CN103902379A - Task scheduling method and device and server cluster - Google Patents

Task scheduling method and device and server cluster Download PDF

Info

Publication number
CN103902379A
CN103902379A CN 201210573053 CN201210573053A CN103902379A CN 103902379 A CN103902379 A CN 103902379A CN 201210573053 CN201210573053 CN 201210573053 CN 201210573053 A CN201210573053 A CN 201210573053A CN 103902379 A CN103902379 A CN 103902379A
Authority
CN
Grant status
Application
Patent type
Prior art keywords
task
server
server node
performance
energy efficiency
Prior art date
Application number
CN 201210573053
Other languages
Chinese (zh)
Inventor
唐华斌
Original Assignee
中国移动通信集团公司
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

Links

Classifications

    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THIR OWN ENERGY USE
    • Y02D10/00Energy efficient computing
    • Y02D10/20Reducing energy consumption by means of multiprocessor or multiprocessing based techniques, other than acting upon the power supply
    • Y02D10/22Resource allocation
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THIR OWN ENERGY USE
    • Y02D10/00Energy efficient computing
    • Y02D10/30Reducing energy consumption in distributed systems
    • Y02D10/36Resource sharing

Abstract

The invention discloses a task scheduling method and device and a server cluster. The method comprises the steps of collecting temperature and power consumption information of nodes of servers, obtaining performance indexes of tasks which operate at the nodes of the servers, calculating the performance of each watt of the nodes of the servers according to the power consumption and temperature information and the performance indexes of the tasks, calculating task distribution weights of the nodes of the servers according to the performance of each watt of the nodes of the servers, and distributing the tasks to the nodes of the servers according to the task distribution weights of the nodes of the servers, so that the total power consumption of a computer cluster is the lowest. Task scheduling is carried out according to the task distributing weights, and the task scheduling policy that power consumption is the optimal for different processing tasks is achieved.

Description

一种任务调度方法、装置及服务器集群 One kind of task scheduling method, apparatus and server cluster

技术领域 FIELD

[0001] 本发明涉及云计算技术领域,尤其涉及一种任务调度方法、装置及服务器集群。 [0001] The present invention relates to cloud computing, and more particularly, to a task scheduling method, apparatus and server cluster.

背景技术 Background technique

[0002] 在大规模运算系统中,通常采用一个包括多台服务器的计算机集群来共同完成一项任务。 [0002] In the large-scale computing systems, usually including a computer cluster of multiple servers to work together to complete a task. 网络中的负载分发设备或软件,起着将任务分配到不同的服务器节点上运行的作用,而任务的具体调度或分配方法,一般称为“调度策略”。 Equipment or software load distribution network, will play the role assigned to the task of running on different server nodes, and specific scheduling or task assignment method, commonly referred to as "scheduling policy."

[0003] 调度策略通常包括均匀分配和加权分配两类。 [0003] scheduling policy typically comprises a uniform distribution and weighting assignment categories. 均匀分配就是通过随机或者顺序轮询的方式,将处理任务基本均匀的分配到多个服务器节点中,也就是“负载均衡”。 Uniform distribution is random or sequential polling mode, the processing tasks to a plurality of substantially uniform distribution server node, i.e. "load balancing." 加权分配又称非均匀分配,是指在任务调度中,各服务器节点不再是完全平等的关系,而是基于某种预先设定的策略或者权重进行任务调度,包括:(I)根据服务器的处理能力、网络通信能力等的不同,给予不同权值,然后根据权重来分配任务。 Weighted distribution, also known as non-uniform distribution, in the task scheduling means, each of the server nodes is no longer entirely equal relationship, but based on some policy set in advance or right heavy task scheduling, comprising: (I) The server different processing capabilities, network communication capability, etc., are given different weights, and then re-allocated tasks according to the weights. 例如,分给4路服务器的负载是2路服务器的两倍;(2)出于管理、节能等原因,对于服务器节点采取“尽量用满”的策略,即先用少数节点承担处理任务,当超出处理能力后,再把任务分配给新的服务器节点。 For example, points to the 4-way server load is twice the 2-way servers; (2) For management, energy conservation and other reasons, for the server node to take "as much as possible with the full" strategy, which is to undertake the processing tasks with a few nodes, when after exceeding capacity, and then assign the task to a new server node. [0004] 随着对IT系统和服务器能耗的逐渐重视,服务器的能源效率(通常定义为Performance/per Watt,缩写为PPW,即“每瓦特性能”)也逐渐成为加权调度策略中考虑的因素。 [0004] With the increasing emphasis on IT system and server energy consumption, energy efficiency of the server (usually defined as Performance / per Watt, abbreviated as PPW, namely "performance per watt") has gradually become a factor in the consideration of weighted scheduling strategy . 由于配置、架构等不同,不同服务器的“每瓦特性能”相差很大,有时达数倍之多。 Depending on the configuration, architecture and so on, "performance per watt" different servers vary widely, sometimes up to several times. 因此,现在已经出现一些技术方案,将服务器能效值作为任务调度的一种权值,以达到在完成同样处理任务的情况下,尽量降低整个系统能耗的目的。 Thus, now some aspect, the energy efficiency of the server as the value of one kind of task scheduling weights, to reach completion in the case where the same processing tasks to minimize energy consumption of the entire system object.

[0005] 现有技术中考虑服务器能效的调度方案中,通常有两种方式: [0005] In consideration of the prior art server energy efficiency in the schedule, there are generally two ways:

[0006] 1、根据服务器设备通常利用率越高能效越高的特点,将负载集中在少数的服务器中,而将其它服务器节点关闭或休眠。 [0006] 1, the higher the energy of the higher utilization efficiency of the server apparatus typically features, the load concentrated in a few servers, and other servers will be closed or dormant node. 这种方式虽然可以节能,但由于服务器从关机或者休眠状态恢复需要时间,因此对于服务器访问无法预料的情况,可能造成现有节点负载过高、服务无法及时响应; Although this method can save, but the server needs time to recover from shutdown or hibernation, so for server access unforeseen circumstances may cause existing node load is too high, the service can not respond in a timely manner;

[0007] 2、根据预先测试得到固定的“服务器能效”,通常与CPU利用率等易于监测的指标对应,然后基于能效权值进行任务调度。 [0007] 2. The obtained pre-test fixed "energy efficiency server" usually corresponds to the CPU utilization indicator easily monitored, and task scheduling based on the weight energy efficiency. 这种方案的主要问题在于无法动态、准确反映当前业务模型下的任务量与功耗的关系,以及不同任务负载、不同环境温度等情况下,能效存在变化的情况,因此无法真正实现能效最优的调度策略。 The main problem with this approach is not dynamic, accurately reflect the current task when the relationship between the amount of power consumption in the business model, as well as loads of different tasks, different ambient temperature, etc., energy efficiency, there is a change, so can not really achieve optimal energy efficiency scheduling policy.

发明内容 SUMMARY

[0008] 为了解决现有技术中CPU利用率并不能真实的反映服务器的实际性能以及服务器能效受负载影响的技术问题,本发明提出一种任务调度方法,基于对多台不同服务器当前负载和能耗、温度等的监控,通过计算服务器节点的能效得到该服务器节点的任务分配权值,根据任务分配权值进行任务调度,实现针对不同处理任务的能耗最优的任务调度策略。 [0008] In order to solve the technical problems of the prior art CPU utilization does not truly reflect the actual performance of the server and the server load is affected by the energy efficiency, the present invention provides a method for task scheduling, and can be based on the current load of multiple different servers monitoring consumption, temperature, etc., by calculating the energy efficiency of the server nodes assign weights to the server node of the task, task scheduling according to the task assigned a weight, power consumption for different processing tasks to realize optimal task scheduling strategy.

[0009] 本发明的一个方面,提供一种任务调度方法,应用于计算机集群,包括以下步骤:采集所述各服务器节点的温度和功耗信息;获取所述各服务器节点运行的所述任务的性能指标;根据所述功耗和温度信息以及所述任务的性能指标,计算所述各服务器节点的每瓦特性能;根据所述各服务器节点的每瓦特性能计算各服务器节点的任务分配权值;根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点。 [0009] An aspect of the present invention, there is provided a scheduling method applied to a computer cluster, comprising the steps of: acquiring said server nodes temperature and power information; obtaining said task of said each node of a server running performance; and temperature information according to the power and performance of the task, calculating the performance per watt of each server node; calculated for each server node according to the performance per watt for each task allocation server node weight; the weights assigned the task of the server nodes of the tasks assigned to each server node.

[0010] 本发明的另一个方面,提供一种任务调度装置,应用于计算机集群,包括任务分配模块、采集模块、性能指标获取模块、能效计算模块、任务分配权值计算模块,其中,所述任务分配模块,用于根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点;所述采集模块,用于采集所述各服务器节点的温度和功耗信息;所述性能指标获取模块,用于获取所述各服务器节点运行的所述任务的性能指标;所述能效计算模块,用于根据所述功耗和温度信息以及所述任务的性能指标,计算所述各服务器节点的每瓦特性能;所述任务分配权值计算模块,用于根据所述各服务器节点的每瓦特性能计算各服务器节点的任务分配权值。 Another of the present invention [0010] aspect, there is provided a scheduling apparatus is applied to a computer cluster comprising a task allocation module acquisition module, an acquisition module performance, energy efficiency calculation module, the task assigned a weight value calculation module, wherein said task assignment module configured to assign weights to the respective server node task assigning the task to each server node; said acquisition module for acquiring the server nodes temperature and power information; said performance indicator obtaining module, configured to obtain the performance indicator of the task running server nodes; the energy efficiency calculation module, based on said temperature information and the power consumption and performance of the task, calculating the servers performance per watt node; said weight calculation task assignment module configured to assign a weight value calculated for each server node according to the performance per watt task server nodes.

[0011] 本发明的又一方面,还提供了一种服务器集群,包括多个服务器节点,以及上述任 [0011] In yet another aspect of the present invention, there is provided a server cluster, comprising a plurality of server nodes, and any of the above

务调度装置。 Traffic scheduling means.

[0012] 本发明的方法、装置及服务器集群,通过对多台服务器节点当前负载和能耗、温度等的监控,动态计算服务器能效,进而得到各服务器节点的任务分配权值并进行任务调度,实现针对不同处理任务动态选择能耗最优的任务调度策略,能够降低整个系统能耗。 [0012] The method, apparatus and a server cluster according to the present invention, by monitoring a plurality of server nodes current load and energy consumption, temperature, etc., the dynamic calculation server energy efficiency, and thus obtain assign weights to the task server nodes and task scheduling, dynamic selection of the optimal energy scheduling policies for different processing tasks, can reduce overall system power consumption.

[0013] 本发明的一些优选实施例在任务调度过程中建立各种类型任务的能效模型,当新的任务需要处理时,计算机集群能够根据当前任务类型、性能需求、服务器进风口温度等信息,通过查询能效模型,快速获得最优调度策略并用于任务分配。 [0013] Examples of energy efficiency model to establish various types of tasks in the task scheduling process some preferred embodiments of the present invention, when a new task to process, the computer cluster can be the type of task, performance requirements, inlet temperature information server according to the current, by querying the model of energy efficiency, and quickly obtain optimal scheduling policy for task assignment. 该能效模型还能够根据各个任务的处理进行自适应修正。 The energy efficiency model can also be corrected according to the adaptive processing of each task.

[0014] 下面通过附图和实施例,对本发明的技术方案做进一步的详细描述。 [0014] The following drawings and embodiments, detailed description of the further aspect of the present invention.

附图说明 BRIEF DESCRIPTION

[0015] 附图用来提供对本发明的进一步理解,并且构成说明书的一部分,与本发明的实施例一起用于解释本发明,并不构成对本发明的限制。 [0015] The accompanying drawings provide a further understanding of the present invention, and constitute part of this specification, the embodiments of the invention, serve to explain the invention, not to limit the present invention. 在附图中: In the drawings:

[0016] 图1是本发明任务调度方法第一实施例的流程图; [0016] FIG. 1 is a flowchart of a scheduling method of a first embodiment of the present invention;

[0017] 图2是服务器功耗和负载之间关系的示意图; [0017] FIG. 2 is a diagram showing the relationship between power consumption and load of the server;

[0018] 图3是服务器功耗与环境温度之间关系的示意图; [0018] FIG. 3 is a diagram showing the relationship between power consumption and ambient temperature server;

[0019] 图4是本发明任务调度方法第二实施例的流程图; [0019] FIG. 4 is a flowchart of a second embodiment of the present invention, task scheduling method;

[0020] 图5是本发明任务调度装置第一实施例的结构示意图; [0020] FIG. 5 is a block diagram of a first embodiment of the apparatus of the present invention, task scheduling;

[0021] 图6是本发明任务调度装置第二实施例的结构示意图。 [0021] FIG. 6 is a schematic view of a second embodiment of the present invention, the task scheduling means.

具体实施方式 detailed description

[0022] 以下结合附图对本发明的优选实施例进行说明,应当理解,此处所描述的优选实施例仅用于说明和解释本发明,并不用于限定本发明。 [0022] Hereinafter, the preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0023] 本发明主要思想是通过监控集群中各服务器节点在处理当前任务时的能耗和温度,动态计算各服务器节点的能效,实现针对不同类型的处理任务选择最优的任务调度策略,以便降低整个集群的能耗。 [0023] The main idea of ​​the present invention by monitoring the consumption server nodes in the cluster and temperature while processing the current task, each server node dynamically calculates the energy efficiency, selecting the optimal scheduling policies for different types of processing tasks, to reduced across the cluster. [0024] 方法实施例 Example [0024] Method

[0025] 实施例一 [0025] Example a

[0026] 计算机集群中的网络负载分发设备或者分发软件能够将任务分配到集群中的各服务器节点。 [0026] The computer cluster or network load distribution apparatus distributing software tasks assigned to each server node in the cluster. 例如BIG-1P链路控制器就是常见的网络负载均衡设备,能够对网络链路的数据流进行管理以及任务的分配,进行任务分配后,各服务器节点开始运行所述任务。 After e.g. BIG-1P link controller is a common network load balancing device, and can be assigned the task of managing the data flow network link, task allocation, each node starts to run the server task.

[0027] 图1为根据本发明任务调度方法实施例的流程图,该方法应用于计算机集群,包括以下步骤: [0027] FIG. 1 is a flowchart of an embodiment according to the present invention, task scheduling method applied to a computer cluster, comprising the steps of:

[0028] 步骤102、采集所述各服务器节点的进风口温度和功耗信息; [0028] Step 102, the intake air temperature and collecting power information of each server node;

[0029]智能平台管理接口(IPMI, Intelligent Platform Management Interface)是一种开放标准的硬件管理接口规格,定义了嵌入式管理子系统进行通信的特定方法。 [0029] Intelligent Platform Management Interface (IPMI, Intelligent Platform Management Interface) is an open standard hardware management interface specification that defines a specific method for embedded management subsystems communicate. IPMI信息通过基板管理控制器(BMC,位于IPMI规格的硬件组件上)进行交流,使用低级硬件智能管理,不使用操作系统进行管理,因此,该配置允许带外服务器管理,操作系统不必负担传输系统状态数据的任务。 IPMI messages through the baseboard management controller (the BMC, hardware components located on the IPMI specification) AC, using low-level hardware intelligent management, without using the operating system to manage, and therefore, this arrangement allows the band management server, the operating system does not load the transmission system task state data. 这样,通过各服务器节点的IPMI接口监视并获取服务器的物理特征,如温度、电压、风扇工作状态、电源状态等。 Thus, through the IPMI interfaces monitoring server nodes and the server acquires physical characteristics, such as temperature, voltage, the fan working status, power status. 在该步骤中,主要通过智能型平台管理接口获取各服务器节点的功耗和温度信息,此处的服务器温度信息主要是服务器进风口温度。 In this step, mainly by the Intelligent Platform Management Interface acquires temperature information and power consumption of each server node, the server here is mainly the temperature information server inlet temperature.

[0030] 步骤104、获取所述各服务器节点运行的所述任务的性能指标; [0030] Step 104, the performance of the task of acquiring the respective nodes running the server;

[0031] 通过集群中的网络负载均衡设备或者软件,能够实时获得各服务器节点与运行任务(Task)相关的性能指标(Performance)。 [0031] through the cluster network load balancing device or software capable of real-time access each server node running the task (Task) related performance indicators (Performance). 以web服务为例,一般包括web服务的同时并发数、搜索服务的每秒能处理的搜索次数、游戏服务的最大同时在线游戏人数等等。 To web services, for example, typically include web services while the number of concurrent, the number of searches search service can handle per second, while the largest number of online games and so game services. 另外,根据运行任务类型的不同所获取的性能指标也会相应改变。 Further, according to the type of task performance metrics obtained different changes accordingly.

[0032] 这与现有技术中各服务器节点CPU使用率等单纯计算机性能指标不同。 [0032] This differs from the prior art simple computer performance server nodes CPU usage and the like. 在很多情况下,CPU利用率并不能真实的反映服务器的实际性能,因为CPU主要反映服务器的计算性能,但很多应用,如Web类、Cache类和存储类等,都不是计算密集型的业务,而是对网络I/ In many cases, CPU utilization does not truly reflect the actual performance of the server, because the CPU computing performance mainly reflects the server, but many applications, such as Web category, Cache and Storage, etc., are not computationally intensive business, but the network I /

O、内存或者磁盘I/O敏感型的业务。 O, memory, or disk I / O-sensitive business. 以Web类服务为例,如果增加一倍的网络吞吐能力,月艮务器的Web访问性能也能够增加将近一倍,但CPU可能仅仅增加10%而已。 Web-like services, for example, if the increase in network throughput double the month that works to filter Web access performance can increase nearly doubled, but the CPU may only increased by 10% only. 因此,web并发访问量、单位时间处理交易量等实际业务量指标,是比CPU利用率等更适合、对用户真正有价值的性能指标。 Therefore, web concurrent access volume per unit of time to process transactions, the actual amount of traffic indicators, is more suitable than the CPU utilization for users really valuable performance.

[0033] 步骤106、根据所述功耗和温度信息以及所述任务的性能指标,计算所述各服务器节点的每瓦特性能; [0033] Step 106, based on the temperature information and the power consumption and performance of the task, calculating the performance per watt of each server node;

[0034] 服务器的功耗通常来说会随着负载的增加而提升,也就是说,服务器利用率越高,服务器提供同样处理能力所消耗的功耗越低。 [0034] Typically server power consumption increases as the load and lifting, that is, the lower the higher the server utilization, server provides the same processing power consumed power. 但测试数据表明,不同类型和配置的服务器的增长曲线不同,在不同利用率情况下,不同服务器的功耗可能存在“交替领先”的情况,如图2所示。 But test data show the growth curve of different server types and configurations, in various utilization case, the power consumption may be cases where different servers "alternate leading", as shown in FIG. 服务器的功耗还受到环境温度(一般指服务器进风口温度)的影响。 Server power consumption is also affected by ambient temperature (inlet temperature generally refers to a server) effects. 根据美国采暖制冷与空调工程师协会(ASHRAE)所公布的服务器测试结果,随着温度升高,服务器的功耗根据优化程度的不同也有不同程度的上升,如图3所示。 According to the results of the test server American Society of Heating Refrigerating and Air-Conditioning Engineers (ASHRAE) announced, as the temperature increases, the power consumption of servers based on different optimization levels have increased to varying degrees, as shown in Fig.

[0035] 因此,通过上述分析,能够得知服务器功耗(Power)与任务类型Task、服务器节点温度以及任务性能指标存在相关关系,其中,功耗与任务性能指标和服务器温度成正比关系,任务类型则决定具体的数值关系,即 [0035] Thus, the above analysis, it is possible that the server power (the Power) is correlated with the task type Task, tasks and server nodes temperature performance, wherein the power and performance metrics and the server task is proportional to the temperature dependence, the task determines the type of the specific numerical relationship, i.e.,

[0036] Power = F(Task, Performance, Temperature)。 [0036] Power = F (Task, Performance, Temperature). [0037] 根据测量和采集得到的服务器节点的功耗值、该节点运行的任务类型、任务性能指标以及服务器节点的温度数值信息,能够得到服务器能效值PPW,即每瓦特性能,具体可以通过下述公式计算得到: [0037] The power consumption value of the server nodes measuring and collecting the resulting task type of the node running, the temperature of the numerical information of task performance metrics and the server node, it is possible to obtain the server energy efficiency value the PPW, i.e., performance per watt, specifically by the following obtained above formula:

[0038] PPff = Performance/Power。 [0038] PPff = Performance / Power.

[0039] 步骤108、根据所述各服务器节点的每瓦特性能计算各服务器节点的任务分配权值; [0039] Step 108, the server calculated for each node according to the performance per watt for each task allocation server node weight;

[0040] 根据步骤106得到的每瓦特性能数值或者当前任务下多台不同服务器节点的每瓦特性能相对值,根据以下公式计算得到各服务器节点的任务分配权值。 [0040] The performance per watt values ​​obtained in step 106 or a plurality of different values ​​of the relative performance per watt server nodes under the current task, the task is calculated weights assigned for each server node according to the following formula.

[0041] Weight = k*PPW,或者Weight = k*PPff/Average (PPff)。 [0041] Weight = k * PPW, or Weight = k * PPff / Average (PPff).

[0042] 式中k为增益系数,用于将各服务器节点的PPW数值之间的差距进行适当放大,以便于根据任务分配权值进行任务调度。 [0042] wherein k is a gain factor for the difference between the value of each server node PPW is appropriately amplified, in order to assign the task according to the task scheduling weights.

[0043] 步骤110、根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点。 [0043] Step 110, in accordance with the task assigned weights of each server node to the task assigned to each server node.

[0044] 按照使所述计算机集群总的功耗最低的基本原则,根据计算得到的各服务器节点任务分配权值进行任务分配。 [0044] The computer cluster in accordance with the lowest total power consumption of the basic principles, the task allocation according to the distribution server nodes task weights calculated. 一般采用将任务分配给任务分配权值最大的服务器节点。 Commonly used to assign tasks to assign weights to the biggest task of server nodes. 如果所述任务分配权值最大的服务器节点超过预设处理能力的阈值,则将所述任务分配给剩余的空闲服务器节点中任务分配权值最大的服务器节点。 If the weights assigned the task of the server node exceeds a preset maximum threshold value processing capability, then the task is assigned to the maximum weight distribution server node remaining idle task server node.

[0045] 根据实际需要,也可以采取其他策略,例如按照各服务器节点任务分配权值的比例将所述任务进行分配。 [0045] According to actual needs, other strategies may be adopted, for example, each node in proportion to the server task allocation value assigned to the task.

[0046] 该实施例中,通过对集群中的各服务器节点的当前负载采用业务量指标,而不是传统的CPU利用率指标,以及能耗、温度等的监控,动态计算服务器能效,实现针对不同类型处理任务的能耗最优的任务调度策略,以达到集群能耗降低的目的。 [0046] In this embodiment, the current load of each server node in the cluster using a traffic indicator, rather than the traditional CPU utilization indicators, and monitoring energy consumption, temperature, etc., the dynamic calculation server energy efficiency, for different energy type processing tasks of optimal scheduling strategy to achieve the goal of reducing the energy consumption of the cluster.

[0047] 实施例二 [0047] Second Embodiment

[0048] 图4为根据本发明任务调度方法另一种优选实施例的流程图,包括以下步骤: [0048] FIG 4 is a flowchart of an embodiment of another method according to the present invention, preferably the task scheduling, comprising the steps of:

[0049] 步骤402、将任务分配给计算机集群的各服务器节点; Server nodes [0049] Step 402, assign tasks to the computer cluster;

[0050] 步骤404、采集所述各服务器节点的进风口温度和功耗信息; [0050] Step 404, the intake air temperature and collecting power information of each server node;

[0051] 步骤406、获取所述各服务器节点运行的所述任务的性能指标; [0051] Step 406, to obtain performance indicators for the task of running the server nodes;

[0052] 步骤408、根据所述功耗和温度信息以及所述任务的性能指标,计算所述各服务器节点的每瓦特性能; [0052] Step 408, based on the temperature information and the power consumption and performance of the task, calculating the performance per watt of each server node;

[0053] 步骤410、根据所述各服务器节点的每瓦特性能计算各服务器节点的任务分配权值; [0053] Step 410, the server calculated for each node according to the performance per watt for each task allocation server node weight;

[0054] 步骤412、根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点,以使所述计算机集群总的功耗最低。 [0054] Step 412, in accordance with the task assigned weights of each server node to the task assigned to each server node, so the lowest total power consumption of a computer cluster.

[0055] 步骤414、针对不同类型的任务,重复上述步骤402至412,保存各任务的类型和对应各服务器节点的任务分配权值,作为所述类型任务的能效模型; [0055] Step 414, for different types of tasks, repeating the above steps 402 to 412, to save the task assigned weights corresponding to each type of server node and each task as a task of the type of energy efficiency model;

[0056] 步骤416、当所述计算机集群处理新的所述类型的任务时,直接调用所述能效模型进行任务调度。 [0056] Step 416, when the new computer clustering the type of task, the energy efficiency of a direct call model scheduling.

[0057] 优选的,在所述能效模型建立之后,针对不同性能和温度等条件下,通过处理新的同种类型任务得到新的数据,并用新的数据对原有的模型数据进行插值等近似计算,对原有能效模型进行补充样本,这样通过闭环调整和修正,使能效模型更精准。 [0057] Preferably, after the energy efficiency model for the different properties and conditions such as temperature, obtained by processing the new same type tasks of the new data, and interpolating the original model data with the new data approximation calculation, the original model catalog supplement energy efficiency, and closed-loop adjustment and correction, enabling more accurate model of efficiency.

[0058] 该实施例能够建立起针对不同类型的任务,各服务器节点的能效模型,在处理新任务时针对任务类型快速调用对应的能效模型进行任务调度,能够提高集群处理任务的能力和效率,并能在处理各种类型的任务时不断对能效模型进行修正和优化,使任务调度更加精准。 [0058] This embodiment can be established for different types of tasks, the energy efficiency of the model of each server node, when processing a new task quick call the corresponding energy efficiency model for a task type task scheduling, to increase capacity and efficiency of the clustering processing tasks, and can continue to modify the model of energy efficiency and optimization in dealing with all kinds of jobs, make more accurate scheduling.

[0059] 装置实施例 [0059] Example apparatus

[0060] 实施例三 [0060] Example three

[0061] 如图5所示,为本发明任务调度装置实施例的结构示意图,该装置设置在计算机集群上,包括任务分配模块502、采集模块504、性能指标获取模块506、能效计算模块508、任务分配权值计算模块510,其中, [0061] As shown in FIG 5, a schematic structural diagram of the embodiment of the present invention, task scheduling means, the apparatus is provided on a computer cluster comprising a task assignment module 502, acquisition module 504, the performance index obtaining module 506, the energy efficiency calculation module 508, task assignment weight calculation module 510, wherein,

[0062] 所述任务分配模块502,用于根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点,以使所述计算机集群总的功耗最低;所述采集模块504,用于采集所述各服务器节点的进风口温度和功耗信息;所述性能指标获取模块506,用于获取所述各服务器节点运行的所述任务的性能指标;所述能效计算模块508,用于根据所述功耗和温度信息以及所述任务的性能指标,计算所述各服务器节点的每瓦特性能;所述任务分配权值计算模块510,用于根据所述各服务器节点的每瓦特性能计算各服务器节点的任务分配权值。 [0062] The task allocation module 502, according to the assigned weights of each server node of the task assigning the task to each server node, so the lowest total power consumption of the computer cluster; the acquisition module 504 , air inlet temperature and power information for acquiring the respective server node; the performance index obtaining module 506, configured to obtain the performance indicator of the task running server nodes; computing the energy efficiency of module 508, based on said temperature information and the power consumption and performance of the task, calculating the performance per watt of each server node; the weights assigned task calculating module 510, according to each of the respective server node watt performance calculating server nodes task allocation weights.

[0063] 优选的,采集模块504通过所述各服务器节点的智能型平台管理接口获取各服务器节点的功耗和温度信息。 [0063] Preferably, the acquisition module 504 acquires the temperature information and the power consumption of each server node through the intelligent platform management interface of each server node. 性能指标获取模块506通过网络负载均衡设备获取相关任务的性能指标。 Performance index obtaining module 506 to obtain performance indicators related tasks through the network load balancing device. 能效计算模块508根据所述任务的类型、所述任务的性能指标和所述服务器进风口温度计算得到各服务器节点的功耗;根据所述各服务器节点的功耗和所述任务的性能指标计算得到各服务器节点的每瓦特性能。 Energy efficiency calculation module 508 according to the type of the task, the task server performance metrics and the calculated intake air temperature of the power consumption of each server node; calculated based on power consumption and performance of the task of the server nodes get performance per watt each server node. 任务分配模块510将所述任务分配给任务分配权值最大的服务器节点。 Task assignment module 510 to assign the tasks to the maximum weight of the task server node. 如果所述任务分配权值最大的服务器节点超过预设处理能力的阈值,则将所述任务分配给剩余的空闲服务器节点中任务分配权值最大的服务器节点。 If the weights assigned the task of the server node exceeds a preset maximum threshold value processing capability, then the task is assigned to the maximum weight distribution server node remaining idle task server node.

[0064] 任务分配模块510还可以按照各服务器节点任务分配权值的比例将所述任务进行分配。 [0064] The task allocation module 510 may also be distributed in proportion to the task task allocation weight of each server node values.

[0065] 实施例四 [0065] Fourth Embodiment

[0066] 如图6所示,除实施例五中的各模块外,本实施例的任务调度装置还包括能效建模模块612,用于保存所述任务的类型和对应各服务器节点的任务分配权值,作为所述类型任务的能效模型;所述任务分配模块602用于,当所述计算机集群处理新的所述类型的任务时,直接调用所述能效模型进行任务调度。 Task scheduling means [0066] As shown in FIG 6, each of the modules in addition to the fifth embodiment, the present embodiment further comprises energy efficiency modeling module 612, and a corresponding type of assignment tasks for each server node of the task stored weight, as the energy efficiency of the task model type; said task allocation module 602 for processing when the computer cluster of the new type of task, the energy efficiency of a direct call model scheduling.

[0067] 在该实施例中,能效建模模块612保存各任务的类型和对应各服务器节点的任务分配权值,作为所述类型任务的能效模型;当所述计算机集群处理新的所述类型的任务时,直接调用所述能效模型进行任务调度。 [0067] In this embodiment, the energy efficiency modeling module 612 and stored for each task type corresponding to each task allocation server node weights, as the energy efficiency of the type of task model; cluster when said computer processing said new type when task, the energy efficiency of a direct call model scheduling. 并且,在所述能效模型建立之后,通过处理新的同种类型任务对所述能效模型进行闭环调整和修正。 And, after the model of energy efficiency, the energy efficiency of the closed-loop model and corrected by the adjustment process of the new task of the same type.

[0068] 实施例五 [0068] Embodiment V

[0069] 本发明实施例还提供了一种服务器集群,包括多个服务器节点,该服务器集群中包括如实施例三和四种任一所述的任务调度装置。 [0069] Embodiments of the present invention further provides a server cluster, comprising a plurality of server nodes, the cluster includes three servers and a scheduling apparatus as claimed in any one of four kinds of embodiments.

[0070] 服务器集群通过运行各种不同类型的任务,得到针对各类型任务的能效模型,并储存在任务调度装置的能效建模模块中,当继续运行相同类型的任务时,集群从能效建模模块中快速调度对应的能效模型,能够进行更加合理的任务调度,从整体上降低服务器集群的功耗,同时更加有效率的运行处理该任务。 [0070] cluster of servers by running various types of tasks, to give energy efficiency model for each type of task, and stored in the energy efficiency modeling module task scheduling apparatus, while continuing to run the same type of task, the cluster from the energy efficiency modeling fast scheduling modules corresponding to the energy efficiency of the model can be made more reasonable task scheduling, reduce the power consumption of the server cluster as a whole, while the more efficient operation of the processing task. 优选的,在服务器集群的运行过程中,还能够通过闭环方式不断的对能效建模模块中的能效模型进行修正,使其不断优化。 Preferably, during operation of the servers in the cluster, it is possible energy efficiency in energy efficiency modeling module to modify the model by continuous closed loop, it continuously optimized.

[0071] 本领域普通技术人员可以理解:实现上述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,执行包括上述方法实施例的步骤;而前述的存储介质包括:服务器、单元、模块、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。 [0071] Those of ordinary skill in the art can be appreciated: realize all or part of the steps of the method described above may be implemented by a program instructing relevant hardware to complete, the program may be stored in a computer readable storage medium, the program execution when, comprising the step of performing the above-described embodiment of the method; and the storage medium comprising: a medium of the various servers may store program code, unit, module, ROM, RAM, magnetic disk, or optical disk.

[0072] 最后应说明的是:以上所述仅为本发明的优选实施例而已,并不用于限制本发明,尽管参照前述实施例对本发明进行了详细的说明,对于本领域的技术人员来说,其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换。 [0072] Finally, it should be noted that: the above embodiments are only preferred embodiments of the present invention, but the present invention is not intended to limit the present invention. Although the detailed description of the embodiments, those skilled in the art that aspect, each of which can still be described embodiments of the foregoing embodiment may be modified, or some technical features equivalents. 凡在本发明的精神和原则之内,所作的任何修改、等同替换、改进等,均应包含在本发明的保护范围之内。 Any modification within the spirit and principle of the present invention, made, equivalent substitutions, improvements, etc., should be included within the scope of the present invention.

Claims (17)

  1. 1.一种任务调度方法,应用于计算机集群,包括:采集所述各服务器节点的温度和功耗信息; 获取所述各服务器节点运行的所述任务的性能指标; 根据所述功耗和温度信息以及所述任务的性能指标,计算所述各服务器节点的每瓦特ί生倉泛; 根据所述各服务器节点的每瓦特性能计算各服务器节点的任务分配权值; 根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点。 A task scheduling method applied to a computer cluster, comprising: acquiring said server nodes temperature and power information; acquiring the performance of the task nodes running server; according to the power and temperature information and performance of the task, calculating each of the server nodes per watt ί green cartridge pan; calculated for each server node according to the performance per watt of each server node weights assigned tasks; according to the respective server node task allocation weights assigned to each server node of the task.
  2. 2.根据权利要求1所述的方法,其特征在于,所述方法还包括: 保存所述任务的类型和对应各服务器节点的任务分配权值,作为所述类型任务的能效模型; 当所述计算机集群处理新的所述类型的任务时,直接调用所述能效模型进行任务调度。 2. The method according to claim 1, wherein said method further comprises: assigning a weight value storage task of the task type corresponding to the respective server node, and as the type of task energy efficiency model; when the when the computer cluster to handle new types of tasks, the energy efficiency of a direct call model scheduling.
  3. 3.根据权利要求2所述的方法,其特征在于,保存所述任务的类型和对应各服务器节点的任务分配权值,作为所述类型任务的能效模型的步骤进一步包括: 在所述能效模型建立之后,通过处理新的相同类型任务对所述能效模型进行闭环调整和修正。 3. The method according to claim 2, wherein the weights assigned the task of the task type and stored corresponding to each server node, the step as the energy efficiency of the type of task model further comprising: energy efficiency in the model after the establishment, the energy efficiency of the closed-loop model and corrected by the adjustment process the same type of a new task.
  4. 4.根据权利要求1至3任一所述的方法,其特征在于,采集所述各服务器节点的温度和能耗信息的步骤包括: 通过所述各服务器节点的智能型平台管理接口获取各服务器节点的功耗和温度信息。 4. The method according to any one of claims 1 to 3, wherein said collecting server nodes temperature and power information comprises: acquiring by the servers of the intelligent platform management interface of each server node power and temperature information node.
  5. 5.根据权利要求1至3任一所述的方法,其特征在于,根据所述功耗和进风口温度以及所述任务的性能指标,计算所述各服务器节点的每瓦特性能的步骤包括: 根据所述任务的类型、所述任务的性能指标和所述服务器温度计算得到各服务器节点的功耗; 根据所述各服务器节点的功耗和所述任务的性能指标计算得到各服务器节点的每瓦特性能。 5. The method according to any one of claims 1 to 3, characterized in that, according to the inlet temperature and power consumption and performance of the task, calculating the performance per watt of each server node comprises: Depending on the type of the task, and the server performance of the temperature calculation for each task obtained power consumption of the server nodes; obtained per each server node according to the performance index calculation power and the task of the server nodes watt performance.
  6. 6.根据权利要求1至3任一所述的方法,其特征在于,根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点,以使所述计算机集群总的功耗最低的步骤包括: 将所述任务分配给任务分配权值最大的服务器节点。 The method according to any one of claim 1 to claim 3, characterized in that the weights assigned according to each of the tasks assigned to the server node of the task to each server node, so the total power consumption of the computer cluster minimum comprises: assigning the tasks to the task allocation server node with the maximum weight.
  7. 7.根据权利要求6所述的方法,其特征在于,所述方法还包括: 如果所述任务分配权值最大的服务器节点超过预设处理能力的阈值,则将所述任务分配给剩余的空闲服务器节点中任务分配权值最大的服务器节点。 7. The method according to claim 6, wherein said method further comprises: if the weight distribution of the maximum task server node exceeds a preset threshold value processing capability, then the task is assigned to the remaining free server node task allocation with the maximum weight of the server node.
  8. 8.根据权利要求1至3任一所述的方法,其特征在于,根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点,以使所述计算机集群总的功耗最低的步骤包括: 按照各服务器节点任务分配权值的比例将所述任务进行分配。 The method according to any one of claim 1 to claim 3, wherein the weights assigned the task in accordance with the respective server node, assigning the task to each server node, so the total power consumption of the computer cluster minimum comprises: each node in proportion to the server task allocation value assigned to the task.
  9. 9.一种任务调度装置,应用于计算机集群,包括任务分配模块、采集模块、性能指标获取模块、能效计算模块、任务分配权值计算模块,其中, 所述任务分配模块,用于根据所述各服务器节点的任务分配权值将所述任务分配给各服务器节点; 所述采集模块,用于采集所述各服务器节点的温度和功耗信息;所述性能指标获取模块,用于获取所述各服务器节点运行的所述任务的性能指标;所述能效计算模块,用于根据所述功耗和温度信息以及所述任务的性能指标,计算所述各服务器节点的每瓦特性能; 所述任务分配权值计算模块,用于根据所述各服务器节点的每瓦特性能计算各服务器节点的任务分配权值。 A task scheduling apparatus is applied to a computer cluster comprising a task allocation module acquisition module, an acquisition module performance, energy efficiency calculation module, assigning a weight value calculation module task, wherein the task allocation module, according to the assign a weight for each task of the task server node assigned to each server node; said acquisition module for acquiring the server nodes temperature and power information; and the performance index obtaining module, configured to obtain the server nodes running performance of the task; the energy efficiency calculation module, based on said temperature information and the power consumption and performance of the task, calculating the performance per watt of each server node; said task assigning a weight value calculating module, for calculating the respective server node according to the performance per watt of each server node weights assigned tasks.
  10. 10.根据权利要求9所述的装置,其特征在于, 所述装置还包括能效建模模块,用于保存所述任务的类型和对应各服务器节点的任务分配权值,作为所述类型任务的能效模型; 所述任务调度模块用于,当所述计算机集群处理新的所述类型的任务时,直接调用所述能效模型进行任务调度。 10. The apparatus according to claim 9, characterized in that said apparatus further comprises energy efficiency modeling module configured to assign a weight saving of the tasks and task type corresponding to each server node as the task type energy efficiency model; the scheduling module is configured to, when said computer clustering of the new type of task, the energy efficiency of a direct call model scheduling.
  11. 11.根据权利要求10所述的装置,其特征在于,能效建模模块进一步用于,在所述能效模型建立之后,通过处理新的同种类型任务对所述能效模型进行闭环调整和修正。 11. The apparatus according to claim 10, wherein the modeling module is further used for energy efficiency, energy efficiency, after the model, closed-loop adjustment and correction of the energy efficiency of new model by processing the same type of task.
  12. 12.根据权利要求9至11任一所述的装置,其特征在于,所述采集模块用于,通过所述各服务器节点的智能型平台管理接口获取各服务器节点的功耗和温度信息。 12. The apparatus according to any one of claims 9 to 11, characterized in that said acquisition module for acquiring the temperature information and the power consumption of each server node through the intelligent platform management interface of each server node.
  13. 13.根据权利要求9至11任一所述的装置,其特征在于,根据能效计算模块用于: 根据所述任务的类型、所述任务的性能指标和所述服务器进风口温度计算得到各服务器节点的功耗; 根据所述各服务器节点的功耗和所述任务的性能指标计算得到各服务器节点的每瓦特性能。 13. The apparatus according to any one of claims 9 to 11, characterized in that the means for calculating the energy efficiency according to: according to the type of the task, the task server performance metrics and the calculated intake air temperature of each server power node; the performance of the task, and the power consumption of each server node calculated per watt for each server node.
  14. 14.根据权利要求9至11任一所述的装置,其特征在于,所述任务分配模块用于:将所述任务分配给任务分配权值最大的服务器节点。 14. The apparatus according to any one of claims 9 to 11, wherein said task distribution module is configured to: assigning the tasks to the task allocation server node with the maximum weight.
  15. 15.根据权利要求14所述的装置,其特征在于,所述任务分配模块还用于: 如果所述任务分配权值最大的服务器节点超过预设处理能力的阈值,则将所述任务分配给剩余的空闲服务器节点中任务分配权值最大的服务器节点。 15. The apparatus according to claim 14, wherein the task assignment module is further configured to: assign weights to the task if the maximum exceeds a preset threshold server node processing capability, then the task is assigned to the remaining idle server node task allocation with the maximum weight of the server node.
  16. 16.根据权利要求9至11任一所述的装置,其特征在于,所述任务分配模块用于: 按照各服务器节点任务分配权值的比例将所述任务进行分配。 16. The apparatus according to any one of claims 9 to 11, wherein said task distribution module is configured to: assign weights to the scale of the task server nodes task allocation.
  17. 17.一种服务器集群,包括多个服务器节点,其特征在于,还包括如权利要求9至16任一所述的任务调度装置。 17. A server cluster comprising a plurality of server nodes, characterized by, further comprising a scheduling apparatus as claimed in any one of claims 9 to 16.
CN 201210573053 2012-12-25 2012-12-25 Task scheduling method and device and server cluster CN103902379A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201210573053 CN103902379A (en) 2012-12-25 2012-12-25 Task scheduling method and device and server cluster

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201210573053 CN103902379A (en) 2012-12-25 2012-12-25 Task scheduling method and device and server cluster

Publications (1)

Publication Number Publication Date
CN103902379A true true CN103902379A (en) 2014-07-02

Family

ID=50993717

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201210573053 CN103902379A (en) 2012-12-25 2012-12-25 Task scheduling method and device and server cluster

Country Status (1)

Country Link
CN (1) CN103902379A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104111875A (en) * 2014-07-03 2014-10-22 重庆大学 Device, system and method for dynamically controlling number of newly-increased tasks at cloud data center
CN104298536A (en) * 2014-10-09 2015-01-21 南京大学镇江高新技术研究院 Dynamic frequency modulation and pressure adjustment technology based data center energy-saving dispatching method
CN104360724A (en) * 2014-11-26 2015-02-18 曙光信息产业股份有限公司 Heat dissipation system and heat dissipation method of blade server based on job scheduling
CN104601704A (en) * 2015-01-19 2015-05-06 深圳市普天宜通技术有限公司 Task based group dispatching data processing method and system
CN104657212A (en) * 2015-02-13 2015-05-27 北京亚信数据有限公司 Task scheduling method and system
CN104702409A (en) * 2014-12-31 2015-06-10 北京天诚盛业科技有限公司 Cloud-platform-based biological recognition system and method for biological recognition system
CN104731660A (en) * 2014-12-31 2015-06-24 北京天诚盛业科技有限公司 Data distribution method, device and system
CN105141541A (en) * 2015-09-23 2015-12-09 浪潮(北京)电子信息产业有限公司 Task-based dynamic load balancing scheduling method and device
CN106020977A (en) * 2016-05-16 2016-10-12 深圳市中业智能系统控制有限公司 Distributed task scheduling method and apparatus used for monitoring system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1779600A (en) * 2004-11-23 2006-05-31 国际商业机器公司 Method for dynamically reprovisioning applications and other server resources in a computer center
US20070260417A1 (en) * 2006-03-22 2007-11-08 Cisco Technology, Inc. System and method for selectively affecting a computing environment based on sensed data
CN102057367A (en) * 2008-04-10 2011-05-11 惠普开发有限公司 Virtual machine migration according to environmental data
EP2472397A1 (en) * 2010-12-28 2012-07-04 POLYTEDA Software Corporation Limited Load distribution scheduling method in data processing system
CN102567086A (en) * 2010-12-30 2012-07-11 中国移动通信集团公司 Task scheduling method, equipment and system
CN102708011A (en) * 2012-05-11 2012-10-03 南京邮电大学 Multistage load estimating method facing task scheduling of cloud computing platform

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1779600A (en) * 2004-11-23 2006-05-31 国际商业机器公司 Method for dynamically reprovisioning applications and other server resources in a computer center
US20070260417A1 (en) * 2006-03-22 2007-11-08 Cisco Technology, Inc. System and method for selectively affecting a computing environment based on sensed data
CN102057367A (en) * 2008-04-10 2011-05-11 惠普开发有限公司 Virtual machine migration according to environmental data
EP2472397A1 (en) * 2010-12-28 2012-07-04 POLYTEDA Software Corporation Limited Load distribution scheduling method in data processing system
CN102567086A (en) * 2010-12-30 2012-07-11 中国移动通信集团公司 Task scheduling method, equipment and system
CN102708011A (en) * 2012-05-11 2012-10-03 南京邮电大学 Multistage load estimating method facing task scheduling of cloud computing platform

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104111875A (en) * 2014-07-03 2014-10-22 重庆大学 Device, system and method for dynamically controlling number of newly-increased tasks at cloud data center
CN104111875B (en) * 2014-07-03 2017-11-28 重庆大学 Cloud data center dynamically control the number of new tasks apparatus, system and method
CN104298536A (en) * 2014-10-09 2015-01-21 南京大学镇江高新技术研究院 Dynamic frequency modulation and pressure adjustment technology based data center energy-saving dispatching method
CN104360724A (en) * 2014-11-26 2015-02-18 曙光信息产业股份有限公司 Heat dissipation system and heat dissipation method of blade server based on job scheduling
CN104360724B (en) * 2014-11-26 2018-10-23 曙光信息产业股份有限公司 Cooling method based job scheduling blade server
CN104702409A (en) * 2014-12-31 2015-06-10 北京天诚盛业科技有限公司 Cloud-platform-based biological recognition system and method for biological recognition system
CN104731660A (en) * 2014-12-31 2015-06-24 北京天诚盛业科技有限公司 Data distribution method, device and system
CN104731660B (en) * 2014-12-31 2018-01-12 北京天诚盛业科技有限公司 Data distribution method, apparatus and system for
CN104702409B (en) * 2014-12-31 2018-04-27 北京天诚盛业科技有限公司 Biometric identification system and method based on cloud platform
CN104601704B (en) * 2015-01-19 2018-05-18 深圳市普天宜通技术股份有限公司 Kind of task-based data processing method and cluster scheduling system
CN104601704A (en) * 2015-01-19 2015-05-06 深圳市普天宜通技术有限公司 Task based group dispatching data processing method and system
CN104657212B (en) * 2015-02-13 2018-02-16 北京亚信数据有限公司 Method and system for task scheduling
CN104657212A (en) * 2015-02-13 2015-05-27 北京亚信数据有限公司 Task scheduling method and system
CN105141541A (en) * 2015-09-23 2015-12-09 浪潮(北京)电子信息产业有限公司 Task-based dynamic load balancing scheduling method and device
CN106020977A (en) * 2016-05-16 2016-10-12 深圳市中业智能系统控制有限公司 Distributed task scheduling method and apparatus used for monitoring system

Similar Documents

Publication Publication Date Title
US7210048B2 (en) Enterprise power and thermal management
Pakbaznia et al. Temperature-aware dynamic resource provisioning in a power-optimized datacenter
US20070186121A1 (en) Server system, power control method, and computer product
US20020152301A1 (en) System and method for determining the state of a service within a computer network
US20120005505A1 (en) Determining Status Assignments That Optimize Entity Utilization And Resource Power Consumption
US8001403B2 (en) Data center power management utilizing a power policy and a load factor
US20140181545A1 (en) Dynamic Balancing Of Power Across A Plurality Of Processor Domains According To Power Policy Control Bias
US20110208911A1 (en) Storage system, control method thereof, and program
Xu et al. Exploring power-performance tradeoffs in database systems
US20100257531A1 (en) Scheduling jobs of a multi-node computer system based on environmental impact
Kipp et al. Layered green performance indicators
Luo et al. A resource scheduling algorithm of cloud computing based on energy efficient optimization methods
US20110131431A1 (en) Server allocation to workload based on energy profiles
Casavant et al. Effects of response and stability on scheduling in distributed computing systems
US20110161696A1 (en) Reducing energy consumption in a cloud computing environment
CN101719081A (en) Method for scheduling virtual machines
Rao et al. QoS guarantees and service differentiation for dynamic cloud applications
JP2005196601A (en) Policy simulator for autonomous management system
Wang et al. Efficient power management of heterogeneous soft real-time clusters
US20070180453A1 (en) On demand application scheduling in a heterogeneous workload environment
CN102281290A (en) Simulation System and method for PaaS cloud platform
CN104065745A (en) Cloud computing dynamic resource scheduling system and method
US20140215076A1 (en) Allocation of Virtual Machines in Datacenters
Cao et al. An energy-aware heuristic framework for virtual machine consolidation in Cloud computing
US8756441B1 (en) Data center energy manager for monitoring power usage in a data storage environment having a power monitor and a monitor module for correlating associative information associated with power consumption

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
RJ01