodiment 2 of a method for controlling power capping according to the present invention;
图 4为本发明功耗封顶的控制方法实施例二中的系统架构示意图; 图 5为本发明功耗封顶的控制方法实施例二中的服务器功耗变化示意 图;  4 is a schematic diagram of a system architecture in a second embodiment of a power consumption capping control method according to the present invention; FIG. 5 is a schematic diagram of a server power consumption change in a second embodiment of a power capping control method according to the present invention;
图 6 为本发明功耗封顶的控制方法实施例二中的整框功耗变化示意 图;  6 is a schematic diagram of a power consumption change of a whole frame in Embodiment 2 of a method for controlling power capping according to the present invention;
图 7为本发明功耗封顶的控制设备实施例一的结构示意图;  7 is a schematic structural diagram of Embodiment 1 of a power consumption capping control device according to the present invention;
图 8为本发明功耗封顶的控制设备实施例二的结构示意图。 具体实施方式  FIG. 8 is a schematic structural diagram of Embodiment 2 of a power capping control device according to the present invention. detailed description
为使本发明实施例的目的、 技术方案和优点更加清楚, 下面将结合本发 明实施例中的附图, 对本发明实施例中的技术方案进行清楚、 完整地描述, 显然, 所描述的实施例是本发明一部分实施例, 而不是全部的实施例。 基于 本发明中的实施例, 本领域普通技术人员在没有作出创造性劳动前提下所获 得的所有其他实施例, 都属于本发明保护的范围。  The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is a partial embodiment of the invention, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.
图 2为本发明功耗封顶的控制方法实施例一的流程图, 如图 2所示, 本 实施例提供了一种功耗封顶的控制方法, 本实施例从管理板一侧对本发明的 技术方案进行说明, 本实施例可以具体执行如下步骤: 步骤 201 , 对整框功耗进行监控, 获取整框监控结果。 本实施例提供的功耗封顶的控制方法主要针对刀片服务器或者具有管理 板的多节点服务器的功耗封顶过程, 即本实施例中的服务器可以具体为刀片 服务器或具有管理板的多节点服务器, 由管理板对各服务器和机框提供管理 功能, 担负机框的功耗封顶一级开关。 当管理板开始执行功耗封顶控制, 即 启动功耗封顶一级开关时, 管理板对整框功耗进行周期性的监控, 并实时获 取整框监控结果。 此处管理板监控的整框功耗为整个机框上处于运行状态的 所有服务器和部件的功耗, 其随着服务器的运行情况以及外部环境的不断变 化, 整框功耗也是不断变化的, 因此本实施例通过监控可以实时获取到当前 的整框功耗的值。 在本实施例中, 整框监控结果可以为资源池的利用情况, 如资源池是否溢出, 即获取到的整框监控结果可以为资源池溢出或者资源池 不溢出。 此处的资源池为插在同一机框上的多个服务器所共享的供电资源, 在本实施例中, 插在同一机框上的多个服务器共享一个资源池, 即多个服务 器共享电源模块提供的供电资源。 步骤 202, 当整框监控结果为资源池不溢出时, 不对各服务器进行功耗 封顶。 2 is a flowchart of Embodiment 1 of a method for controlling power capping according to the present invention. As shown in FIG. 2, this embodiment provides a control method for power capping, which is from the side of the management board to the present invention. The technical solution is described. In this embodiment, the following steps may be specifically implemented: Step 201: Monitor the power consumption of the entire frame to obtain the entire frame monitoring result. The control method of the power consumption capping provided in this embodiment is mainly directed to the power consumption capping process of the blade server or the multi-node server with the management board, that is, the server in this embodiment may be specifically a blade server or a multi-node server with a management board. The management board provides management functions for each server and chassis, and is responsible for the power consumption of the chassis. When the management board starts to perform power capping control, that is, when the power capping level switch is started, the management board periodically monitors the power consumption of the entire frame, and obtains the entire frame monitoring result in real time. Here, the power consumption of the whole frame monitored by the management board is the power consumption of all the servers and components in the running state of the entire chassis. As the server operates and the external environment changes continuously, the power consumption of the entire frame is also constantly changing. Therefore, in this embodiment, the current value of the entire frame power consumption can be obtained in real time by monitoring. In this embodiment, the monitoring result of the entire frame may be the utilization of the resource pool. For example, if the resource pool overflows, the obtained whole frame monitoring result may be a resource pool overflow or the resource pool does not overflow. The resource pool here is a power resource shared by multiple servers inserted in the same chassis. In this embodiment, multiple servers inserted in the same chassis share one resource pool, that is, multiple servers share power modules. Power supply resources provided. Step 202: When the whole frame monitoring result is that the resource pool does not overflow, the power consumption capping is not performed on each server.
在本实施例中, 通过上述步骤的监控过程, 当获取到的整框监控结果为 资源池不溢出时, 管理板不会对各服务器进行功耗封顶操作, 而是继续执行 上述整框功耗的监控操作。 在本实施例中, 在各服务器共享的资源池不溢出 的前提下, 不执行服务器的功耗封顶操作, 允许服务器无限制运行, 即使部 分服务器的功耗远超过设定的服务器功耗封顶值也不对该服务器进行控制, 这样在不影响其他服务器的情况下, 可以大大提高部分服务器的运行速度。 步骤 203 , 当整框监控结果为资源池溢出时, 向各服务器发送功耗封顶 控制指令, 以使各服务器收到该功耗封顶控制指令后进行功耗封顶控制。 当获取到的整框监控结果为资源池溢出时, 管理板向各服务器发送功耗 封顶控制指令, 各服务器在收到功耗封顶控制指令后进行功耗封顶控制, 相 当于开启各服务器的二级封顶开关, 使得功耗封顶后各服务器的服务器功耗 不超过服务器功耗封顶值。 由此可见, 本实施例中的功耗封顶控制是结合实 时监控得到的整个机框的整框功耗和各服务器的服务器功耗来进行的 , 各服 务器共享资源池, 它们之间的资源分配和使用不是相互独立的, 因此, 本实 施例相对于现有技术的封顶功能更加灵活, 能够对资源进行最大化的合理利 用。 本实施例提供了一种功耗封顶的控制方法, 通过对整框功耗进行监控, 当获取到的整框监控结果为资源池未溢出时, 不对各服务器进行功耗封顶操 作, 当获取到的整框监控结果为资源池溢出时, 向各服务器发送功耗封顶控 制指令,以使所述各服务器收到所述功耗封顶控制指令后进行功耗封顶控制。 本实施例实现了更加灵活的封顶功能, 最大化地合理使用资源, 大大降低了 资源的浪费。 在本实施例中, 在每个服务器上分别设置一个封顶开关, 当管理板通过 监控整框功耗获取到资源池溢出时, 开启各服务器的封顶开关, 此处的封顶 开关具体为功耗封顶二级开关。 对于每一个服务器来说, 当服务器的封顶开 关被开启后, 服务器中的单板管理控制 ( Board Management Controller; 以下 筒称: BMC )单元对服务器的服务器功耗进行周期性的监控, 以实时获取到 服务器当前的服务器功耗。 在本实施例中, 服务器中的 BMC单元为服务器 的功耗封顶二级开关, 开启功耗封顶二级开关即启动 BMC单元进行功耗封 顶控制。 在 BMC单元对服务器的功耗进行监控的过程中, BMC单元根据监控得 到的服务器功耗和服务器功耗封顶值进行功耗封顶控制。 其中, 本步骤中的 服务器功耗封顶值为管理板根据监控得到的整框功耗实时更新得到的, 具体 为在管理板对整框功耗进行监控的过程中, 管理板根据监控得到的整框功耗 实时更新服务器功耗封顶值。 在本步骤中, 服务器中的 BMC单元在进行功 耗封顶控制时, 结合实时监控得到的服务器功耗和实时更新得到的服务器功 耗封顶值, 因此, 本实施例的功耗封顶控制过程是灵活的, 能够对资源进行 最大化的合理利用。 图 3为本发明功耗封顶的控制方法实施例二的流程图, 如图 3所示, 本 实施例提供了一种功耗封顶的控制方法, 可以具体包括如下步骤: 步骤 301 , 管理板根据机拒配电和业务负载情况设置整框功耗封顶值。 图 4为本发明功耗封顶的控制方法实施例二中的系统架构示意图, 如图 4所示, 假设本实施例中的机框上插设有 N个刀片服务器, 即刀片服务器 1、 刀片服务器 2、 …刀片服务器 N。 图中的管理板为刀片服务器的机框管理板, 用于提供刀片服务器和机框的管理功能, 相当于包含所有刀片服务器在内的 整个机框的功耗封顶一级开关, 本实施例中开启功耗封顶一级开关相当于启 动管理板, 由管理板开始执行本实施例中的功耗封顶的控制过程。 从图 5 中 可以看出, 管理板通过供电单元( Power Supply Unit; 以下筒称: PSU ) 实时 获取机框的整框功耗, 此处的整框功耗是指插设在机框上的所有刀片服务器 和部件在运行过程中的功耗总和; 管理板还通过 "管理通信" 通道与各个刀 片服务器进行通信, 从而实时下发各刀片服务器的服务器功耗封顶值。 PSU 用于为各刀片服务器供电, 并通过 "管理信号线" 实时向管理板上报整框功 耗。 每个刀片服务器可以主要由 BMC单元、 BIOS、 功耗检测单元和 CPU构 成, BMC单元为刀片服务器的带外管理单元, 其与刀片服务器的 BIOS配合 实现功耗封顶, 是本实施例中的功耗封顶二级开关。 BIOS用于接收 BMC单 元的控制命令, 从而对 CPU的工作频率状态 ( Performance state; 以下筒称: P-state )和时钟占空比状态( Throttle state; 以下筒称: T-state )、 内存的 P-state 和 T-state以及其他部件的工作状态进行调整, 实现刀片服务器的封顶动作的 执行。 功耗检测单元用于检测整个刀片服务器的服务器功耗, 将检测数据实 时上报给 BMC单元。 本步骤为在启动管理板进行功耗封顶控制之前, 管理板先根据机拒配电 和业务负载情况设置整框功耗封顶值, 管理板可以具体根据机拒配电要求、 刀片服务器正常运行的实际功耗、 业务压力需求等几个方面来配置整框功耗 封顶值, 即以机拒配电要求为条件, 参考刀片服务器正常运行的实际功耗的 最大值、 平均值等, 还可以结合业务负载情况等业务压力需求, 来配置整框 功耗封顶值, 具体的配置方法可以采用现有技术中本领域技术人员熟知的方 法, 此处不再赘述。 此处假设配置的整框功耗封顶值为 P。, 即 P。为整框功耗 封顶值的一个初始值。 步骤 302, 管理板根据整框功耗封顶值计算各刀片服务器的服务器功耗 封顶值。 管理板在对整框功耗封顶值进行设置之后, 可以根据该整框功耗封顶值 来具体计算分发到各刀片服务器的服务器功耗封顶值, 此处假设服务器功耗 封顶值为 Pm。 本实施例中的服务器功耗封顶值的计算方法为用整框功耗封顶 值 P。减去机框上除刀片服务器之外的其他部件的功耗值 p得到一个差值, 再 将该差值按照各刀片服务器的在位状态均分到各在位的刀片服务器上。其中, 机框上除刀片服务器之外的其他部件例如可以包括机框风扇、 电源、 管理板、 交换板等等。 在管理板计算得到刀片服务器的服务器功耗封顶值 Pm之后, 管 理板通过 "管理通信" 通道将 Pm下发到各在位的刀片服务器。 例如, 假设 N 个刀片服务器均在位, 其他部件的功耗为 p, 则可以采用下述公式(1 )来计 算得到的服务器功耗封顶值 Pm: Pm = ^ ( 1 ) In this embodiment, through the monitoring process of the foregoing steps, when the obtained monitoring result of the entire frame is that the resource pool does not overflow, the management board does not perform power consumption capping operation on each server, but continues to execute the above-mentioned full frame power consumption. Monitoring operation. In this embodiment, under the premise that the resource pool shared by each server does not overflow, the power consumption capping operation of the server is not performed, and the server is allowed to run without limit, even if the power consumption of some servers far exceeds the set server power consumption cap value. The server is also not controlled, so that the speed of some servers can be greatly improved without affecting other servers. Step 203: When the whole frame monitoring result is that the resource pool overflows, send a power consumption capping control instruction to each server, so that each server receives the power consumption capping control instruction and performs power consumption capping control. When the obtained monitoring result of the entire frame is a resource pool overflow, the management board sends a power capping control command to each server, and each server performs power capping control after receiving the power capping control command, which is equivalent to turning on each server. The level capping switch allows the server power consumption of each server after power capping to not exceed the server power capping value. It can be seen that the power capping control in this embodiment is combined with real When monitoring the entire frame power consumption of the entire chassis and the server power consumption of each server, each server shares a resource pool, and resource allocation and use between them are not independent of each other. Therefore, this embodiment is relative to the present The technical capping function is more flexible and can maximize the rational use of resources. This embodiment provides a control method for power capping. By monitoring the power consumption of the entire frame, when the obtained monitoring result of the entire frame is that the resource pool is not overflowed, the power capping operation is not performed on each server. When the resource pool overflows, the power consumption capping control command is sent to each server, so that the servers receive the power consumption capping control command and then perform power capping control. This embodiment implements a more flexible capping function, maximizes the rational use of resources, and greatly reduces the waste of resources. In this embodiment, a capping switch is respectively disposed on each server. When the management board obtains the resource pool overflow by monitoring the power consumption of the entire frame, the capping switch of each server is turned on, where the capping switch is specifically a power cap. Secondary switch. For each server, when the server's top switch is turned on, the board management controller (BMC) unit in the server periodically monitors the server's server power consumption to obtain real-time access. Current server power consumption to the server. In this embodiment, the BMC unit in the server is a power consumption capping secondary switch of the server, and the power consumption capping secondary switch is started to start the BMC unit for power capping control. During the monitoring of the power consumption of the server by the BMC unit, the BMC unit performs power capping control according to the monitored server power consumption and the server power cap value. The server power consumption cap value in this step is obtained by real-time updating of the entire frame power consumption obtained by the management board according to the monitoring, and specifically, in the process of monitoring the power consumption of the entire frame by the management board, the management board obtains the whole according to the monitoring. The box power consumption updates the server power cap value in real time. In this step, the BMC unit in the server combines the server power consumption obtained by real-time monitoring with the server power consumption cap value obtained by real-time monitoring when performing power capping control. Therefore, the power capping control process of the embodiment is flexible. , can maximize the rational use of resources. FIG. 3 is a flowchart of a second embodiment of a method for controlling power capping according to the present invention. As shown in FIG. 3, the embodiment provides a control method for power capping, which may specifically include the following steps: Step 301: The management board is configured according to the following steps: The machine rejects power distribution and service load conditions to set the entire frame power consumption cap value. 4 is a schematic diagram of a system architecture in a second embodiment of a power consumption capping control method according to the present invention. As shown in FIG. 4, it is assumed that N blade servers are inserted in a chassis of the embodiment, that is, a blade server and a blade server. 2, ... blade server N. The management board in the figure is the chassis management board of the blade server. It is used to provide the management function of the blade server and the chassis. It is equivalent to the power capping level switch of the entire chassis including all the blade servers. In this embodiment, Turning on the power consumption capping level one switch is equivalent to starting the management board, and the management board starts to execute the power consumption capping control process in this embodiment. As can be seen from Figure 5, the management board obtains the power consumption of the entire frame of the chassis in real time through the power supply unit (Power Supply Unit; the following package: PSU). The power consumption of the entire frame here is inserted in the chassis. The total power consumption of all blade servers and components during operation; the management board also communicates with each blade server through the "Management Communication" channel to deliver the server power cap value of each blade server in real time. The PSU is used to power each blade server and report the power consumption of the entire frame to the management board in real time through the "management signal line". Each of the blade servers may be composed of a BMC unit, a BIOS, a power consumption detecting unit, and a CPU. The BMC unit is an out-of-band management unit of the blade server, and cooperates with the BIOS of the blade server to implement power capping. Consumption of the top switch. The BIOS is used to receive the control commands of the BMC unit, so as to the operating frequency state of the CPU (the performance state; the following cylinder: P-state) and the clock duty state (Throttle state; the following cylinder: T-state), the memory The P-state and T-state and other components are adjusted to implement the capping action of the blade server. The power consumption detection unit is configured to detect the server power consumption of the entire blade server, and report the detection data to the BMC unit in real time. In this step, before the power management capping control is started on the management board, the management board first sets the power consumption capping value of the entire frame according to the machine rejection and service load. The management board can be specifically based on the machine rejection power supply requirement and the blade server is running normally. The actual power consumption and service pressure requirements are used to configure the power consumption capping value of the entire frame, that is, the machine power rejection requirements are met, and the maximum power and average value of the actual power consumption of the blade server can be combined. For the service pressure requirement, such as the service load situation, the entire frame power consumption capping value is configured. The specific configuration method may be a method well known to those skilled in the art in the prior art, and details are not described herein again. It is assumed here that the configured full frame power consumption cap value is P. , ie P. An initial value for the full frame power consumption cap value. Step 302: The management board calculates a server power consumption capping value of each blade server according to the entire frame power consumption capping value. After the management board sets the power consumption cap value of the entire frame, it can be based on the power consumption cap value of the entire frame. To calculate the server power capping value distributed to each blade server, it is assumed here that the server power capping value is P m . The calculation method of the server power consumption capping value in this embodiment is to use the entire frame power consumption capping value P. Subtracting the power consumption value p of the components other than the blade server on the chassis to obtain a difference, and then dividing the difference into the in-position blade servers according to the in-position state of each blade server. The components on the chassis other than the blade server may include, for example, a chassis fan, a power supply, a management board, a switchboard, and the like. After the management board calculates the server power cap value P m of the blade server, the management board delivers the P m to each in- position blade server through the "management communication" channel. For example, assuming that all N blade servers are in place and the power consumption of other components is p, the server power consumption capping value P m can be calculated by the following formula (1): P m = ^ ( 1 )
其中, Pm为服务器功耗封顶值, P。为整框功耗封顶值, p为其他部件的 功耗值, N为服务器的个数。 此处计算得到的服务器功耗封顶值 Pm也是服务 器功耗封顶值的初始值, 后续根据服务器的运行状况来更行该服务器功耗封 顶值。 或者, 在本实施例中, 也可以根据业务负载情况, 通过手动方式来设 置部分或者全部刀片服务器的服务器功耗封顶值。 步骤 303, 管理板按照预设的整框监控周期对整框功耗进行监控, 并根 据监控得到的整框功耗更新各服务器的服务器功耗封顶值。 Where P m is the server power cap value, P. For the entire frame power cap value, p is the power consumption value of other components, and N is the number of servers. The calculated server power capping value P m here is also the initial value of the server power capping value, and the server power capping value is further changed according to the running condition of the server. Alternatively, in this embodiment, the server power consumption cap value of some or all of the blade servers may also be manually set according to the traffic load condition. Step 303: The management board monitors the power consumption of the entire frame according to the preset whole frame monitoring period, and updates the server power consumption capping value of each server according to the monitored full frame power consumption.
在完成整框功耗封顶值和服务器功耗封顶值的设置之后, 可以启动管理 板进行功耗封顶的控制过程。 在启动管理板进行功耗封顶的控制过程后, 管 理板按照预设的整框监控周期对整框功耗进行监控, 此处可以假设整框功耗 为 该整框监控周期可以根据实际情况来具体设定, 例如可以设定为每秒 对整框功耗监控 10次。 同时, 管理板根据监控得到的整框功耗更新各服务器 的服务器功耗封顶值, 具体地, 管理板可以根据机框当前的整框功耗实时刷 新各刀片服务器的服务器功耗封顶值。 根据预设时间段内监控得到的多个整 框功耗的值和服务器总功耗的值计算所述整框功耗的平均值和所述服务器总 功耗的平均值, 例如, 管理板可以每秒刷新一次服务器功耗封顶值, 即预设 时间段为 1秒, 1秒内可以分别监控得到 10个整框功耗的值和 10个服务器 总功耗的值; 分别实时获取整框功耗和服务器总功耗在一秒内的平均值, 将 两个平均值相减得到除刀片服务器之外的所有部件的总功耗 P_other; 再用整 框功耗封顶值 P。减去 P_other得到一个差值,再将该差值均分到各在位的刀片 服务器上, 即得到更新后的服务器功耗封顶值 Pm。 具体可以采用下述工公式After completing the setting of the whole frame power cap value and the server power cap value, the management board can be started to control the power capping. After the management board is enabled to control the power capping process, the management board monitors the power consumption of the entire frame according to the preset frame monitoring period. It can be assumed that the entire frame power consumption period can be based on the actual situation. Specific settings, for example, can be set to monitor the power consumption of the entire frame 10 times per second. At the same time, the management board updates the server power consumption cap value of each server according to the power consumption of the whole frame obtained by the monitoring. Specifically, the management board can refresh the server power consumption cap value of each blade server in real time according to the current frame power consumption of the chassis. Calculating an average value of the power consumption of the whole frame and an average value of the total power consumption of the server according to the values of the power consumption of the whole frame and the total power consumption of the server monitored in the preset time period, for example, the management board may The server power capping value is refreshed every second, that is, the preset time period is 1 second, and the value of 10 full-frame power consumption and the total power consumption of 10 servers can be separately monitored in one second; The average power consumption of the server and the average power consumption in one second, the two average values are subtracted to obtain the total power consumption of all components except the blade server P_other; The box power cap value P. Subtract P_other to get a difference, and then divide the difference into each in-position blade server, that is, get the updated server power capping value P m . Specifically, the following formula can be used.
( 2 )来更新服务器功耗封顶值 Pm : (2) to update the server power cap value P m :
p = [P0 - (^ - PN )] ( 2 ) p = [P 0 - (^ - P N )] ( 2 )
m N  m N
其中, Pm为服务器功耗封顶值, P。为整框功耗封顶值, 为整框功耗的 平均值, 为所述服务器总功耗的平均值, N为服务器的个数, 本实施例以 预设时间段为单位, 采用上述公式对所述各服务器的服务器功耗封顶值进行 更新。 在本实施例中, 在刀片服务器的运行过程中, 不同时间段的整框功耗 不同, 本实施例通过监控到的整框功耗来实时刷新计算服务器功耗封顶值, 以使得服务器功耗封顶值可以更加接近于当前的服务器运行情况, 从而能够 大幅降低因服务器功耗封顶值不准确带来的封顶误差, 提高后续步骤中封顶 动作的精确度。 Where P m is the server power cap value, P. The average power consumption of the entire frame is the average value of the power consumption of the entire frame, which is the average value of the total power consumption of the server, and N is the number of servers. In this embodiment, the preset formula is used. The server power consumption cap value of each server is updated. In this embodiment, during the operation of the blade server, the power consumption of the entire frame is different in different time periods. In this embodiment, the power consumption capping value of the computing server is refreshed in real time by monitoring the power consumption of the entire frame, so that the server power consumption is obtained. The capping value can be closer to the current server operation, which can greatly reduce the capping error caused by the inaccuracy of the server power capping value, and improve the accuracy of the capping action in the subsequent steps.
步骤 304, 管理板判断监控得到的整框功耗是否小于预设的区间上限系 数与整框功耗封顶值之积, 如果是, 则执行步骤 305 , 否则执行步骤 306。 在每监控得到一个整框功耗 时, 管理板判断该整框功耗 是否小于预 设的区间上限系数与整框功耗封顶值 P。之积, 其中, 区间上限系数可以根据 实际情况来设定或修改, 例如设定为 0.7, 则本步骤具体为管理板判断监控得 到的整框功耗!^是否小于 Ρ。' 0.7 , 如果是, 则执行步骤 305 , 否则执行步骤 306。本实施例中的区间上限系数与整框功耗封顶值之积为刀片服务器共享的 资源池的上门限值, 后续步骤中还会涉及到区间下限系数, 区间下限系数与 整框功耗封顶值之积为资源池的下门限值, 上门限值和下门限值之间的范围 则构成资源池的防震荡区间。 步骤 305 , 管理板获取整框监控结果为资源池未溢出, 并返回执行步骤 Step 304: The management board determines whether the power consumption of the entire frame obtained by the monitoring is less than the product of the preset upper limit coefficient and the entire frame power consumption cap value. If yes, step 305 is performed; otherwise, step 306 is performed. When each monitor receives a full frame power consumption, the management board determines whether the power consumption of the entire frame is less than a preset upper limit coefficient and a full frame power consumption cap value P. The product, where the upper limit coefficient of the interval can be set or modified according to the actual situation, for example, set to 0.7, then this step is specifically for the management panel to determine the power consumption of the whole frame obtained by monitoring! ^ Is it less than Ρ. '0.7, if yes, go to step 305, otherwise go to step 306. The product of the interval upper limit coefficient and the entire frame power consumption cap value in this embodiment is the upper threshold of the resource pool shared by the blade server, and the interval lower limit coefficient, the interval lower limit coefficient and the entire frame power consumption cap value are also involved in the subsequent steps. The product is the lower threshold of the resource pool, and the range between the upper threshold and the lower threshold constitutes the anti-shock interval of the resource pool. Step 305: The management board obtains the whole frame monitoring result that the resource pool does not overflow, and returns to the execution step.
303。 在本实施例中, 当管理板监控得到的整框功耗 小于预设的区间上限系 数与整框功耗封顶值之积 Ρ。' 0.7时, 表明资源池当前没有溢出, 则管理板获 取到整框监控结果为资源池未溢出, 管理板可以不执行进一步的操作, 而返 回执行上述步骤 303 , 继续对整框功耗进行监控。 步骤 306, 管理板获取整框监控结果为资源池溢出。 当管理板监控得到的整框功耗 大于或等于预设的区间上限系数与整框 功耗封顶值之积 Ρ。' 0.7时, 表明资源池当前已溢出, 则管理板获取到整框监 控结果为资源池溢出, 并执行后续步骤 307。 在本实施例中, 当整框功耗 低 于区间上限系数与整框功耗封顶值之积 Ρ。' 0.7时, 不对服务器执行功耗封顶 动作, 允许所有刀片服务器无限制运行, 其中部分刀片服务器还可以远超服 务器封顶值运行; 当整框功耗 达到区间上限系数与整框功耗封顶值之积 P。' 0.7时, 才会开启刀片服务器的功耗封顶控制。 步骤 307, 管理板关闭资源池, 并向各服务器发送功耗封顶控制指令。 当管理板获知资源池溢出时, 管理板关闭资源池, 并向各服务器发送功 耗封顶控制指令, 以使所述各服务器收到所述功耗封顶控制指令后进行功耗 封顶控制, 即启动各刀片服务器中作为功耗封顶二级开关的 BMC单元开始 执行功耗封顶控制的操作。 步骤 308,刀片服务器中的 BMC单元对刀片服务器的服务器功耗进行监 控, 判断监控得到的服务器功耗是否小于服务器功耗封顶值, 如果是, 则继 续执行本步骤 308, 否则执行步骤 309。 在接收到管理板发送的功耗封顶控制指令后, 各刀片服务器中的 BMC 单元分别对各自刀片服务器的服务器功耗进行监控, BMC单元判断监控得到 的服务器功耗 Pn是否小于上述步骤 307更新后的服务器功耗封顶值 Pm , 如果 是, 则表明该刀片服务器的服务器功耗尚未超过服务器功耗封顶值, BMC单 元不执行任何控制功耗的动作, 而继续执行本步骤 308, 继续监控服务器功 耗, 否则执行步骤 309。 步骤 309, BMC单元根据服务器功耗与服务器功耗封顶值之差向刀片服 务器的基本输入输出系统(Basic Input Output System; 以下筒称: BIOS )发 送封顶执行通知。 303. In this embodiment, the power consumption of the entire frame obtained by the management board is less than the accumulation of the preset upper limit coefficient and the entire frame power consumption cap value. If the resource pool is not overflowed, the management board obtains the monitoring result of the entire frame. The management board does not perform any further operations. The management board can perform the above steps 303 and continue to monitor the power consumption of the entire frame. . Step 306, the management board obtains the whole box monitoring result as a resource pool overflow. When the power consumption of the entire frame obtained by the management board is greater than or equal to the preset upper limit coefficient of the interval and the power consumption cap value of the entire frame. If the resource pool is overflowed, the management board obtains the entire box monitoring result as the resource pool overflow, and performs the following step 307. In this embodiment, when the power consumption of the entire frame is lower than the accumulation of the upper limit coefficient of the interval and the power consumption cap value of the entire frame. '0.7, does not perform power capping on the server, allowing all blade servers to run unrestricted, some of them can also run far beyond the server cap value; when the whole frame power consumption reaches the upper limit coefficient and the whole frame power consumption cap value Product P. At 0.7, the power capping control of the blade server is turned on. In step 307, the management board closes the resource pool and sends a power capping control command to each server. When the management board learns that the resource pool overflows, the management board closes the resource pool, and sends a power consumption capping control instruction to each server, so that the servers receive the power consumption capping control command and then perform power consumption capping control, that is, start The BMC unit, which is a power-capped secondary switch in each blade server, begins the operation of power capping control. Step 308: The BMC unit in the blade server monitors the server power consumption of the blade server, and determines whether the monitored server power consumption is less than the server power consumption cap value. If yes, continue to perform step 308, otherwise step 309 is performed. After receiving the power capping control command sent by the management board, the BMC unit in each blade server separately monitors the server power consumption of the respective blade server, and the BMC unit determines whether the monitored server power consumption Pn is smaller than the above step 307 update. After the server power cap value P m , if yes, it indicates that the server power consumption of the blade server has not exceeded the server power cap value, and the BMC unit does not perform any action of controlling power consumption, and continues to perform this step 308 to continue monitoring. Server power consumption, otherwise go to step 309. Step 309: The BMC unit sends a cap execution notification to the basic input/output system (Basic Input Output System; BIOS) of the blade server according to the difference between the server power consumption and the server power consumption cap value.
当 BMC单元监控得到的服务器功耗 Pn大于或等于上述步骤 307更新后的 服务器功耗封顶值 Pm时, 表明该刀片服务器的服务器功耗已经达到服务器功 耗封顶值, BMC单元开始执行功耗封顶动作。 具体地, BMC根据服务器功 耗与服务器功耗封顶值之差向刀片服务器的 BIOS发送封顶执行通知。 步骤 310, BIOS 根据所述封顶执行通知调整刀片服务器的中央处理器 ( Central Processing Unit; 以下筒称: CPU )的工作频率状态和时钟占空比状 态以及内存的工作频率状态和时钟占空比状态。 刀片服务器的 BIOS在接收到封顶执行通知后, 根据该封顶执行通知调 整该刀片月良务器的 CPU的工作频率状态 P-state和时钟占空比状态 T-state、该 刀片服务器的内存的 P-state和 T-state以及该刀片服务器的其他部件的工作状 态。 When the resulting BMC unit monitoring server power P n is greater than or equal to said power capping step 307 the server the updated value P m, power consumption of the blade server indicating that the server has reached the cap value server power consumption, BMC unit begins execution function The topping action is consumed. Specifically, the BMC sends a cap execution notification to the blade server's BIOS based on the difference between the server power consumption and the server power cap value. Step 310, the BIOS adjusts the central processor of the blade server according to the capping execution notification. (Central Processing Unit; The following cylinders are called: CPU) operating frequency state and clock duty cycle state, as well as the operating frequency state of the memory and the clock duty cycle state. After receiving the capping execution notification, the BIOS of the blade server adjusts the operating frequency state P-state of the CPU of the blade server and the clock duty state T-state according to the capping execution notification, and the memory of the blade server. -state and T-state and the working state of other components of the blade server.
步骤 311 , BMC单元判断当前监控得到的服务器功耗是否小于服务器功 耗封顶值, 如果是, 则执行步骤 312, 否则返回执行步骤 310。  Step 311: The BMC unit determines whether the current monitored server power consumption is less than the server power consumption cap value. If yes, step 312 is performed; otherwise, the process returns to step 310.
在本实施例中, BMC单元对服务器功耗的监控操作不会由于功耗封顶执 行动作而停止, 即在执行上述步骤 309-311的过程中, BMC单元仍然同时监 控该刀片服务器的服务器功耗。 BMC 单元每执行一次功耗封顶动作, BMC 单元便会对当前监控得到的服务器功耗 Pn与服务器功耗封顶值 Pm进行比较, 判断当前监控得到的服务器功耗 Pn是否小于服务器功耗封顶值 Pm , 如果是, 则执行步骤 312, 否则返回执行步骤 310, 继续执行功耗封顶动作。 In this embodiment, the monitoring operation of the power consumption of the server by the BMC unit is not stopped due to the power capping execution action, that is, in the process of performing the above steps 309-311, the BMC unit still monitors the server power consumption of the blade server at the same time. . Each time the BMC unit performs the power capping action, the BMC unit compares the current server power consumption P n with the server power cap value P m to determine whether the current server power consumption P n is less than the server power consumption. cap value P m, if yes, step 312 is performed, otherwise step 310 is performed, power capping operation continues.
步骤 312, BMC单元向刀片服务器的 BIOS发送封顶执行停止通知。 在执行功耗封顶动作之后,当 BMC单元获知服务器功耗 Pn小于当前更新 的服务器功耗封顶值 Pm时, 表明之前的功耗封顶动作已经将该刀片服务器的 功耗控制在服务器功耗封顶值以下, BMC单元向该刀片服务器的 BIOS发送 封顶执行停止通知。 Step 312: The BMC unit sends a capping execution stop notification to the BIOS of the blade server. When the power after performing capping operation, when the BMC power unit P n recognizes server updates the server power less than the current capping value P m, power capping operation has been previously indicated that the power consumption of the blade server in the server power control Below the cap value, the BMC unit sends a cap execution stop notification to the blade server's BIOS.
步骤 313, BIOS根据封顶执行停止通知停止执行功耗封顶的相关操作。 刀片服务器的 BIOS在接收到封顶停止执行通知后, 根据该封顶停止执 行通知停止执行功耗封顶的相关操作。  Step 313: The BIOS stops performing the power consumption capping related operation according to the capping execution stop notification. After receiving the notification of the capping stop execution, the BIOS of the blade server stops the execution of the power capping according to the capping stop execution notification.
步骤 314, 管理板判断当前监控得到的整框功耗是否小于或等于预设的 区间下限系数与所述整框功耗封顶值之积, 如果是, 则执行步骤 315, 否则 返回执行步骤 308。  Step 314: The management board determines whether the current frame power consumption obtained by the current monitoring is less than or equal to a product of the preset interval lower limit coefficient and the whole frame power consumption capping value. If yes, step 315 is performed; otherwise, the process returns to step 308.
在本实施例中, 管理板对整框功耗的监控操作不会由于 BMC单元的服 务器功耗监控以及功耗封顶动作的执行而停止, 即在执行上述步骤 307-313 的过程中, 管理板仍然同时监控整框功耗。 当一个或几个刀片服务器的服务 器功耗下降后, 整框功耗也会随之下降。 本步骤为管理板判断当前监控得到 的整框功耗 是否小于或等于预设的区间下限系数与整框功耗封顶值 P。之 积, 如果是, 则执行步骤 315, 否则返回执行步骤 308。 本实施例中的区间下 限系数与整框功耗封顶值之积为刀片服务器共享的资源池的下门限值, 例如 可以设定区间下限系数为 0.6, 区间上限系数与整框功耗封顶值之积为刀片服 务器共享的资源池的上门限值, 上门限值和下门限值之间的范围则构成资源 池的防震荡区间。 本实施例中的防震荡区间可以根据实际情况修改, 如果资 源池的打开和关闭的频率较高, 则可以适当下调下门限值, 以增大防震荡区 间。 In this embodiment, the monitoring operation of the management board on the power consumption of the entire frame is not stopped due to the server power consumption monitoring of the BMC unit and the execution of the power consumption capping action, that is, in the process of performing the above steps 307-313, the management board The entire frame power consumption is still monitored at the same time. When one or several blade servers are served After the power consumption of the device is reduced, the power consumption of the entire frame will also decrease. In this step, the management board determines whether the power consumption of the entire frame obtained by the current monitoring is less than or equal to the preset interval lower limit coefficient and the entire frame power consumption capping value P. The product, if yes, proceeds to step 315, otherwise returns to step 308. The product of the interval lower limit coefficient and the entire frame power consumption capping value in this embodiment is the lower threshold of the resource pool shared by the blade server. For example, the interval lower limit coefficient may be set to 0.6, the interval upper limit coefficient and the entire frame power consumption cap value. The product is the upper threshold of the resource pool shared by the blade server, and the range between the upper threshold and the lower threshold constitutes the anti-shock interval of the resource pool. The anti-shock interval in this embodiment can be modified according to actual conditions. If the frequency of opening and closing the resource pool is high, the lower threshold can be appropriately lowered to increase the anti-shock interval.
步骤 315, 管理板开启资源池, 向各服务器发送功耗封顶停止指令, 以 使各服务器收到所述功耗封顶停止指令后停止进行功耗封顶控制。  Step 315: The management board starts the resource pool, and sends a power consumption capping stop instruction to each server, so that each server stops the power consumption capping control after receiving the power consumption capping stop instruction.
如果当前监控得到的整框功耗 大于预设的区间下限系数与整框功耗封 顶值之积 P。' 0.6 , 则管理板重新开启资源池, 并向各服务器发送功耗封顶停 止指令, 以使各服务器收到功耗封顶停止指令后停止进行功耗封顶控制, 此 时完成一轮封顶控制。 在本实施例中, 假设上述步骤 301配置的整框功耗封顶值 P。为 3500瓦, 机框中共有 10个在位的刀片服务器, 机框中风扇、 交换板、 管理板、 电源等 部件的总功耗为 500瓦不变, 则计算的每个刀片服务器的服务器功耗封顶值 Pm为 300瓦。 此时, 采用本实施例的上述功耗封顶的控制方法进行功耗封顶 控制, 则随着负载不断增加, 可以得到如图 5所示的资源池中功耗变化结果。 图 5和图 6分别为本发明功耗封顶的控制方法实施例二中的服务器功耗和整 框功耗变化示意图, 从图 5和图 6中可以看出, 本实施例可以实现对资源的 最大化地利用。 本实施例提供的功耗封顶的控制方法可以适用于刀片服务器或有管理板 的节点服务器, 通过管理板实时监控整框功耗的运行情况来决定刀片服务器 的功耗封顶二级开关的状态, 各个刀片服务器的 BMC单元作为刀片服务器 自身的二级开关, 与 BIOS 配合执行刀片服务器本身的功耗封顶动作, 实现 功耗封顶灵活监控设计, 且通过功耗封顶将刀片服务器的额定功耗限定为刀 片服务器实际能达到的功耗, 不仅可以避免刀片服务器实际运行时只能达到 50%的额定功耗所引起的资源浪费, 还可以大幅提高机房和机拒的设备密度, 提高空间资源的使用率。 本实施例通过两级功耗封顶开关构造出共享的资源 池, 在功耗封顶过程中结合整框功耗和服务器功耗来进行功耗封顶的控制, 在资源池未溢出的情况下确保高负载刀片超额运行, 同时业务性能也不会受 影响; 在资源池溢出的情况下确保整框功耗不超过预设的整框功耗封顶值, 从而在保证供电安全的基础上最大化提高供电资源的利用率, 即打开资源池 时可以最大程度利用供电资源, 关闭资源池时可以确保整框刀片系统的供电 安全; 且本实施例通过资源池的上下门限设定了资源池的防震荡区间, 避免 了业务负载波动引起的封顶开关频繁地打开或关闭, 提高了功耗封顶的稳定 性。 另外, 本实施例通过管理板实时监控机框风扇、 电源、 管理板、 交换板 等部件的功耗, 采用实时刷新各刀片服务器的服务器功耗封顶值的方案, 最 大幅度地降低功耗封顶的误差。 本领域普通技术人员可以理解: 实现上述各方法实施例的全部或部分步 骤可以通过程序指令相关的硬件来完成。 前述的程序可以存储于一计算机可 读取存储介质中。 该程序在执行时, 执行包括上述各方法实施例的步骤; 而 前述的存储介质包括: ROM, RAM, 磁碟或者光盘等各种可以存储程序代码 的介质。 If the power consumption of the whole frame obtained by the current monitoring is greater than the product P of the preset interval lower limit coefficient and the entire frame power consumption capping value. ' 0.6 , the management board re-opens the resource pool and sends a power capping stop command to each server, so that each server stops the power capping control after receiving the power capping stop command, and completes one round of capping control. In this embodiment, the entire frame power consumption capping value P configured in the above step 301 is assumed. For 3500 watts, there are 10 in-plane blade servers in the chassis. The total power consumption of the fan, switch board, management board, power supply, etc. in the chassis is 500 watts, and the server power of each blade server is calculated. The consumption capping value P m is 300 watts. At this time, the power consumption capping control method of the power consumption capping method of the present embodiment is used, and as the load is continuously increased, the power consumption variation result in the resource pool as shown in FIG. 5 can be obtained. FIG. 5 and FIG. 6 are respectively schematic diagrams showing changes in server power consumption and overall frame power consumption in the second embodiment of the power capping control method of the present invention. As can be seen from FIG. 5 and FIG. 6, the embodiment can implement resources. Maximize use. The control method of the power consumption capping provided in this embodiment may be applied to a blade server or a node server having a management board, and the status of the power consumption of the blade server is determined by the management board monitoring the running power of the entire frame in real time. The BMC unit of each blade server acts as the secondary switch of the blade server itself, and cooperates with the BIOS to perform the power capping action of the blade server itself, realizes the power capping flexible monitoring design, and limits the rated power consumption of the blade server to the power cap by power capping. The actual power consumption of the blade server can not only avoid the actual operation of the blade server. The waste of resources caused by 50% of rated power consumption can also greatly increase the density of equipment in the equipment room and machine rejection, and increase the utilization rate of space resources. In this embodiment, a shared resource pool is constructed by using a two-stage power capping switch, and the power capping is controlled by combining the whole frame power consumption and the server power consumption in the power capping process, and ensuring high in the case that the resource pool does not overflow. The load blade is running excessively, and the service performance is not affected. In the case of overflow of the resource pool, ensure that the power consumption of the entire frame does not exceed the preset power consumption capping value of the entire frame, thereby maximizing the power supply on the basis of ensuring power supply security. When the resource pool is opened, the power supply resource can be utilized to the maximum extent. When the resource pool is closed, the power supply of the entire rack blade system can be ensured. In this embodiment, the anti-shock interval of the resource pool is set by the upper and lower thresholds of the resource pool. The capping switch caused by the fluctuation of the business load is frequently turned on or off, which improves the stability of the power cap. In addition, in this embodiment, the management board monitors the power consumption of the fan, the power supply, the management board, and the switch board in real time, and adopts a scheme of real-time refreshing the server power consumption cap value of each blade server, thereby minimizing the power consumption capping. error. It will be understood by those skilled in the art that all or part of the steps of implementing the above method embodiments may be performed by hardware related to the program instructions. The aforementioned program can be stored in a computer readable storage medium. When the program is executed, the steps including the foregoing method embodiments are performed; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.
图 7为本发明功耗封顶的控制设备实施例一的结构示意图,如图 7所示, 本实施例提供了一种功耗封顶的控制设备, 可以具体执行上述方法实施例一 中的各个步骤, 此处不再赘述。 本实施例提供的功耗封顶的控制设备可以具 体为管理板, 其可以具体包括监控模块 701和第一封顶控制模块 702。 其中, 监控模块 701用于对整框功耗进行监控, 获取整框监控结果, 所述整框监控 结果为资源池溢出或者资源池不溢出, 所述资源池为插在同一机框上的多个 服务器所共享的供电资源。 第一封顶控制模块 702用于当所述整框监控结果 为资源池不溢出时, 不对各服务器进行功耗封顶; 当所述整框监控结果为资 源池溢出时, 向各服务器发送功耗封顶控制指令, 以使所述各服务器收到所 述功耗封顶控制指令后进行功耗封顶控制。 图 8为本发明功耗封顶的控制设备实施例二的结构示意图,如图 8所示, 本实施例提供了一种功耗封顶的控制设备, 可以具体执行上述方法实施例二 中的各个步骤, 此处不再赘述。 本实施例提供的功耗封顶的控制设备在上述 图 7所示的基础之上, 监控模块 701可以具体包括监控单元 711、 第一获取 单元 721和第二获取单元 731。 其中, 监控单元 711用于按照预设的整框监 控周期对整框功耗进行监控。 第一获取单元 721用于当所述监控得到的整框 功耗小于预设的区间上限系数与所述整框功耗封顶值之积时, 获取整框监控 结果为资源池未溢出。 第二获取单元 731用于当所述监控得到的整框功耗大 于或等于预设的区间上限系数与所述整框功耗封顶值之积时, 获取整框监控 结果为资源池溢出。 进一步地, 本实施例提供的功耗封顶的控制设备还可以包括第二封顶控 制模块 801。 第二封顶控制模块 801用于在向各服务器发送功耗封顶指令之 后, 当所述监控得到的整框功耗小于或等于预设的区间下限系数与所述整框 功耗封顶值之积时, 向所述各服务器发送功耗封顶停止指令, 以使所述各服 务器收到所述功耗封顶停止指令后停止进行功耗封顶控制, 其中, 所述区间 上限指数大于所述区间下限指数。 更进一步地, 本实施例提供的功耗封顶的控制设备还可以包括第一计算 模块 802和更新模块 803。 其中, 第一计算模块 802用于根据预设时间段内 监控得到的多个整框功耗的值和服务器总功耗的值计算所述整框功耗的平均 值和所述服务器总功耗的平均值。 更新模块 803用于根据所述整框功耗的平 均值、所述服务器总功耗的平均值和所述整框功耗封顶值,采用上述公式(2 ) 更新各服务器的服务器功耗封顶值; 并以所述预设时间段为单位, 采用上述 公式对所述各服务器的服务器功耗封顶值进行更新。 更进一步地, 本实施例提供的功耗封顶的控制设备还可以包括设置模块 804和第二计算模块 805。 其中, 设置模块 804用于在所述对整框功耗进行监 控之前, 根据机拒配电和业务负载情况设置所述整框功耗封顶值。 第二计算 模块 805用于根据所述整框功耗封顶值和机框上除所述各服务器之外的其他 部件的功耗值, 采用上述公式(1 )计算所述各服务器的服务器功耗封顶值。 本实施例提供了一种功耗封顶的控制设备, 通过对整框功耗进行监控, 当获取到的整框监控结果为资源池未溢出时, 不对各服务器进行功耗封顶操 作, 当获取到的整框监控结果为资源池溢出时, 向各服务器发送功耗封顶控 制指令,以使所述各服务器收到所述功耗封顶控制指令后进行功耗封顶控制。 本实施例实现了更加灵活的封顶功能, 最大化地合理使用资源, 大大降低了 资源的浪费。 本实施例还提供了一种功耗封顶的控制系统,可以具体如上述图 4所示, 该功耗封顶的控制系统可以具体包括管理板、 供电单元 PSU和多个刀片服务 器。 其中, 管理板可以具体包括上述图 7或图 8所示的功耗封顶的控制设备, 刀片服务器可以具体包括单板管理控制单元、 基本输入输出系统 BIOS和中 央处理器 CPU。 最后应说明的是: 以上各实施例仅用以说明本发明的技术方案, 而非对 其限制; 尽管参照前述各实施例对本发明进行了详细的说明, 本领域的普通 技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改, 或者对其中部分或者全部技术特征进行等同替换; 而这些修改或者替换, 并 不使相应技术方案的本质脱离本发明各实施例技术方案的范围。 FIG. 7 is a schematic structural diagram of Embodiment 1 of a power consumption capping control device according to the present invention. As shown in FIG. 7 , the present embodiment provides a power consumption capping control device, which may specifically perform the steps in the first embodiment of the foregoing method. , will not repeat them here. The power consumption capping control device provided in this embodiment may be specifically a management board, which may specifically include a monitoring module 701 and a first capping control module 702. The monitoring module 701 is configured to monitor the power consumption of the entire frame and obtain the monitoring result of the entire frame. The monitoring result of the whole frame is that the resource pool overflows or the resource pool does not overflow, and the resource pool is inserted in the same chassis. Power resources shared by servers. The first top control module 702 is configured to: when the monitoring result of the whole frame is that the resource pool does not overflow, do not perform power consumption capping on each server; when the monitoring result of the whole frame is a resource pool overflow, send power consumption capping to each server. And controlling the instructions to enable the each server to perform power consumption capping control after receiving the power capping control command. FIG. 8 is a schematic structural diagram of Embodiment 2 of a power consumption capping control device according to the present invention. As shown in FIG. 8 , the present embodiment provides a power consumption capping control device, which may specifically perform the steps in the second embodiment of the foregoing method. , will not repeat them here. The power consumption capping control device provided in this embodiment is as described above On the basis of the data shown in FIG. 7, the monitoring module 701 may specifically include a monitoring unit 711, a first obtaining unit 721, and a second acquiring unit 731. The monitoring unit 711 is configured to monitor the power consumption of the entire frame according to a preset entire frame monitoring period. The first obtaining unit 721 is configured to: when the power consumption of the whole frame obtained by the monitoring is less than a product of a preset upper limit coefficient and a full-frame power cap value, obtain a whole frame monitoring result that the resource pool does not overflow. The second obtaining unit 731 is configured to obtain the whole box monitoring result as a resource pool overflow when the power consumption of the whole frame obtained by the monitoring is greater than or equal to a product of the preset upper limit coefficient and the full frame power capping value. Further, the power capping control device provided in this embodiment may further include a second capping control module 801. The second capping control module 801 is configured to: when the power consumption capping command is sent to each server, when the monitored full frame power consumption is less than or equal to a product of a preset interval lower limit coefficient and the entire frame power capping value And sending, to the servers, a power consumption capping stop instruction, so that each server stops the power consumption capping control after receiving the power consumption capping stop command, wherein the interval upper limit index is greater than the interval lower limit index. Further, the power capping control device provided in this embodiment may further include a first calculating module 802 and an updating module 803. The first calculating module 802 is configured to calculate an average value of the power consumption of the whole frame and a total power consumption of the server according to values of multiple frame power consumptions monitored by the preset time period and values of total server power consumption. average value. The updating module 803 is configured to update the server power consumption cap value of each server according to the average value of the whole frame power consumption, the average value of the total power consumption of the server, and the entire frame power consumption capping value by using the above formula (2). And in the unit of the preset time period, the server power consumption cap value of each server is updated by using the above formula. Further, the power capping control device provided in this embodiment may further include a setting module 804 and a second calculating module 805. The setting module 804 is configured to set the whole frame power consumption capping value according to the machine rejection power distribution and the service load condition before the power consumption of the entire frame is monitored. The second calculating module 805 is configured to calculate the server power consumption of each server by using the above formula (1) according to the whole frame power consumption capping value and the power consumption value of other components except the servers on the chassis. The cap value. The embodiment provides a power consumption capping control device. When the power consumption of the entire frame is monitored, when the obtained monitoring result of the entire frame is not overflowed by the resource pool, the power capping operation is not performed on each server. When the resource pool overflows, the power consumption capping control command is sent to each server, so that the servers receive the power consumption capping control command and then perform power capping control. This embodiment implements a more flexible capping function, maximizes the rational use of resources, and greatly reduces the waste of resources. The power consumption capping control system may further include a management board, a power supply unit PSU, and a plurality of blade servers, as shown in FIG. 4 above. The management board may specifically include the power consumption capping control device shown in FIG. 7 or FIG. 8 above. The blade server may specifically include a board management control unit, a basic input/output system BIOS, and a central processing unit CPU. Finally, it should be noted that the above embodiments are only for explaining the technical solutions of the present invention, and are not intended to be limiting thereof; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that The technical solutions described in the foregoing embodiments may be modified, or some or all of the technical features may be equivalently replaced; and the modifications or substitutions do not deviate from the technical solutions of the embodiments of the present invention. range.

Claims

权 利 要 求 书 Claim
1、 一种功耗封顶的控制方法, 其特征在于, 包括: 对整框功耗进行监控, 获取整框监控结果, 所述整框监控结果为资源 池溢出或者资源池不溢出, 所述资源池为插在同一机框上的多个服务器所 共享的供电资源; A power consumption capping control method, comprising: monitoring a whole frame power consumption, and obtaining an entire frame monitoring result, where the whole frame monitoring result is a resource pool overflow or a resource pool does not overflow, the resource The pool is a power supply resource shared by multiple servers inserted in the same chassis;
当所述整框监控结果为资源池不溢出时, 不对各服务器进行功耗封 顶;  When the monitoring result of the whole frame is that the resource pool does not overflow, the power consumption of each server is not capped;
当所述整框监控结果为资源池溢出时, 向各服务器发送功耗封顶控制 指令, 以使所述各服务器收到所述功耗封顶控制指令后进行功耗封顶控 制。  When the whole box monitoring result is that the resource pool overflows, the power consumption capping control instruction is sent to each server, so that the servers receive the power consumption capping control command and then perform power consumption capping control.
2、 根据权利要求 1 所述的方法, 其特征在于, 所述对整框功耗进行 监控, 获取整框监控结果包括:  2. The method according to claim 1, wherein the monitoring of the power consumption of the entire frame, and obtaining the monitoring result of the entire frame includes:
按照预设的整框监控周期对整框功耗进行监控; 当监控得到的整框功耗小于预设的区间上限系数与所述整框功耗封 顶值之积时, 获取整框监控结果为资源池未溢出; 当监控得到的整框功耗大于或等于预设的区间上限系数与所述整框 功耗封顶值之积时, 获取整框监控结果为资源池溢出。  The power consumption of the entire frame is monitored according to the preset frame monitoring period. When the power consumption of the entire frame is less than the product of the preset upper limit coefficient and the power consumption cap value of the entire frame, the monitoring result of the whole frame is obtained. If the power consumption of the entire frame is greater than or equal to the product of the preset upper limit coefficient and the power consumption cap value of the entire frame, the monitoring result of the entire frame is the resource pool overflow.
3、 根据权利要求 2所述的方法, 其特征在于, 在所述向各服务器发 送功耗封顶控制指令之后, 还包括: 当所述监控得到的整框功耗小于或等于预设的区间下限系数与所述 整框功耗封顶值之积时, 向所述各服务器发送功耗封顶停止指令, 以使所 述各服务器收到所述功耗封顶停止指令后停止进行功耗封顶控制;  The method according to claim 2, after the sending the power consumption capping control instruction to each server, the method further includes: when the power consumption of the whole frame obtained by the monitoring is less than or equal to a preset lower limit Sending a power consumption capping stop instruction to each of the servers when the coefficient is a product of the full frame power consumption capping value, so that the servers stop power consumption capping control after receiving the power consumption capping stop command;
其中, 所述区间上限系数大于所述区间下限系数。  The interval upper limit coefficient is greater than the interval lower limit coefficient.
4、 根据权利要求 1 所述的方法, 其特征在于, 所述收到所述功耗封 顶控制指令后进行功耗封顶控制包括: The method according to claim 1, wherein the performing power consumption capping control after receiving the power consumption capping control instruction comprises:
收到所述功耗封顶控制指令后, 当监控得到的服务器功耗大于或等于 服务器功耗封顶值时, 根据所述服务器功耗与所述服务器功耗封顶值之差 向所述服务器的基本输入输出系统 BIOS 发送封顶执行通知, 以使所述 BIOS根据所述封顶执行通知调整所述服务器的中央处理器 CPU的工作频 率状态和时钟占空比状态以及内存的工作频率状态和时钟占空比状态。After receiving the power capping control command, when the monitored server power consumption is greater than or equal to When the server power consumption cap value, sending a cap execution notification to the basic input/output system BIOS of the server according to the difference between the server power consumption and the server power capping value, so that the BIOS performs notification adjustment according to the capping The operating frequency state and clock duty cycle state of the central processing unit CPU of the server, and the operating frequency state and clock duty cycle state of the memory.
5 5、 根据权利要求 4所述的方法, 其特征在于, 所述收到所述功耗封 顶控制指令后进行功耗封顶控制还包括: 当监控得到的服务器功耗小于服务器功耗封顶值时, 向所述服务器的 BIOS发送封顶执行停止通知,以使所述 BIOS根据所述封顶执行停止通知 停止功耗封顶的执行操作。 The method according to claim 4, wherein the performing the power consumption capping control after receiving the power consumption capping control instruction further comprises: when the monitored server power consumption is less than the server power consumption capping value Sending a capping execution stop notification to the BIOS of the server, so that the BIOS stops the execution of the power capping according to the capping execution stop notification.
10 6、 根据权利要求 1所述的方法, 其特征在于, 还包括: 根据预设时间段内监控得到的多个整框功耗的值和服务器总功耗的 值计算所述整框功耗的平均值和所述服务器总功耗的平均值; 根据所述整框功耗的平均值、 所述服务器总功耗的平均值和整框功耗 封顶值, 采用下述公式更新各服务器的服务器功耗封顶值: The method according to claim 1, further comprising: calculating the power consumption of the whole frame according to values of a plurality of whole frame power consumptions monitored in a preset time period and values of total server power consumption Average value of the average value of the server and the total power consumption of the server; according to the average value of the whole frame power consumption, the average value of the total power consumption of the server, and the entire frame power consumption cap value, the following formula is used to update each server. Server power cap value:
t _ [P0 - (^ - PN )] . t _ [P 0 - (^ - P N )] .
m _ N ,  m _ N ,
其中, Pm为所述服务器功耗封顶值, P。为所述整框功耗封顶值, 为 所述整框功耗的平均值, 为所述服务器总功耗的平均值, N为服务器的 个数; Where P m is the server power consumption cap value, P. The average power consumption of the entire frame is the average value of the total power consumption of the server, and N is the number of servers;
以所述预设时间段为单位, 采用上述公式对所述各服务器的服务器功 20 耗封顶值进行更新。  In the unit of the preset time period, the server function 20 consumption cap value of each server is updated by using the above formula.
7、 根据权利要求 1 所述的方法, 其特征在于, 在所述对整框功耗进 行监控之前, 还包括:  7. The method according to claim 1, wherein before the monitoring of the power consumption of the entire frame, the method further includes:
根据机拒配电和业务负载情况设置所述整框功耗封顶值; 根据所述整框功耗封顶值和机框上除所述各服务器之外的其他部件 25 的功耗值, 采用下述公式计算所述各服务器的服务器功耗封顶值:
Figure imgf000018_0001
Setting the power consumption capping value of the entire frame according to the machine rejection power distribution and the service load condition; according to the power consumption cap value of the entire frame and the power consumption value of the other components 25 except the servers on the chassis, The formula calculates the server power capping value of each server:
Figure imgf000018_0001
其中, Pm为所述服务器功耗封顶值, P。为所述整框功耗封顶值, p为 所述其他部件的功耗值, N为服务器的个数。 Where P m is the server power consumption cap value, P. For the full frame power consumption cap value, p is The power consumption value of the other components, N is the number of servers.
8、 一种功耗封顶的控制设备, 其特征在于, 包括: 监控模块, 用于对整框功耗进行监控, 获取整框监控结果, 所述整框 监控结果为资源池溢出或者资源池不溢出, 所述资源池为插在同一机框上 的多个服务器所共享的供电资源;  A power consumption capping control device, comprising: a monitoring module, configured to monitor power consumption of the entire frame, and obtain a whole frame monitoring result, where the monitoring result of the whole frame is a resource pool overflow or a resource pool is not Overflow, the resource pool is a power supply resource shared by multiple servers inserted in the same chassis;
第一封顶控制模块, 用于当所述整框监控结果为资源池不溢出时, 不 对各服务器进行功耗封顶; 当所述整框监控结果为资源池溢出时, 向各服 务器发送功耗封顶控制指令, 以使所述各服务器收到所述功耗封顶控制指 令后进行功耗封顶控制。  The first top control module is configured to: when the monitoring result of the whole frame is that the resource pool does not overflow, do not perform power capping on each server; when the monitoring result of the whole frame is a resource pool overflow, send power consumption capping to each server. And controlling the instructions to enable the each server to perform power consumption capping control after receiving the power capping control command.
9、 根据权利要求 8所述的设备, 其特征在于, 所述监控模块包括: 监控单元, 用于按照预设的整框监控周期对整框功耗进行监控; 第一获取单元, 用于当监控得到的整框功耗小于预设的区间上限系数 与所述整框功耗封顶值之积时, 获取整框监控结果为资源池未溢出;  The device according to claim 8, wherein the monitoring module comprises: a monitoring unit, configured to monitor power consumption of the entire frame according to a preset entire frame monitoring period; When the power consumption of the entire frame is less than the product of the preset upper limit coefficient and the power consumption cap value of the entire frame, the monitoring result of the whole frame is obtained as the resource pool does not overflow;
第二获取单元, 用于当监控得到的整框功耗大于或等于预设的区间上 限系数与所述整框功耗封顶值之积时, 获取整框监控结果为资源池溢出。  The second obtaining unit is configured to: when the monitored full frame power consumption is greater than or equal to a product of the preset interval upper limit coefficient and the whole frame power consumption capping value, obtain the whole frame monitoring result as a resource pool overflow.
10、 根据权利要求 9所述的设备, 其特征在于, 还包括:  10. The device according to claim 9, further comprising:
第二封顶控制模块, 用于在所述向各服务器发送功耗封顶控制指令之 后, 当所述监控得到的整框功耗小于或等于预设的区间下限系数与所述整 框功耗封顶值之积时, 向所述各服务器发送功耗封顶停止指令, 以使所述 各服务器收到所述功耗封顶停止指令后停止进行功耗封顶控制, 其中, 所 述区间上限指数大于所述区间下限指数。  a second capping control module, configured to: when the power consumption capping control command is sent to each server, when the power consumption of the whole frame obtained by the monitoring is less than or equal to a preset interval lower limit coefficient and the whole frame power consumption capping value And generating, by the respective servers, a power consumption capping stop instruction, so that the servers stop the power consumption capping control after receiving the power consumption capping stop command, wherein the interval upper limit index is greater than the interval Lower limit index.
11、 根据权利要求 8所述的设备, 其特征在于, 还包括:  The device according to claim 8, further comprising:
第一计算模块, 用于根据预设时间段内监控得到的多个整框功耗的值 和服务器总功耗的值计算所述整框功耗的平均值和所述服务器总功耗的 平均值;  a first calculating module, configured to calculate an average value of the whole frame power consumption and an average of the total power consumption of the server according to values of multiple frame power consumptions and total server power consumption values monitored in a preset time period Value
更新模块, 用于根据所述整框功耗的平均值、 所述服务器总功耗的平 均值和整框功耗封顶值, 采用下述公式更新各服务器的服务器功耗封顶 值: An update module, configured to update a server power consumption cap of each server according to an average value of the whole frame power consumption, an average value of the total power consumption of the server, and a full frame power consumption capping value Value:
= [P0 - (^ - PN )] . = [P 0 - (^ - P N )] .
m N '  m N '
其中, Pm为所述服务器功耗封顶值, P。为所述整框功耗封顶值, 为 所述整框功耗的平均值, 为所述服务器总功耗的平均值, N为服务器的 个数; 以所述预设时间段为单位, 采用上述公式对所述各服务器的服务器 功耗封顶值进行更新。 Where P m is the server power consumption cap value, P. The average value of the power consumption of the whole frame is the average value of the total power consumption of the server, where N is the number of servers; and the preset time period is used as a unit. The above formula updates the server power cap value of each server.
12、 根据权利要求 8所述的设备, 其特征在于, 还包括: 设置模块, 用于在所述对整框功耗进行监控之前, 根据机拒配电和业 务负载情况设置所述整框功耗封顶值; The device according to claim 8, further comprising: a setting module, configured to set the whole frame function according to the machine rejection power distribution and the service load condition before monitoring the power consumption of the entire frame Consumption cap value
第二计算模块, 用于根据所述整框功耗封顶值和机框上除所述各服务 器之外的其他部件的功耗值, 采用下述公式计算所述各服务器的服务器功 耗封顶值:
Figure imgf000020_0001
a second calculating module, configured to calculate, according to the power consumption cap value of the entire frame and the power consumption value of other components except the servers, on the chassis, calculate a server power consumption cap value of each server by using the following formula :
Figure imgf000020_0001
其中, Pm为所述服务器功耗封顶值, P。为所述整框功耗封顶值, p为 所述其他部件的功耗值, N为服务器的个数。 Where P m is the server power consumption cap value, P. For the full frame power consumption cap value, p is the power consumption value of the other components, and N is the number of servers.
13、 一种功耗封顶的控制系统, 其特征在于, 包括管理板、 供电单元 和多个刀片服务器,所述管理板包括权利要求 8-13中任一项所述的功耗封 顶的控制设备, 所述刀片服务器包括单板管理控制单元、 基本输入输出系 统 BIOS和中央处理器 CPU。 A power capping control system, comprising: a management board, a power supply unit, and a plurality of blade servers, the management board comprising the power capping control device according to any one of claims 8-13 The blade server includes a board management control unit, a basic input/output system BIOS, and a central processing unit CPU.
PCT/CN2012/079107 2012-02-28 2012-07-24 Power consumption capping control method, device and system WO2013127151A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201210048013.7A CN102624546B (en) 2012-02-28 2012-02-28 Control method, control equipment and control system for capping power consumption
CN201210048013.7 2012-02-28

Publications (1)

Publication Number Publication Date
WO2013127151A1 true WO2013127151A1 (en) 2013-09-06

Family

ID=46564237

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2012/079107 WO2013127151A1 (en) 2012-02-28 2012-07-24 Power consumption capping control method, device and system

Country Status (2)

Country Link
CN (1) CN102624546B (en)
WO (1) WO2013127151A1 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112558747A (en) * 2020-11-20 2021-03-26 山东云海国创云计算装备产业创新中心有限公司 Power capping method, system and related components of server
CN112817746A (en) * 2021-01-15 2021-05-18 浪潮电子信息产业股份有限公司 CPU power adjusting method, device, equipment and readable storage medium

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103926994A (en) * 2014-04-04 2014-07-16 浪潮电子信息产业股份有限公司 ME based dynamic server energy consumption management and correction method
CN106371546A (en) * 2016-08-30 2017-02-01 浪潮电子信息产业股份有限公司 Method and device for limiting power dissipation of whole cabinet
CN108983946B (en) * 2018-06-13 2021-04-27 烽火通信科技股份有限公司 Server power consumption control method, system and equipment
CN109062618B (en) * 2018-06-29 2022-01-11 深圳市同泰怡信息技术有限公司 Development method, system and medium for server single-node power consumption capping firmware
CN109032324B (en) * 2018-07-03 2021-01-26 北京百度网讯科技有限公司 Data center power management and control method, device, equipment and computer readable medium
CN109032325A (en) * 2018-07-17 2018-12-18 郑州云海信息技术有限公司 A kind of acquisition methods, system, device and the readable storage medium storing program for executing of controller power consumption
CN113826078B (en) * 2019-09-11 2023-01-10 阿里云计算有限公司 Resource scheduling and information prediction method, device, system and storage medium
CN111913802B (en) * 2020-07-17 2022-09-30 烽火通信科技股份有限公司 Multi-node server power consumption control method and system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101277200A (en) * 2007-03-30 2008-10-01 联想(北京)有限公司 Method and device for managing multiserver power supply
CN101689070A (en) * 2007-06-25 2010-03-31 惠普开发有限公司 Dynamicizer control for high efficiency manipulation
CN102096460A (en) * 2009-12-14 2011-06-15 英特尔公司 Method and apparatus for dynamically allocating power in a data center

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101277200A (en) * 2007-03-30 2008-10-01 联想(北京)有限公司 Method and device for managing multiserver power supply
CN101689070A (en) * 2007-06-25 2010-03-31 惠普开发有限公司 Dynamicizer control for high efficiency manipulation
CN102096460A (en) * 2009-12-14 2011-06-15 英特尔公司 Method and apparatus for dynamically allocating power in a data center

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112558747A (en) * 2020-11-20 2021-03-26 山东云海国创云计算装备产业创新中心有限公司 Power capping method, system and related components of server
CN112817746A (en) * 2021-01-15 2021-05-18 浪潮电子信息产业股份有限公司 CPU power adjusting method, device, equipment and readable storage medium

Also Published As

Publication number Publication date
CN102624546A (en) 2012-08-01
CN102624546B (en) 2015-04-29

Similar Documents

Publication Publication Date Title
WO2013127151A1 (en) Power consumption capping control method, device and system
US8473768B2 (en) Power control apparatus and method for cluster system
US10509456B2 (en) Server rack power management
TWI526843B (en) Adaptive interrupt coalescing for energy efficient mobile platforms
US8250382B2 (en) Power control of servers using advanced configuration and power interface (ACPI) states
US8504690B2 (en) Method and system for managing network power policy and configuration of data center bridging
TWI551081B (en) Remote server managing system and remote server managing method
WO2021078144A1 (en) Power management method and device
JP2009501496A5 (en)
US20090003229A1 (en) Adaptive Bandwidth Management Systems And Methods
TW201203936A (en) Method and system for managing network power policy and configuration of data center bridging
US9143403B2 (en) Autonomous metric tracking and adjustment
CN103645795A (en) Cloud computing data center energy saving method based on ANN (artificial neural network)
Kliazovich et al. Energy consumption optimization in cloud data centers
JP6263995B2 (en) Information processing system, management apparatus, information processing system control method, and management apparatus control program
WO2017148253A1 (en) Energy-saving management implementation method and apparatus, and network device
CN101833366B (en) Low-power-consumption dynamic node controlling method for cluster job management system
Carrega et al. OpenStack extensions for QoS and energy efficiency in edge computing
CN105116987A (en) Universal power supply and performance management system of cloud computing center
WO2023236727A1 (en) Temperature regulation and control method and apparatus for gateway controller, and device and medium
US9274587B2 (en) Power state adjustment
US20160179183A1 (en) Power State Adjustment
US20130061072A1 (en) Power saving node controller
CN114327017A (en) Server control method and device and server
US12007734B2 (en) Datacenter level power management with reactive power capping

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 12869609

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 12869609

Country of ref document: EP

Kind code of ref document: A1