WO2015188504A1 - 对角加边模型分解协调计算的数据中心求解方法 - Google Patents

对角加边模型分解协调计算的数据中心求解方法 Download PDF

Info

Publication number
WO2015188504A1
WO2015188504A1 PCT/CN2014/087114 CN2014087114W WO2015188504A1 WO 2015188504 A1 WO2015188504 A1 WO 2015188504A1 CN 2014087114 W CN2014087114 W CN 2014087114W WO 2015188504 A1 WO2015188504 A1 WO 2015188504A1
Authority
WO
WIPO (PCT)
Prior art keywords
server
data center
calculation
energy consumption
virtual machine
Prior art date
Application number
PCT/CN2014/087114
Other languages
English (en)
French (fr)
Inventor
杨挺
向文平
冯瑛敏
盆海波
徐明玉
游金阔
王洪涛
Original Assignee
天津大学
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 天津大学 filed Critical 天津大学
Priority to US14/901,726 priority Critical patent/US10235341B2/en
Publication of WO2015188504A1 publication Critical patent/WO2015188504A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/11Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00Error detection; Error correction; Monitoring
    • G06F11/008Reliability or availability analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00Error detection; Error correction; Monitoring
    • G06F11/36Preventing errors by testing or debugging software
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/16Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/4557Distribution of virtual machine instances; Migration and load balancing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/45583Memory management, e.g. access or allocation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/45591Monitoring or debugging support
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/45595Network integration; Enabling network access in virtual machine instances

Definitions

  • the invention relates to a diagonal coordinate model decomposition and coordination calculation of a power system.
  • it relates to a data center solving method for diagonal angled model decomposition coordination calculation.
  • Cloud computing data center is a new type of Internet computing model.
  • VM Virtual Machine
  • enhancement technology provides ideas for complex computing in power systems. Data centers have more physical machines. The actual computation of a complex power system is broken down into multiple small tasks, the tasks are mapped to virtual machines, and the virtual machines are placed in the server for calculation. Through such calculations, the computation time of complex operations is greatly shortened.
  • the technical problem to be solved by the present invention is to provide a packing model that establishes an energy efficiency priority in the solution process, and shortens the diagonalization and edge model decomposition of the power system diagonal addition model decomposition and coordination calculation time and the data center energy consumption reduction. Coordinated computing data center solution method.
  • the technical solution adopted by the invention is: a data center solving method for diagonally-edge model decomposition and coordination calculation, comprising the following steps:
  • the E server is the energy consumption of a single server in the data center
  • the E switch is the energy consumption of a single switch in the data center
  • N is the number of data center servers used
  • T is the number of data center switches used
  • the energy consumption model for a single server is:
  • P baseline represents the power consumption of the server during no-load operation
  • t max represents the continuous running time of the server
  • P VM represents the power consumption of a virtual machine in the server
  • t k represents the running time of a virtual machine in the server
  • M Indicates the number of virtual machines on the server
  • the energy consumption model for a single switch is:
  • P switch represents the running power consumption of the switch
  • t max represents the running time of the switch
  • the collection of virtual machines is: with Represents the MIPS size and memory size of a single virtual machine.
  • the capacity of a single server is with Represents the MIPS size and memory size of a single server.
  • the total energy consumption model of the data center server is as follows:
  • X j,i 1 means that virtual machine j is placed in server i
  • X j,i 0 means that virtual machine j is not placed in server machine i
  • H i 1 means that server i is used
  • H i 0 means that server i is not being used
  • step 7) loading the virtual machine set in step 3) into the server of the data center;
  • Step 7) is boxed according to the constraints on the total energy consumption model of the data center server as described in step 5).
  • Each task is loaded into a virtual machine in the data center server as described in step 8), using an indiscriminate order placement method or a binding method as described below:
  • the data center solving method of the diagonal plus-edge model decomposition and coordination calculation of the invention can solve the diagonal-edge model decomposition and coordination algorithm widely applied to the power system through the data center, and the traditional single-machine multi-process and multi-machine Compared with the simple scheduling calculation mode, the invention can shorten the calculation time of the decomposition coordination algorithm and reduce the data center energy consumption, and the advantages of using the data center calculation become more obvious as the calculation scale and complexity of the power node network increases.
  • 1 is a specific flow chart of a data center solving method for diagonal plus edge model decomposition coordination calculation
  • Figure 3 is a workflow diagram of a diagonal-edge model decomposition coordination algorithm.
  • the data center solving method for the diagonal plus-edge model decomposition coordination calculation of the present invention comprises the following steps:
  • the power network can be partitioned by a node tearing method, a branch cutting method, or a unified network blocking method.
  • the server's MIPS is set to 2580
  • the memory is 512MB
  • the virtual machine MIPS size is set to [700,900]
  • the virtual memory is 128MB.
  • the E server is the energy consumption of a single server in the data center
  • the E switch is the energy consumption of a single switch in the data center
  • N is the number of data center servers used
  • T is the number of data center switches used
  • the energy consumption model for a single server is:
  • P baseline represents the power consumption of the server during no-load operation
  • t max represents the continuous running time of the server
  • P VM represents the power consumption of a virtual machine in the server
  • t k represents the running time of a virtual machine in the server
  • M Indicates the number of virtual machines on the server
  • the energy consumption model for a single switch is:
  • P switch represents the running power consumption of the switch
  • t max represents the running time of the switch
  • the energy consumption of the data center server accounts for about 80% of the total energy consumption of the IT equipment, according to the hierarchical sequence method, we can first calculate the energy consumption of the server, then calculate the energy consumption of the switch, and finally get the total IT equipment of the data center. Energy consumption can therefore minimize the energy consumption of the server and attribute the energy consumption model of the server to a packing model to reduce the total energy consumption.
  • the server's energy consumption model is reduced to a boxing model, and the server's minimum energy model is given:
  • the collection of virtual machines is: with Represents the MIPS size and memory size of a single virtual machine.
  • the capacity of a single server is with Represents the MIPS size and memory size of a single server.
  • the total energy consumption model of the data center server is as follows:
  • X j,i 1 means that virtual machine j is placed in server i
  • X j,i 0 means that virtual machine j is not placed in server machine i
  • H i 1 means that server i is used
  • H i 0 means that server i is not being used
  • Best adaptation algorithm The best adaptation algorithm is to put the items into the box in order, first put the first item into the first box, then consider the second item, if the first box consumes the second Items, put the second item into the first box, if you can't put it down, reopen a box, in order, when the i-th item is put, put it in all the items that can hold the item and put In the box with the smallest remaining space after entering, when all the boxes do not meet the requirements, reopen a box.
  • Descending optimal adaptation algorithm Before packing, sort the items in descending order according to their size, then put the first item into the first box, then consider the second item, if the first box can put down the first Two items, put the second item into the first box, if you can't put it down, reopen a box, in order, when the i-th item is placed, put it in all the items that can hold the item and In the box with the smallest remaining space after the insertion, when all the boxes do not meet the requirements, reopen a box;
  • step 7) loading the virtual machine set in step 3) into the server of the data center, specifically according to the constraint on the total energy consumption model of the data center server described in step 5);
  • Each task in step 2) is correspondingly loaded into a virtual machine in the data center server, and the task is mapped to the data center server, as shown in FIG. 2, and each task is loaded into a virtual machine.
  • the workflow diagram can be obtained according to the calculation process of the decomposition and coordination algorithm of the diagonal plus edge model.
  • the workflow diagram is shown in FIG. 3.
  • the rectangle in the figure is a task, which represents a calculation step, and the ellipse is data, and the task is directed to the task.
  • the input data of the task, the arrow indicated by the task indicates that the task calculates the output result, and since each data is connected to the arrow of one task and the arrow of the other task, it is the communication data between the two tasks.
  • the system is divided into three sub-areas, the number of nodes of the three sub-areas is 35, 35, 48, respectively, and the number of nodes of the virtual border network is 7.
  • the data center is a Fat-tree structure, which includes 54 servers and 45 6-port switches.
  • Each server is configured as CPU: Pentium4 (2.8GHz), memory: 512Mb, baseline power consumption: 145w, load virtual machine operation power consumption: 10w, switch uses Huawei S3552F-EA three-layer switch, running power: 54w.
  • the switch port rate fluctuates between [0,8]Mb;
  • the total energy consumption of the data center IT equipment is obtained according to the total energy consumption calculation formula given in step 4).
  • the virtual machine is put into the server using the best adaptive algorithm and the optimal ordering algorithm for descending order.
  • the task is placed in the virtual machine and placed in the order of the box binding. Therefore, the combination generates four algorithms, namely, the indiscriminate order placement-optimal adaptation algorithm.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Quality & Reliability (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Computer Hardware Design (AREA)
  • Computing Systems (AREA)
  • Operations Research (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Power Sources (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

公开了一种对角加边模型分解协调计算的数据中心求解方法,在求解过程中,首先用已有的网络分块法对电力网络进行分块和设定虚拟机参数,建立能效优先的装箱模型,将对角加边模型分解协调算法计算流程中的每一个步骤设定为一个任务,通过服务器承载虚拟机,虚拟机承载任务的映射方式,用数据中心对对角加边模型分解协调算法进行求解,最后得到整个计算过程的时间和数据中心的能耗。能够缩短对角加边模型分解协调算法的计算时间和降低数据中心能耗,并且,随着电力节点网络计算规模和复杂度的增加,采用数据中心计算的优势就越明显。

Description

对角加边模型分解协调计算的数据中心求解方法 技术领域
本发明涉及一种电力系统对角加边模型分解协调计算。特别是涉及一种对角加边模型分解协调计算的数据中心求解方法。
背景技术
随着超大规模区域互联电网的发展,电力系统中的潮流计算,节点电压计算,暂态计算,无功优化计算等对计算机的速度提出了更高的要求,当运算规模达到一定程度时,其计算时间变得越来越长,甚至有时很难用现有的计算方法对其求解。对角加边模型广泛应用于电力系统的各个方面,在对角加边模型的求解算法中,最常见的求解算法为分解协调并行算法,目前分解协调并行算法主要考虑单机并行或者简单调度的多机并行,效果较差,运算时间较长。
云计算数据中心是一种新型的互联网计算模式,随着虚拟化的提出,VM(Virtual Machine)迁移和加强技术为电力系统中复杂的计算提供了思路,数据中心拥有较多的物理机,可以将一个实际的电力系统复杂计算分解成多个小的任务,将任务映射到虚拟机,再将虚拟机放置于服务器中计算,通过这样的计算方式,复杂运算的运算时间将大大缩短。
在数据中心并行计算时,与计算加速比同等重要的是能耗问题。这是因为随着大数据、云计算的广泛应用,数据中心能耗变得惊人可观:2011年,世界数据中心的耗电量已达到6358亿度(中国为568亿度);2012年耗电量增加到7202亿度(中国为664亿度)。因此,只有能量有效计算模型和映射方法才能被业界认同,并移植到数据中心中应用。虽然已有部分学者对数据中心节能问题进行研究,但多针对通用模型,还没有针对电力系统广泛应用的对角加边模型分解协调并行计算的数据中心映射研究。
发明内容
本发明所要解决的技术问题是,提供一种在求解过程中,建立能效优先的装箱模型,缩短电力系统对角加边模型分解协调计算时间和降低数据中心能耗的对角加边模型分解协调计算的数据中心求解方法。
本发明所采用的技术方案是:一种对角加边模型分解协调计算的数据中心求解方法,包括如下步骤:
1)通过已有的网络分块法将电力网络进行划分得到电力网络的分块;
2)将电力系统对角加边模型分解协调算法的计算流程中的每一个计算步骤设定为一个任务,确定整个计算过程的任务数;
3)获得数据中心服务器的MIPS大小和内存大小,设定虚拟机的个数以及虚拟机中CPU的MIPS大小和内存大小;
4)计算数据中心的IT设备总能耗Etotal
Figure PCTCN2014087114-appb-000001
式中,Eserver为数据中心单台服务器的能耗,Eswitch为数据中心单个交换机的能耗,N为数据中心服务器的使用台数,T为数据中心交换机的使用个数,其中,
单台服务器的能耗模型为:
Figure PCTCN2014087114-appb-000002
式中:Pbaseline表示服务器的空载运行时的功耗,tmax表示服务器的持续运行时间,PVM表示服务器中一个虚拟机的功耗,tk表示服务器中一个虚拟机的运行时间,M表示服务器上虚拟机的数量;
单个交换机的能耗模型为:
Eswitch=Pswitchtmax      (3)
式中:Pswitch表示交换机运行功耗,tmax表示交换机运行时间;
5)将服务器的能耗模型归结为装箱模型,并给出服务器的最小能耗模型:
虚拟机的集合为:
Figure PCTCN2014087114-appb-000003
Figure PCTCN2014087114-appb-000004
Figure PCTCN2014087114-appb-000005
分别表示单个虚拟机的MIPS大小和内存大小,单台服务器的容量为
Figure PCTCN2014087114-appb-000006
Figure PCTCN2014087114-appb-000007
Figure PCTCN2014087114-appb-000008
分别表示单台服务器的MIPS大小和内存大小,数据中心服务器的总能耗模型如下:
Figure PCTCN2014087114-appb-000009
对数据中心服务器的总能耗模型的约束如下:
Figure PCTCN2014087114-appb-000010
Figure PCTCN2014087114-appb-000011
Xj,i=0or1     (7)
Hi=0or1     (8)
式中,Xj,i=1表示虚拟机j放置于服务器i中,Xj,i=0表示虚拟机j未放置于服务器机i中;Hi=1表示服务器i被使用,Hi=0表示服务器i未被使用;
6)采用最佳适应算法或降序最佳适应算法对数据中心服务器的总能耗模型求解;
7)将步骤3)中所设定的虚拟机装入数据中心的服务器中;
8)将步骤2)中的每一个任务对应装入数据中心服务器中的一个虚拟机里。
步骤7)中是根据步骤5)中所述的对数据中心服务器的总能耗模型的约束进行的装箱的。
步骤8)中所述的将每一个任务对应装入数据中心服务器中的一个虚拟机里,是采用无差别顺序放置方法或者是采用如下所述的绑定方法进行:
(1)根据对角加边模型的分块情况以及任务的指令长度,将计算协调量后的执行过程中每一个块的计算步骤所代表的任务进行绑定,从而实现不同块的并行计算,节省计算时间;
(2)对于计算协调量前的各任务,将有输入输出关系的任务组放置在相邻虚拟机中,减 少交互数据量,降低通信开销。
本发明的对角加边模型分解协调计算的数据中心求解方法,能够将广泛应用于电力系统的对角加边模型分解协调算法通过数据中心中进行求解,与传统的在单机多进程和多机简单调度的计算模式相比,本发明能够缩短分解协调算法的计算时间和降低数据中心能耗,并且,随着电力节点网络计算规模和复杂度的增加,采用数据中心计算的优势就越明显。
附图说明
图1是对角加边模型分解协调计算的数据中心求解方法具体流程图;
图2是对角加边模型分解协调算法与数据中心映射过程图;
图3是对角加边模型分解协调算法的工作流图。
具体实施方式
下面结合实施例和附图对本发明的对角加边模型分解协调计算的数据中心求解方法做出详细说明。
本发明的对角加边模型分解协调计算的数据中心求解方法,包括如下步骤:
1)通过已有的网络分块法将电力网络进行划分得到电力网络的分块;
可以采用节点撕裂法,或支路切割法,或统一的网络分块法对电力网络进行分块。
2)将电力系统对角加边模型分解协调算法的计算流程中的每一个计算步骤设定为一个任务,确定整个计算过程的任务数;
3)根据对角加边模型分解协调算法的分块以及数据中心服务器的MIPS大小和内存大小,设定虚拟机的个数以及虚拟机中CPU的MIPS大小和内存大小,例如,在计算IEEE118节点图时,设定服务器的MIPS为2580,内存为512MB,设定虚拟机MIPS大小在[700,900]产生,虚拟内存为128MB;
4)计算数据中心的IT设备总能耗Etotal
Figure PCTCN2014087114-appb-000012
式中,Eserver为数据中心单台服务器的能耗,Eswitch为数据中心单个交换机的能耗,N为数据中心服务器的使用台数,T为数据中心交换机的使用个数,其中,
单台服务器的能耗模型为:
Figure PCTCN2014087114-appb-000013
式中:Pbaseline表示服务器的空载运行时的功耗,tmax表示服务器的持续运行时间,PVM表示服务器中一个虚拟机的功耗,tk表示服务器中一个虚拟机的运行时间,M表示服务器上虚拟机的数量;
单个交换机的能耗模型为:
Eswitch=Pswitchtmax      (3)
式中:Pswitch表示交换机运行功耗,tmax表示交换机运行时间;
5)由于数据中心服务器能耗占据IT设备总能耗的80%左右,根据分层序列法我们可以先求得服务器的能耗后,再计算交换机的能耗,最后得到数据中心的IT设备总能耗,因此可以使服务器的能耗降为最小,并将服务器的能耗模型归结为装箱模型,达到降低总能耗的目的。将服务器的能耗模型归结为装箱模型,并给出服务器的最小能耗模型:
虚拟机的集合为:
Figure PCTCN2014087114-appb-000014
Figure PCTCN2014087114-appb-000015
Figure PCTCN2014087114-appb-000016
分别表示单个虚拟机的MIPS大小和内存大小,单台服务器的容量为
Figure PCTCN2014087114-appb-000017
Figure PCTCN2014087114-appb-000018
Figure PCTCN2014087114-appb-000019
分别表示单台服务器的MIPS大小和内存大小,数据中心服务器的总能耗模型如下:
Figure PCTCN2014087114-appb-000020
对数据中心服务器的总能耗模型的约束如下:
Figure PCTCN2014087114-appb-000021
Figure PCTCN2014087114-appb-000022
Xj,i=0or1     (7)
Hi=0or1     (8)
式中,Xj,i=1表示虚拟机j放置于服务器i中,Xj,i=0表示虚拟机j未放置于服务器机i中;Hi=1表示服务器i被使用,Hi=0表示服务器i未被使用;
6)采用最佳适应算法或降序最佳适应算法对数据中心服务器的总能耗模型求解,其中:
最佳适应算法:最佳适应算法是按顺序依次将物品放入箱子,首先将第一个物品放入第一个箱子中,接着考虑第二个物品,如果第一个箱子能耗放下第二个物品,就将第二个物品放入第一个箱子,如果不能放下,就重新开启一个箱子,依次按顺序,当第i个物品放入时,将其放在所有能够容纳该物品并且放入之后剩余空间最小的箱子中,当所有的箱子都不满足要求时,重新开启一个箱子。
降序最佳适应算法:在进行装箱之前,先将物品按其大小进行降序排序,接着将第一个物品放入第一个箱子,然后考虑第二个物品,如果第一个箱子能放下第二个物品,就将第二个物品放入第一个箱子,如果不能放下,就重新开启一个箱子,依次按顺序,当第i个物品放入时,将其放在所有能够容纳该物品并且放入之后剩余空间最小的箱子中,当所有的箱子都不满足要求时,重新开启一个箱子;
7)将步骤3)中设定好的虚拟机装入数据中心的服务器中,具体是根据步骤5)中所述的对数据中心服务器的总能耗模型的约束进行的装机的;
8)将步骤2)中的每一个任务对应装入数据中心服务器中的一个虚拟机里,任务映射到数据中心服务器,如图2所示,所述的将每一个任务装入一个虚拟机里是采用无差别顺序放置方法或者是采用如下所述的绑定放置方法进行:
(1)如图3所示,根据对角加边模型的分块情况以及任务的指令长度,将计算协调量(Xt)后的执行过程中每一个块的计算步骤所代表的任务(3-乘法运算、2-加法运算和4-乘法运算)进行绑定,从而实现不同块的并行计算,节省计算时间;
(2)对于计算协调量(Xt)前的各任务,将有输入输出关系的任务组放置在相邻虚拟机中,减少交互数据量,降低通信开销;
具体可根据对角加边模型的分解协调算法的计算流程得到其工作流图,工作流图如图3所示,图中矩形是一个任务,表示一个计算步骤,椭圆是数据,指向任务的为该任务的输入数据,任务指出的箭头表示该任务计算输出结果,由于每个数据都连接某个任务的出箭头和另一个任务的入箭头,因此也就是这两个任务间的通信数据。
以IEEE118节点系统为例,将系统分成3个子区域,3个子区域的节点数分别为35、35、48,虚拟边界网络的节点数为7。得到运行时间和数据中心IT设备的总能耗:
(1)计算环境:数据中心为Fat-tree结构,包含54台服务器,45台6口的交换机。每台服务器配置为CPU:Pentium4(2.8GHz),内存:512Mb,基线功耗:145w,加载虚拟机运算功耗:10w,交换机选用华为S3552F-EA三层交换机,运行功率:54w。交换机端口速率在[0,8]Mb之间波动;
(2)根据电力系统节点网络的分块计算出每一步的指令长度和通信间的数据量大小,指令长度的计算公式为:指令长度=操作码的长度+操作数地址×操作数地址个数,计算的数据为浮点型,从而得到表一。
表一IEEE118节点系统数据
Figure PCTCN2014087114-appb-000023
Figure PCTCN2014087114-appb-000024
所述的数据中心IT设备的总能耗是根据步骤4)中给出的总能耗计算公式得到的。
虚拟机放入服务器采用了最佳适应算法和降序最佳适应算法,任务放入虚拟机采用了顺序放置盒绑定放置,因此组合产生四种算法,即无差别顺序放置-最佳适应算法,无差别顺序放置-降序最佳适应算法,绑定放置-最佳适应算法,绑定放置-降序最佳适应算法。采用本发明中所述的4种不同算法计算得到的计算时间和能耗。
表二IEEE118节点数据中心运算结果
Figure PCTCN2014087114-appb-000025

Claims (3)

  1. 一种对角加边模型分解协调计算的数据中心求解方法,其特征在于,包括如下步骤:
    1)通过已有的网络分块法将电力网络进行划分得到电力网络的分块;
    2)将电力系统对角加边模型分解协调算法的计算流程中的每一个计算步骤设定为一个任务,确定整个计算过程的任务数;
    3)获得数据中心服务器的MIPS大小和内存大小,设定虚拟机的个数以及虚拟机中CPU的MIPS大小和内存大小;
    4)计算数据中心的IT设备总能耗Etotal
    Figure PCTCN2014087114-appb-100001
    式中,Eserver为数据中心单台服务器的能耗,Eswitch为数据中心单个交换机的能耗,N为数据中心服务器的使用台数,T为数据中心交换机的使用个数,其中,
    单台服务器的能耗模型为:
    Figure PCTCN2014087114-appb-100002
    式中:Pbaseline表示服务器的空载运行时的功耗,tmax表示服务器的持续运行时间,PVM表示服务器中一个虚拟机的功耗,tk表示服务器中一个虚拟机的运行时间,M表示服务器上虚拟机的数量;
    单个交换机的能耗模型为:
    Eswitch=Pswitchtmax     (3)
    式中:Pswitch表示交换机运行功耗,tmax表示交换机运行时间;
    5)将服务器的能耗模型归结为装箱模型,并给出服务器的最小能耗模型:
    虚拟机的集合为:
    Figure PCTCN2014087114-appb-100003
    Figure PCTCN2014087114-appb-100004
    Figure PCTCN2014087114-appb-100005
    分别表示单个虚拟机的MIPS大小和内存大小,单台服务器的容量为
    Figure PCTCN2014087114-appb-100006
    Figure PCTCN2014087114-appb-100007
    Figure PCTCN2014087114-appb-100008
    分别表示单台服务器的MIPS大小和内存大小,数据中心服务器的总能耗模型如下:
    Figure PCTCN2014087114-appb-100009
    对数据中心服务器的总能耗模型的约束如下:
    Figure PCTCN2014087114-appb-100010
    Figure PCTCN2014087114-appb-100011
    Xj,i=0or1     (7)
    Hi=0or1     (8)
    式中,Xj,i=1表示虚拟机j放置于服务器i中,Xj,i=0表示虚拟机j未放置于服务器机i中;Hi=1表示服务器i被使用,Hi=0表示服务器i未被使用;
    6)采用最佳适应算法或降序最佳适应算法对数据中心服务器的总能耗模型求解;
    7)将步骤3)中所设定的虚拟机装入数据中心的服务器中;
    8)将步骤2)中的每一个任务对应装入数据中心服务器中的一个虚拟机里。
  2. 根据权利要求1所述的对角加边模型分解协调计算的数据中心求解方法,其特征在于,步骤7)中是根据步骤5)中所述的对数据中心服务器的总能耗模型的约束进行的装箱的。
  3. 根据权利要求1所述的对角加边模型分解协调计算的数据中心求解方法,其特征在于,步骤8)中所述的将每一个任务对应装入数据中心服务器中的一个虚拟机里,是采用无差别顺序放置方法或者是采用如下所述的绑定方法进行:
    (1)根据对角加边模型的分块情况以及任务的指令长度,将计算协调量后的执行过程中每一个块的计算步骤所代表的任务进行绑定,从而实现不同块的并行计算,节省计算时间;
    (2)对于计算协调量前的各任务,将有输入输出关系的任务组放置在相邻虚拟机中,减少交互数据量,降低通信开销。
PCT/CN2014/087114 2014-06-12 2014-09-22 对角加边模型分解协调计算的数据中心求解方法 WO2015188504A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US14/901,726 US10235341B2 (en) 2014-06-12 2014-09-22 Method for solving the decomposition-coordination calculation based on block bordered diagonal form (BBDF) model using data center

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410260630.2A CN104035868B (zh) 2014-06-12 2014-06-12 对角加边模型分解协调计算的数据中心求解方法
CN201410260630.2 2014-06-12

Publications (1)

Publication Number Publication Date
WO2015188504A1 true WO2015188504A1 (zh) 2015-12-17

Family

ID=51466641

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/087114 WO2015188504A1 (zh) 2014-06-12 2014-09-22 对角加边模型分解协调计算的数据中心求解方法

Country Status (3)

Country Link
US (1) US10235341B2 (zh)
CN (1) CN104035868B (zh)
WO (1) WO2015188504A1 (zh)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104035868B (zh) 2014-06-12 2016-08-24 天津大学 对角加边模型分解协调计算的数据中心求解方法
CN110968953B (zh) * 2019-11-29 2022-03-25 浙江大学 一种基于嵌套对角加边形式的电力系统暂态稳定仿真并行计算方法
CN110991034B (zh) * 2019-11-29 2022-03-22 浙江大学 基于全并行嵌套bbdf的电力系统暂态稳定仿真并行计算方法
CN111488052B (zh) * 2020-04-16 2022-03-08 中国工商银行股份有限公司 应用于物理机集群的容器启用方法和装置、计算机系统
CN112235859B (zh) * 2020-09-22 2022-08-05 国家卫星气象中心(国家空间天气监测预警中心) 一种基于多目标约束的动态能耗控制方法
CN115237241B (zh) * 2022-09-26 2022-12-09 张北云联数据服务有限责任公司 一种数据中心节能调度方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102567076A (zh) * 2011-12-12 2012-07-11 华中科技大学 一种基于能耗的服务器整合方案选择方法
US20120191439A1 (en) * 2011-01-25 2012-07-26 Power Analytics Corporation Systems and methods for automated model-based real-time simulation of a microgrid for market-based electric power system optimization
CN102662750A (zh) * 2012-03-23 2012-09-12 上海交通大学 基于弹性虚拟机池的虚拟机资源优化控制方法及其系统
CN103617455A (zh) * 2013-11-29 2014-03-05 广东电网公司电力科学研究院 基于虚拟机组子群的网厂两级负荷优化调度方法
CN104035868A (zh) * 2014-06-12 2014-09-10 天津大学 对角加边模型分解协调计算的数据中心求解方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8170856B2 (en) * 2006-04-12 2012-05-01 Power Analytics Corporation Systems and methods for real-time advanced visualization for predicting the health, reliability and performance of an electrical power system
CN101169743A (zh) * 2007-11-27 2008-04-30 南京大学 电力网格中基于多核计算机实现并行潮流计算的方法
CN103294521B (zh) * 2013-05-30 2016-08-10 天津大学 一种降低数据中心通信负载及能耗的方法
CN103296680A (zh) * 2013-06-09 2013-09-11 河海大学 一种互联电网经济调度的分布式计算方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120191439A1 (en) * 2011-01-25 2012-07-26 Power Analytics Corporation Systems and methods for automated model-based real-time simulation of a microgrid for market-based electric power system optimization
CN102567076A (zh) * 2011-12-12 2012-07-11 华中科技大学 一种基于能耗的服务器整合方案选择方法
CN102662750A (zh) * 2012-03-23 2012-09-12 上海交通大学 基于弹性虚拟机池的虚拟机资源优化控制方法及其系统
CN103617455A (zh) * 2013-11-29 2014-03-05 广东电网公司电力科学研究院 基于虚拟机组子群的网厂两级负荷优化调度方法
CN104035868A (zh) * 2014-06-12 2014-09-10 天津大学 对角加边模型分解协调计算的数据中心求解方法

Also Published As

Publication number Publication date
CN104035868B (zh) 2016-08-24
CN104035868A (zh) 2014-09-10
US10235341B2 (en) 2019-03-19
US20170083478A1 (en) 2017-03-23

Similar Documents

Publication Publication Date Title
WO2015188504A1 (zh) 对角加边模型分解协调计算的数据中心求解方法
Acun et al. Parallel programming with migratable objects: Charm++ in practice
CN104636187B (zh) 基于负载预测的numa架构中虚拟机调度方法
CN102981910B (zh) 虚拟机调度的实现方法和装置
WO2016062117A1 (zh) 虚拟机迁移处理方法及装置
CN106125888B (zh) 云数据中心中基于虚拟机迁移的资源利用高效的节能方法
CN107515663A (zh) 调整中央处理器内核运行频率的方法和装置
CN103473115B (zh) 虚拟机放置方法和装置
TW201324357A (zh) 虛擬機叢集之綠能管理方法
WO2023036144A1 (zh) 一种集成电路板仿真多级分布式并行计算方法
CN104123171A (zh) 基于numa架构的虚拟机迁移方法及系统
Quang-Hung et al. Epobf: energy efficient allocation of virtual machines in high performance computing cloud
Tang et al. CPU–GPU utilization aware energy-efficient scheduling algorithm on heterogeneous computing systems
Jena et al. Performance evaluation of energy efficient power models for digital cloud
CN108874508A (zh) 一种云计算虚拟服务器系统负载均衡调度方法
Alonso et al. Improving power efficiency of dense linear algebra algorithms on multi-core processors via slack control
Yang Key technologies and optimization for dynamic migration of virtual machines in cloud computing
CN104299170B (zh) 间歇性能源海量数据处理方法
CN109800084A (zh) 释放虚拟机资源的方法及终端设备
CN105049499A (zh) 一种基于多立方体映射的网络功能虚拟化资源分配方法
WO2024021475A1 (zh) 一种容器调度方法及装置
CN105208099A (zh) 一种云服务器内利用sdn技术智能节电的体系架构
CN103020008A (zh) 计算能力增强的可重构微服务器
Mikram et al. Server consolidation algorithms for cloud computing: taxonomies and systematic analysis of literature
CN104699520B (zh) 一种基于虚拟机迁移调度的节能方法

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 14901726

Country of ref document: US

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

Ref document number: 14894527

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: 14894527

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 14894527

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 19.10.2017)

122 Ep: pct application non-entry in european phase

Ref document number: 14894527

Country of ref document: EP

Kind code of ref document: A1