[go: up one dir, main page]

CN106569575A - Energy consumption control system of computing cluster based o event response - Google Patents

Energy consumption control system of computing cluster based o event response Download PDF

Info

Publication number
CN106569575A
CN106569575A CN201510668454.0A CN201510668454A CN106569575A CN 106569575 A CN106569575 A CN 106569575A CN 201510668454 A CN201510668454 A CN 201510668454A CN 106569575 A CN106569575 A CN 106569575A
Authority
CN
China
Prior art keywords
computer
control
energy consumption
model predictive
control system
Prior art date
Legal status (The legal status 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 status listed.)
Pending
Application number
CN201510668454.0A
Other languages
Chinese (zh)
Inventor
王海峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Linyi University
Original Assignee
Linyi University
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 Linyi University filed Critical Linyi University
Priority to CN201510668454.0A priority Critical patent/CN106569575A/en
Publication of CN106569575A publication Critical patent/CN106569575A/en
Pending legal-status Critical Current

Links

Landscapes

  • Feedback Control In General (AREA)

Abstract

本发明公开了一种异构计算机集群能耗控制系统,主要包括:计算机、状态监控器、控制时间发生器、模型预测控制器和能耗状态调整执行器;所述的状态监控器连接在计算机和控制事件发生器之间;所述的控制事件发生器在状态监控器和模型预测控制器之间;所述的模型预测器在控制事件发生器和能耗状态调整执行器之间;所述的能耗状态调整执行器在模型预测控制器和计算机之间。状态监控器和能耗状态调整执行器两者都以软件进程的方式驻留在计算机中;在整个控制系统中只存在一个控制时间发生器和一个模型预测控制器。本发明具备了实时能耗优化的控制装置、方便了进行集群计算中冗余能耗的消减,实时优化效果好。

The invention discloses a heterogeneous computer cluster energy consumption control system, which mainly includes: a computer, a state monitor, a control time generator, a model predictive controller and an energy consumption state adjustment actuator; the state monitor is connected to the computer and the control event generator; the control event generator is between the state monitor and the model predictive controller; the model predictor is between the control event generator and the energy consumption state adjustment actuator; the The energy consumption state adjustment actuator is between the model predictive controller and the computer. Both the state monitor and the energy consumption state adjustment executor reside in the computer as a software process; there is only one control time generator and one model predictive controller in the whole control system. The present invention is equipped with a control device for real-time energy consumption optimization, facilitates reduction of redundant energy consumption in cluster computing, and has good real-time optimization effect.

Description

一种基于事件响应的计算集群能耗控制系统 A Computing Cluster Energy Consumption Control System Based on Event Response

技术领域 technical field

本发明属于高性能计算技术领域,尤其涉及一种基于事件响应的计算集群能耗控制系统。 The invention belongs to the technical field of high-performance computing, and in particular relates to an event response-based computing cluster energy consumption control system.

背景技术 Background technique

随着大数据应用需求的发展,高性能计算集群成为解决大数据处理的有效工具。然而高性能计算在处理大数据时处于一种持续计算状态,要消耗大量的计算能耗和冷却能耗,因此高性能计算集群的节能优化成为一个重要的研究趋势。在高性能计算集群中存在大量的随机计算任务,CPU密集型任务、I/O密集型任务和两者兼顾的任务。目前存在的问题:集群能耗控制系统不考虑计算负载的变化特征,特别是无法根据计算负载变化来实时调整集群的能耗状态,导致节能效果不明显;另外,现存的集群能耗控制系统主要是根据时间基来触发能耗控制,导致集群中大量计算机节点需要反复调整其自身能耗状态,出现节点能耗状态抖动的现象,因此浪费了大量的计算性能和能耗。目前,传统的计算集群能耗控制系统存在控制稳定性差、控制状态切换代价高的问题。 With the development of big data application requirements, high-performance computing clusters have become an effective tool to solve big data processing. However, high-performance computing is in a continuous computing state when processing big data, which consumes a lot of computing energy and cooling energy. Therefore, energy-saving optimization of high-performance computing clusters has become an important research trend. There are a large number of random computing tasks in a high-performance computing cluster, including CPU-intensive tasks, I/O-intensive tasks, and tasks that take into account both. Current problems: the cluster energy consumption control system does not consider the changing characteristics of the computing load, especially the energy consumption status of the cluster cannot be adjusted in real time according to the change of the computing load, resulting in insignificant energy saving effects; in addition, the existing cluster energy consumption control system mainly Energy consumption control is triggered based on the time base, resulting in a large number of computer nodes in the cluster needing to repeatedly adjust their own energy consumption status, resulting in node energy consumption status jitter, thus wasting a lot of computing performance and energy consumption. At present, the traditional computing cluster energy consumption control system has the problems of poor control stability and high cost of control state switching.

发明内容 Contents of the invention

本发明实施例的目的在于提供一种基于事件响应的计算集群能耗实时控制系统,旨在解决传统的能耗控制系统采用时间周期触发控制机制造成的控制稳定性差和状态切换代价高的缺点。 The purpose of the embodiments of the present invention is to provide a real-time control system for energy consumption of computing clusters based on event response, which aims to solve the shortcomings of poor control stability and high cost of state switching caused by the traditional energy consumption control system using a time period trigger control mechanism.

本发明实施例是这样实现的,一种基于事件响应的计算集群能耗实时控制系统,该计算集群能耗实时控制系统主要包括:计算机、状态监控器、控制时间发生器、模型预测控制器和能耗状态调整执行器;所述的状态监控器连接在计算机和控制事件发生器之间;所述的控制事件发生器在状态监控器和模型预测控制器之间;所述的模型预测器在控制事件发生器和能耗状态调整执行器之间;所述的能耗状态调整执行器在模型预测控制器和计算机之间。 The embodiment of the present invention is implemented in this way, a real-time control system for energy consumption of computing clusters based on event response, the real-time control system for energy consumption of computing clusters mainly includes: a computer, a state monitor, a control time generator, a model predictive controller and energy consumption state adjustment executor; the state monitor is connected between the computer and the control event generator; the control event generator is between the state monitor and the model predictive controller; the model predictor is in Between the control event generator and the energy consumption state adjustment executor; the energy consumption state adjustment executor is between the model predictive controller and the computer.

进一步,所述的模型预测控制器具体采用单片机控制芯片,将模型预测控制算法固化到单片机存储芯片中,可以对集群各计算机发出控制指令。 Further, the model predictive controller specifically adopts a single-chip microcomputer control chip, solidifies the model predictive control algorithm into the single-chip memory chip, and can issue control instructions to each computer in the cluster.

进一步,所述的控制事件发生器具体采用单片机控制芯片,通过Zigbee无线通信协议与集群中节点通信,收集各状态监控器传来的数据信号,并综合各种计算机的状态信号,根据事件产生规则来创建控制响应事件,并将控制事件通过Zigbee无线通信协议传送给模型预测控制器,以此启动模型预测控制器工作。 Further, the control event generator specifically adopts a single-chip microcomputer control chip, communicates with the nodes in the cluster through the Zigbee wireless communication protocol, collects the data signals from each state monitor, and synthesizes the state signals of various computers, according to the event generation rules To create a control response event, and transmit the control event to the model predictive controller through the Zigbee wireless communication protocol, so as to start the model predictive controller to work.

进一步,所述的状态监控器具体采用软件设计的守护进程形式,提供计算机的实时频率、CPU利用率和CPU温度等状态数据;另外通过Zigbee无线通信协议与控制事件发生器通讯,将计算机状态数据已字符串方式发送到控制事件发生器中。 Further, the state monitor specifically adopts the daemon form of software design to provide state data such as the real-time frequency of the computer, CPU utilization rate and CPU temperature; in addition, communicate with the control event generator through the Zigbee wireless communication protocol, and the computer state data Sent to the control event generator as a string.

进一步,所述的能耗状态调整执行器具体采用软件设计的操作系统内核驱动程序形式,根据具体控制指令可以实时改变计算机核心计算频率值,通过计算机频率的变化来达到调整计算机能耗状态的目标。 Further, the energy consumption state adjustment executor specifically adopts the form of a software-designed operating system kernel driver, which can change the computer core calculation frequency value in real time according to specific control instructions, and achieve the goal of adjusting the computer energy consumption state by changing the computer frequency .

效果汇总 Effect summary

本发明提供的基于事件响应的计算集群能耗实时控制系统,设置模型预测控制器,实现了动态消减集群内冗余能耗的目标,提高了能耗控制系统的准确性;设置控制事件发生器,减少了能耗状态切换的代价损失,提高了计算集群能耗控制系统的稳定性;设置状态监控器,提高了计算机节点信息收集的准确性和实时性;设置能耗状态调整执行器,实现了计算机节点动态改变核心计算频率的功能;设置计算机,实现了实时改变计算机能耗状态,减少计算过程中冗余能耗的功能。基于事件响应的计算集群能耗实时控制系统具备了实时监控计算集群能耗信息和动态优化能耗的控制装置,方便了进行冗余能耗的消减,控制稳定性好。 The event response-based real-time control system of computing cluster energy consumption provided by the present invention is equipped with a model predictive controller, which realizes the goal of dynamically reducing redundant energy consumption in the cluster and improves the accuracy of the energy consumption control system; and sets a control event generator , reducing the cost loss of energy consumption state switching, and improving the stability of the energy consumption control system of the computing cluster; setting the state monitor, improving the accuracy and real-time performance of computer node information collection; setting the energy consumption state adjustment actuator to realize The computer node has the function of dynamically changing the core computing frequency; setting up the computer realizes the function of changing the energy consumption state of the computer in real time and reducing redundant energy consumption in the computing process. The real-time control system of computing cluster energy consumption based on event response is equipped with a control device for real-time monitoring of computing cluster energy consumption information and dynamic optimization of energy consumption, which facilitates the reduction of redundant energy consumption and has good control stability.

附图说明 Description of drawings

图1是本发明实施例提供的基于事件响应的计算集群能耗控制系统的结构示意图; FIG. 1 is a schematic structural diagram of an event response-based computing cluster energy consumption control system provided by an embodiment of the present invention;

图中:1、计算机;2、状态监控器;3、控制事件发生器;4、模型预测控制器;5、能耗状态调整执行器。 In the figure: 1. Computer; 2. State monitor; 3. Control event generator; 4. Model predictive controller; 5. Energy consumption state adjustment actuator.

具体实施方式 detailed description

为了使本发明的目的、技术方案及优点更加清楚明白,以下结合实施例,对本发明进行进一步详细说明。应当理解,此处所描述的具体实施例仅仅用以解释本发明,并不用于限定本发明。 In order to make the object, technical solution and advantages of the present invention more clear, the present invention will be further described in detail below in conjunction with the examples. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

下面结合附图及具体实施例对本发明的应用原理作进一步描述。 The application principle of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

如图1所示,本发明实施例是这样实现的,基于事件响应的计算集群能耗控制系统,该基于事件响应的计算集群能耗控制系统主要包括:1、计算机;2、状态监控器;3、控制事件发生器;4、模型预测控制器;5、能耗状态调整执行器; As shown in Figure 1, the embodiment of the present invention is implemented in this way, the event response-based computing cluster energy consumption control system, the event response-based computing cluster energy consumption control system mainly includes: 1, a computer; 2, a state monitor; 3. Control event generator; 4. Model predictive controller; 5. Energy consumption state adjustment actuator;

所述的状态监控器连接在计算机和控制事件发生器之间;所述的控制事件发生器在状态监控器和模型预测控制器之间;所述的模型预测器在控制事件发生器和能耗状态调整执行器之间;所述的能耗状态调整执行器在模型预测控制器和计算机之间。 The state monitor is connected between the computer and the control event generator; the control event generator is between the state monitor and the model predictive controller; the model predictor is between the control event generator and the energy consumption Between the state adjustment executors; the energy consumption state adjustment executor is between the model predictive controller and the computer.

进一步,所述的模型预测控制器具体采用单片机控制芯片,将模型预测控制算法固化到单片机存储芯片中,可以对集群各计算机发出控制指令。 Further, the model predictive controller specifically adopts a single-chip microcomputer control chip, solidifies the model predictive control algorithm into the single-chip memory chip, and can issue control instructions to each computer in the cluster.

进一步,所述的控制事件发生器具体采用单片机控制芯片,通过Zigbee无线通信协议与集群中节点通信,收集各状态监控器传来的数据信号,并综合各种计算机的状态信号,根据事件产生规则来创建控制响应事件,并将控制事件通过Zigbee无线通信协议传送给模型预测控制器,以此启动模型预测控制器工作。 Further, the control event generator specifically adopts a single-chip microcomputer control chip, communicates with the nodes in the cluster through the Zigbee wireless communication protocol, collects the data signals from each state monitor, and synthesizes the state signals of various computers, according to the event generation rules To create a control response event, and transmit the control event to the model predictive controller through the Zigbee wireless communication protocol, so as to start the model predictive controller to work.

进一步,所述的状态监控器具体采用软件设计的守护进程形式,提供计算机的实时频率、CPU利用率和CPU温度等状态数据;另外通过Zigbee无线通信协议与控制事件发生器通讯,将计算机状态数据已字符串方式发送到控制事件发生器中。 Further, the state monitor specifically adopts the daemon form of software design to provide state data such as the real-time frequency of the computer, CPU utilization rate and CPU temperature; in addition, communicate with the control event generator through the Zigbee wireless communication protocol, and the computer state data Sent to the control event generator as a string.

进一步,所述的能耗状态调整执行器具体采用软件设计的操作系统内核驱动程序形式,根据具体控制指令可以实时改变计算机核心计算频率值,通过计算机频率的变化来达到调整计算机能耗状态的目标。 Further, the energy consumption state adjustment executor specifically adopts the form of a software-designed operating system kernel driver, which can change the computer core calculation frequency value in real time according to specific control instructions, and achieve the goal of adjusting the computer energy consumption state by changing the computer frequency .

工作原理 working principle

如图1所示,基于事件响应的计算集群能耗控制系统主要包括:1、计算机;2、状态监控器;3、控制事件发生器;4、模型预测控制器;5、能耗状态调整执行器;工作时,利用状态监控器2对计算机1的状态信息进行采集,利用控制事件发生器3对采集信息进行汇总,通过事件发生规则产生控制事件,控制事件传给模型预测控制器4后,通过模型预测控制器5进行能耗的综合控制和管理,将控制执行指令传给能耗状态调整执行器5,能耗状态调整执行器5动态调整计算机1的核心频率,降低计算过程中的冗余能耗。 As shown in Figure 1, the event response-based computing cluster energy consumption control system mainly includes: 1. Computer; 2. State monitor; 3. Control event generator; 4. Model predictive controller; 5. Energy consumption state adjustment execution When working, utilize the state monitor 2 to collect the state information of the computer 1, utilize the control event generator 3 to collect information, generate control events through the event occurrence rules, and after the control events are passed to the model predictive controller 4, The comprehensive control and management of energy consumption is carried out through the model predictive controller 5, and the control execution instruction is transmitted to the energy consumption state adjustment executor 5, and the energy consumption state adjustment executor 5 dynamically adjusts the core frequency of the computer 1, reducing the redundancy in the calculation process residual energy consumption.

以上所述仅为本发明的较佳实施例而已,并不用以限制本发明,凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等,均应包含在本发明的保护范围之内。 The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. within range.

Claims (7)

1. a kind of heterogeneous computer cluster energy consumption control system, it is characterised in that the heterogeneous computer cluster energy consumption control system mainly includes:Computer, status monitor, control time generator, model predictive controller and power consumption state adjustment executor;Described status monitor is connected between computer and control event generator;Described control event generator is between status monitor and model predictive controller;Described model prediction device is between control event generator and power consumption state adjustment executor;Described power consumption state adjusts executor between model predictive controller and computer.
2. a large amount of computers, status monitor and power consumption state adjustment executor there is in whole Heterogeneous Computing cluster, its Computer, status monitor and power consumption state adjustment executor three must exist simultaneously, and both status monitor and power consumption state adjustment executor are resident in a computer in the way of software process;A control time generator and a model predictive controller are only existed in whole control system.
3. a kind of heterogeneous computer cluster energy consumption control system as claimed in claim 1, it is characterized in that, described model predictive controller specifically adopts singlechip control chip, and Model Predictive Control Algorithm is cured in single-chip microcomputer storage chip, can send control instruction to each computer of cluster.
4. a kind of heterogeneous computer cluster energy consumption control system as claimed in claim 1, it is characterized in that, described control event generator specifically adopts singlechip control chip, communicated with cluster interior joint by Zigbee wireless communication protocols, collect the data signal that each status monitor is transmitted, and the status signal of comprehensive various computers, control response event is created according to event generation rule, and control event is sent to model predictive controller by Zigbee wireless communication protocols, with this Boot Model predictive controller work.
5. a kind of heterogeneous computer cluster energy consumption control system as claimed in claim 1, it is characterised in that described condition monitoring implement body using software design finger daemon form, there is provided the status data such as the real-time frequency of computer, cpu busy percentage and cpu temperature;Additionally by Zigbee wireless communication protocols and control event generator communication, by computer mode data, character string mode is sent in control event generator.
6. a kind of heterogeneous computer cluster energy consumption control system as claimed in claim 1, it is characterized in that, described power consumption state adjustment performs operating system nucleus driver form of the implement body using software design, computer core can in real time be changed according to concrete control instruction and calculate frequency values, by the target for calculating the change of unit frequency to reach Adjustable calculation machine power consumption state.
7. a kind of heterogeneous computer cluster energy consumption control system as claimed in claim 1, it is characterised in that single calculate node of the described computer specifically in computing cluster, needs its status information to transmit to control time generator;Simultaneously when model predictive controller transmits control command, power consumption state adjustment executor then changes the core frequency of computer, in this way realizing cutting down the redundancy energy consumption during computer operation.
CN201510668454.0A 2015-10-13 2015-10-13 Energy consumption control system of computing cluster based o event response Pending CN106569575A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510668454.0A CN106569575A (en) 2015-10-13 2015-10-13 Energy consumption control system of computing cluster based o event response

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510668454.0A CN106569575A (en) 2015-10-13 2015-10-13 Energy consumption control system of computing cluster based o event response

Publications (1)

Publication Number Publication Date
CN106569575A true CN106569575A (en) 2017-04-19

Family

ID=58508763

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510668454.0A Pending CN106569575A (en) 2015-10-13 2015-10-13 Energy consumption control system of computing cluster based o event response

Country Status (1)

Country Link
CN (1) CN106569575A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US12248352B2 (en) 2020-12-02 2025-03-11 British Telecommunications Public Limited Company Computer orchestration

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101477403A (en) * 2009-01-22 2009-07-08 浪潮电子信息产业股份有限公司 Automatic control method for system power consumption
US20100070784A1 (en) * 2008-09-15 2010-03-18 Vmware, Inc. Reducing Power Consumption in a Server Cluster
CN101916220A (en) * 2010-08-27 2010-12-15 上海交通大学 Monitoring System of Power Consumption of CPU in Cluster
CN102016748A (en) * 2008-04-21 2011-04-13 自适应计算企业股份有限公司 System and method for managing energy consumption in a compute environment
CN103488540A (en) * 2013-09-30 2014-01-01 华南理工大学 Energy consumption optimization resource scheduling method for heterogeneous cloud data center

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102016748A (en) * 2008-04-21 2011-04-13 自适应计算企业股份有限公司 System and method for managing energy consumption in a compute environment
US20100070784A1 (en) * 2008-09-15 2010-03-18 Vmware, Inc. Reducing Power Consumption in a Server Cluster
CN101477403A (en) * 2009-01-22 2009-07-08 浪潮电子信息产业股份有限公司 Automatic control method for system power consumption
CN101916220A (en) * 2010-08-27 2010-12-15 上海交通大学 Monitoring System of Power Consumption of CPU in Cluster
CN103488540A (en) * 2013-09-30 2014-01-01 华南理工大学 Energy consumption optimization resource scheduling method for heterogeneous cloud data center

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李杰 等: "基于蓝牙4.0的GPU集群功耗测量系统设计", 《电子测量与仪器学报》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US12248352B2 (en) 2020-12-02 2025-03-11 British Telecommunications Public Limited Company Computer orchestration

Similar Documents

Publication Publication Date Title
CN104216502B (en) SoC (system on chip) dynamic voltage frequency scaling method with foresight
CN102591250B (en) Dual-CPU (central processing unit) combined type motion controller
CN113535409B (en) A serverless computing resource allocation system for energy consumption optimization
CN114710563B (en) Cluster energy saving method and device
CN103595773A (en) System performance optimization method based on cloud desktop
CN109031131B (en) The real-time host system of the heterogeneous polynuclear of battery testing and analog meter and method
US20140317267A1 (en) High-Density Server Management Controller
CN103108039B (en) Service quality guarantee method in low-energy cluster environment
CN103298087B (en) Based on the wireless sensor network node low power consumption method of state machine
CN103914066A (en) Agricultural machine automatic navigation controller and control method
CN104796673B (en) A kind of cloud video monitoring system task cut-in method towards energy optimization
WO2022262819A1 (en) Edge computing module and power consumption control method therefor
US9448617B2 (en) Systems and methods for messaging-based fine granularity system-on-a-chip power gating
CN106569575A (en) Energy consumption control system of computing cluster based o event response
CN111291958B (en) A device and implementation method for interactive power supply and demand between power grid and industrial users
CN201674386U (en) Double-CPU redundancy fault-tolerant system based on high-voltage frequency converter
CN102760079B (en) Adaptive interrupt control method and device for mobile application processor
CN103559092B (en) The peak power control method of heterogeneous Computing node
CN206863739U (en) A kind of data communication machine of band storage
CN106708242A (en) Hard real-time system energy consumption optimizing method
CN104536833A (en) Method for improving high-frequency trading performance
CN104601480A (en) Zigbee turn 3G switching router
CN204707289U (en) A kind of Internet of Things intelligent node
CN105045338B (en) A kind of more peripheral hardware low energy consumption computer architectures and control method
CN109478087A (en) The method and apparatus for maintaining node power budget for the system to share power supply

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20170419