CN106095565B - Cloud computing system resource distribution backward inference system and configuration method based on backward stochastic differential equation - Google Patents

Cloud computing system resource distribution backward inference system and configuration method based on backward stochastic differential equation Download PDF

Info

Publication number
CN106095565B
CN106095565B CN201610363496.8A CN201610363496A CN106095565B CN 106095565 B CN106095565 B CN 106095565B CN 201610363496 A CN201610363496 A CN 201610363496A CN 106095565 B CN106095565 B CN 106095565B
Authority
CN
China
Prior art keywords
resource allocation
module
cloud computing
backward
historical data
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.)
Active
Application number
CN201610363496.8A
Other languages
Chinese (zh)
Other versions
CN106095565A (en
Inventor
吕宏武
郭盛开
王慧强
冯光升
郭方方
林俊宇
徐俊波
李冰洋
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.)
Harbin Engineering University
Original Assignee
Harbin Engineering 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 Harbin Engineering University filed Critical Harbin Engineering University
Priority to CN201610363496.8A priority Critical patent/CN106095565B/en
Publication of CN106095565A publication Critical patent/CN106095565A/en
Application granted granted Critical
Publication of CN106095565B publication Critical patent/CN106095565B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明属于云计算系统资源配置领域，具体涉及基于倒向随机微分方程的云计算系统资源配置逆向推理系统及配置方法。基于倒向随机微分方程的云计算资源配置系统，由用户处理请求模块、历史数据处理模块、逆向推理模块和用户交互模块组成，用户请求处理模块首先按照指定格式接收用户对云计算服务器的访问请求，以及对资源配置的约束条件，并根据云计算系统计算节点的配置和网络条件，将前述访问请求和约束条件解析为中央处理器需求、带宽需求、内存需求的参数。能够根据未来确定时刻的计算资源需求情况，确定当前需要的计算资源，并保证当前准备的计算资源是“最节省的”；能够积极应对未来资源配置的随机波动性，提高云计算系统的稳定性和可用性。

The invention belongs to the field of cloud computing system resource configuration, in particular to a cloud computing system resource configuration reverse reasoning system and configuration method based on a backward stochastic differential equation. The cloud computing resource allocation system based on the backward stochastic differential equation is composed of a user processing request module, a historical data processing module, a reverse reasoning module and a user interaction module. The user request processing module first receives the user's access request to the cloud computing server according to the specified format , and the constraints on resource configuration, and according to the configuration and network conditions of computing nodes in the cloud computing system, the aforementioned access requests and constraints are parsed into parameters of CPU requirements, bandwidth requirements, and memory requirements. It can determine the currently required computing resources according to the computing resource demand at a certain time in the future, and ensure that the currently prepared computing resources are "the most economical"; can actively deal with the random fluctuation of future resource allocation and improve the stability of the cloud computing system and availability.

Description

Cloud computing system resource distribution backward inference system based on backward stochastic differential equation And configuration method
Technical field
The invention belongs to cloud computing system resource distribution fields, and in particular to the cloud computing based on backward stochastic differential equation System resource configures backward inference system and configuration method.
Background technique
Cloud computing is that a large amount of computing resources (including narrow sense computing resource, storage resource and Internet resources etc.) are virtually turned to One shared resource pond, service needed for user is obtained by rent mode.The maximum feature of cloud computing is to use as needed, i.e. client It can apply for cloud resource or expansion/reduction resource at any time as needed, and operator completes related resource according to customer demand and matches It sets.The allowable resource that service provider provides is more, and the service experience and availability of user is higher, and reply pop-up mission is asked The ability asked is stronger.But assignable cloud computing resources are more, put into also more, expense is also bigger.Both how to balance It is faced with huge challenge.
Currently, the resource distribution of cloud computing system is faced with both sides challenge.On the one hand, the important spy of cloud computing system Property be randomness, either service request is initiated or resource distribution deployment, is filled with uncertain and chance phenomenon everywhere, not The resource distribution come often has stochastic volatility.This randomness and fluctuation must be thus considered during resource distribution Property.On the other hand, existing cloud computing system predicts that can future time point meet use only according to current computing resource The demand at family, and the problem of cloud computing system is increased sharply there may be pop-up mission and user at any time, do not ensure that user and clothes The service-level agreement SLA (Serice Level Agreement) that business provider signs is certain to be met just, does not generate resource Waste.It is then desired to which a kind of reverse inference method, determines that current time is least according to the resources requirement of future time instance Resource distribution amount.It is existing in terms of the method for cloud computing system resource distribution does not all also consider the problems of this.
" backward stochastic differential equation " theory is exactly the target according to future time instance, by the formulation of strategy gradually random The uncertain counteracting that fluctuation introduces, thus a kind of method that risk averse is fallen.It has set up between " random " and " determination " Bridge, allow people go to solve the problems, such as with determining strategy, method it is random uncertain, or random uncertain Thing carries out optimization processing.Backward stochastic differential equation at present " theory is mainly used for solving finance, economics and engineering science The practical problem in equal fields, does not apply to this field.
Summary of the invention
It pushes away the purpose of the present invention is to provide a kind of randomness feature for having fully considered cloud computing resources configuration and inversely The demand of reason, resource allocation proposal are more reasonable；The waste of cloud computing system resource distribution is advantageously reduced, resource distribution is promoted Efficiency saves the cloud computing system resource distribution backward inference system based on backward stochastic differential equation of cost.Of the invention Purpose, which also resides in, provides a kind of cloud computing system resource distribution backward inference method based on backward stochastic differential equation.
The object of the present invention is achieved like this:
Cloud computing resources based on backward stochastic differential equation configure system, handle request module, historical data by user Processing module, backward inference module and user interactive module composition, user's request processing module are received first, in accordance with specified format User is to the access request of cloud computing server, and to the constraint condition of resource distribution, and is calculated and is saved according to cloud computing system The configuration of point and network condition, aforementioned access is requested and constrained condition resolution is central processing unit demand, bandwidth demand, memory The parameter of demand, form are a triple ζ=(CPU (T), MEM (T), BW (T)), and wherein T is user and cloud service provider The future arranged in service-level agreement SLA determines the moment, and CPU (T) is the central processing unit demand at T moment, and MEM (T) is T The bandwidth demand at moment, BW (T) are the memory requirements at T moment, and triple ζ is passed to backward inference module；At historical data Manage on the one hand resource distribution historical data that module collects cloud computing system operation；Another aspect historical data processing module is collected Cloud computing system operation resource distribution historical data, determine the generating function G (*) of backward stochastic differential equation；History number According to processing module tool, there are two Boolean type flag bits: G (*) pre-sets flag bit PRECONFIG_G and updates flag bit UPDATE_FLAG determines G (*) analytic expression according to the generating function setting method of backward stochastic differential equation；By generating function G (*) passes to backward inference module；Backward inference module receives the parameter ζ from user's request processing module, and comes from In the generating function G (*) of historical data processing module；Using ζ as terminal condition, is established using generating function G (*) and be based on swinging to The resource distribution model of stochastic differential equation；According to the aforementioned resource distribution model based on backward stochastic differential equation, by swinging to The numerical solution of stochastic differential equation obtains one group of (Y (0), Z (0)) uniquely determined, and then show that cloud computing system initially provides Source configuration；The resource distribution condition (Y (0), Z (0)) of initial time is sent to user interactive module；Wherein Y (0), Z (0) points Not Wei initial time normal resource configuration and cope with the anticipating risk since resource random fluctuation is influenced and generated by environment Resource distribution；User interactive module receives the resource distribution condition of the initial time from backward inference module, and resource Allocation plan feeds back to user.
Cloud computing resources configuration method based on backward stochastic differential equation, includes the following steps:
(1) user's request processing module receives user to the access request of cloud computing server first, in accordance with specified format, And the constraint condition to resource distribution；And according to the configuration of cloud computing system calculate node and network condition, by aforementioned access Request and constraint condition resolution be central processing unit demand, bandwidth demand, memory requirements parameter, form be a triple ζ =(CPU (T), MEM (T), BW (T)), wherein T is that user and cloud service provider are arranged not in service-level agreement SLA Determine the moment, CPU (T) is the central processing unit demand at T moment, MEM (T) is the bandwidth demand at T moment, and BW (T) is the T moment Memory requirements；And ζ is passed to backward inference module；
(2) on the one hand historical data processing module collects the resource distribution historical data of cloud computing system operation；Another party The resource distribution historical data that face historical data processing module is run according to industry experience value or the cloud computing system of collection, really Determine the generating function G (*) of backward stochastic differential equation；There are two Boolean type flag bits for historical data processing module tool: G (*) It pre-sets flag bit PRECONFIG_G and updates flag bit UPDATE_FLAG, according to the generating function of backward stochastic differential equation Setting method determines G (*) analytic expression；And generating function G (*) is passed into backward inference module；
The generating function setting method of backward stochastic differential equation above-mentioned is specific further include:
(2.1) historical data processing module first checks for G (*) and pre-sets flag bit PRECONFIG_G, generates if 1 Function G (*) is set by industry experience value, is turned (2.7), is otherwise turned (2.2)；
(2.2) it checks and updates whether flag bit UPDATE_FLAG is 0, if UPDATE_FLAG is 0, turns (2.3), otherwise turn (2.4)；
(2.3) the resource distribution historical data that historical data processing module is run according to the cloud computing system of collection utilizes The generating function G (*) of one preset function set fitting backward stochastic differential equation, and UPDATE_FLAG is set as 1, The time Lasttime of record at this time simultaneously turns (2.7)；
(2.4) UPDATE_FLAG is 1 at this time, current time system time Nowtime is read, if Nowtime- Lasttime >=Interval, wherein Interval > 0 is default update cycle constant, turns (2.5), otherwise turns (2.6)；
(2.5) generating function G (*) is fitted again at this time, completes the update of G (*)；The time of record at this time simultaneously Lasttime turns (2.7)；
(2.6) at this point, Nowtime-Lasttime < Interval, records time Lasttime at this time, turn (2.7)；
(2.7) generating function G (*) is passed to backward inference module by historical data processing module；
(3) backward inference module receives the parameter ζ from user's request processing module, and at historical data Manage the generating function G (*) of module；And using ζ as terminal condition, is established using generating function G (*) and be based on swinging to stochastic differential side The resource distribution model of journey；And according to the aforementioned resource distribution model based on backward stochastic differential equation, by swinging to stochastic differential The numerical solution of equation obtains one group of (Y (0), Z (0)) uniquely determined, and then show that cloud computing system initial resource configures；And The resource distribution condition (Y (0), Z (0)) of initial time is sent to user interactive module；Wherein Y (0), Z (0) are respectively initial The normal resource at moment configures and copes with the anticipating risk resource distribution since resource random fluctuation is influenced and generated by environment；
Resource distribution model above-mentioned based on backward stochastic differential equation has the feature that
(3.1) the resource distribution model based on backward stochastic differential equation established meets following equation
Wherein, [0, T] t ∈, t is the time, and at the time of T is following determines, W is the Brownian movement of d dimension；
(3.2) resource configuration amount can be expressed as the processing capacity CPU (t), interior of central processing unit within [0, the T] period The capacity BW (t) of the capacity MEM (t), bandwidth that deposit；
(3.3) Y (t) is t moment resource distribution, i.e. Y (t)=(CPU (t), MEM (t), BW (t))；
(3.4) Z (t) is set as coping with since the resource for being influenced by environment and generating the anticipating risk of resource random fluctuation is matched It sets (the hereinafter referred to as resource distribution of anticipating risk), and wherein environment influences master to Z (t)=(CPU (t) ', MEM (t) ', BW (t) ') Refer to the factors such as pop-up mission request, the temperature surge of calculate node, main board power supply deficiency；Z (t) can be random with coping resources Disturbance, provides redundancy for resource distribution；
(3.5) generating function G (*) is the relation function of Y (t), Z (t), t, is matched according to the resource of each node of cloud computing system Historical data processing result or the setting of industry experience value are set, can be provided by historical data processing module；
(3.6) ζ is the terminal condition of backward stochastic differential equation, is one group of measurable stochastic variable, ζ is by user's request Reason module provides；
(4) user interactive module receives the resource distribution condition of the initial time from backward inference module, and money Source allocation plan feeds back to user.
The beneficial effects of the present invention are: (1) can according to the following computational resource requirements situation that determines moment, determination it be worked as The computing resource of preceding needs, and guarantee that the computing resource currently prepared is " most saving "；(2) Future can be responded actively The stochastic volatility of configuration improves the stability and availability of cloud computing system.
Detailed description of the invention
Fig. 1 is the module map of the cloud computing system resource distribution backward inference system based on backward stochastic differential equation；
Fig. 2 is the generating function setting method flow chart that the present invention implements the backward stochastic differential equation provided.
Specific embodiment
The present invention is described further with reference to the accompanying drawing.
The existing method about cloud computing system resource distribution can only configure the possibility shape for calculating future according to Current resource State, and present resource deployment demand cannot be calculated according to the random fluctuation in future with swinging to, this makes at analysis, calculating and place When managing cloud computing resources allocation problem, it not can guarantee the computing resource currently prepared and meet just.Base disclosed by the invention In the cloud computing system resource distribution backward inference system and method for backward stochastic differential equation, target is configured according to Future And random fluctuation, current resource allocation proposal is calculated by backward inference, advantageously reduces cloud computing system resource distribution Waste, promoted Allocation Efficiency, save cost.
Cloud computing resources configuration system of the present invention based on backward stochastic differential equation includes that user handles request 4 module compositions such as module, historical data processing module, backward inference module and user interactive module.
1, user's request processing module receives user to the access request of cloud computing server first, in accordance with specified format, with And the constraint condition to resource distribution.And according to the configuration of cloud computing system calculate node and network condition, aforementioned access is asked Summation constraint condition resolve to central processing unit demand, bandwidth demand, memory requirements parameter, form be a triple ζ= (CPU (T), MEM (T), BW (T)), wherein T is user and the future that cloud service provider is arranged in service-level agreement SLA Determine the moment, CPU (T) is the central processing unit demand at T moment, and MEM (T) is the bandwidth demand at T moment, and BW (T) is the T moment Memory requirements.And ζ is passed to backward inference module.
2, on the one hand historical data processing module collects the resource distribution historical data of cloud computing system operation；On the other hand The resource distribution historical data that historical data processing module is run according to industry experience value or the cloud computing system of collection, determines The generating function G (*) of backward stochastic differential equation.There are two Boolean type flag bits for historical data processing module tool: G (*) is pre- Flag bit PRECONFIG_G is set and updates flag bit UPDATE_FLAG, is set according to the generating function of backward stochastic differential equation Determine method, determines G (*) analytic expression.And generating function G (*) is passed into backward inference module.
The generating function setting method of backward stochastic differential equation above-mentioned is specific further include:
(1) historical data processing module first checks for G (*) and pre-sets flag bit PRECONFIG_G, generates letter if 1 Number G (*) is set by industry experience value, is turned (7), is otherwise turned (2).
(2) it checks and updates whether flag bit UPDATE_FLAG is 0, if UPDATE_FLAG is 0, turns (3), otherwise turn (4)；
(3) the resource distribution historical data that historical data processing module is run according to the cloud computing system of collection, utilizes one The generating function G (*) of a preset function set fitting backward stochastic differential equation, and UPDATE_FLAG is set as 1, together The time Lasttime of Shi Jilu at this time turns (7).
(4) UPDATE_FLAG is 1 at this time, reads current time system time Nowtime, if Nowtime-Lasttime > =Interval, wherein Interval > 0 is default update cycle constant, turns (5), otherwise turns (6).
(5) generating function G (*) is fitted again at this time, completes the update of G (*).The time of record at this time simultaneously Lasttime turns (7).
(6) at this point, Nowtime-Lasttime < Interval, records time Lasttime at this time, turn (7).
(7) generating function G (*) is passed to backward inference module by historical data processing module.
3, backward inference module receives the parameter ζ from user's request processing module, and at historical data Manage the generating function G (*) of module.And using ζ as terminal condition, is established using generating function G (*) and be based on swinging to stochastic differential side The resource distribution model of journey.And according to the aforementioned resource distribution model based on backward stochastic differential equation, by swinging to stochastic differential The numerical solution of equation obtains one group of (Y (0), Z (0)) uniquely determined, and then show that cloud computing system initial resource configures.And The resource distribution condition (Y (0), Z (0)) of initial time is sent to user interactive module.Wherein Y (0), Z (0) are respectively initial The normal resource at moment configures and copes with the anticipating risk resource distribution since resource random fluctuation is influenced and generated by environment.
Resource distribution model above-mentioned based on backward stochastic differential equation has the feature that
(1) the resource distribution model based on backward stochastic differential equation established meets following equation
Wherein, [0, T] t ∈, t are the time, and at the time of T is following determines, W is that (present invention is set as the Brownian movement of d dimension 3)。
(2) resource configuration amount can be expressed as the processing capacity CPU (t) of central processing unit, memory within [0, the T] period Capacity MEM (t), bandwidth capacity BW (t).
(3) Y (t) is t moment resource distribution, i.e. Y (t)=(CPU (t), MEM (t), BW (t)).
(4) Z (t) is set as coping with the resource distribution since the anticipating risk of resource random fluctuation is influenced and generated by environment (the hereinafter referred to as resource distribution of anticipating risk), and wherein environment influences mainly Z (t)=(CPU (t) ', MEM (t) ', BW (t) ') Refer to the factors such as pop-up mission request, the temperature surge of calculate node, main board power supply deficiency.Z (t) can be disturbed at random with coping resources It is dynamic, redundancy is provided for resource distribution.
(5) generating function G (*) is the relation function of Y (t), Z (t), t, according to the resource distribution of each node of cloud computing system Historical data processing result or the setting of industry experience value, can be provided by historical data processing module.
(6) ζ is the terminal condition of backward stochastic differential equation, is one group of measurable stochastic variable, and ζ is handled by user's request Module provides.
4, user interactive module receives the resource distribution condition of the initial time from backward inference module, and resource Allocation plan feeds back to user.
The example of this method be a simple cloud computing server system, service can occupy resource have CPU, bandwidth, Memory.
Cloud computing resources configuration system of the present invention based on backward stochastic differential equation includes that user handles request 4 module compositions such as module, historical data processing module, backward inference module and user interactive module.
1, user's request processing module receives user to the access request of cloud computing server first, in accordance with specified format, with And the constraint condition to resource distribution.And according to the configuration of cloud computing system calculate node and network condition, aforementioned access is asked Summation constraint condition resolve to central processing unit demand, bandwidth demand, memory requirements parameter, form be a triple ζ= (CPU (T), MEM (T), BW (T)), wherein T is user and the future that cloud service provider is arranged in service-level agreement SLA Determine the moment, CPU (T) is the central processing unit demand at T moment, and MEM (T) is the bandwidth demand at T moment, and BW (T) is the T moment Memory requirements.And ζ is passed to backward inference module.
2, on the one hand historical data processing module collects the resource distribution historical data of cloud computing system operation；On the other hand The resource distribution historical data that historical data processing module is run according to industry experience value or the cloud computing system of collection, determines The generating function G (*) of backward stochastic differential equation.There are two Boolean type flag bits for historical data processing module tool: G (*) is pre- Flag bit PRECONFIG_G is set and updates flag bit UPDATE_FLAG, is set according to the generating function of backward stochastic differential equation Determine method, determines G (*) analytic expression.And generating function G (*) is passed into backward inference module.
The generating function setting method of backward stochastic differential equation above-mentioned is specific further include:
(1) historical data processing module first checks for G (*) and pre-sets flag bit PRECONFIG_G, generates letter if 1 Number G (*) is set by industry experience value, is turned (7), is otherwise turned (2).
(2) it checks and updates whether flag bit UPDATE_FLAG is 0, if UPDATE_FLAG is 0, turns (3), otherwise turn (4)；
(3) the resource distribution historical data that historical data processing module is run according to the cloud computing system of collection, utilizes one The generating function G (*) of a preset function set fitting backward stochastic differential equation, and UPDATE_FLAG is set as 1, together The time Lasttime of Shi Jilu at this time turns (7).
(4) UPDATE_FLAG is 1 at this time, reads current time system time Nowtime, if Nowtime-Lasttime > =Interval, wherein Interval > 0 is default update cycle constant, turns (5), otherwise turns (6).
(5) generating function G (*) is fitted again at this time, completes the update of G (*).The time of record at this time simultaneously Lasttime turns (7).
(6) at this point, Nowtime-Lasttime < Interval, records time Lasttime at this time, turn (7).
(7) generating function G (*) is passed to backward inference module by historical data processing module.
3, backward inference module receives the parameter ζ from user's request processing module, and at historical data Manage the generating function G (*) of module.And using ζ as terminal condition, is established using generating function G (*) and be based on swinging to stochastic differential side The resource distribution model of journey.And according to the aforementioned resource distribution model based on backward stochastic differential equation, by swinging to stochastic differential The numerical solution of equation obtains one group of (Y (0), Z (0)) uniquely determined, and then show that cloud computing system initial resource configures.And The resource distribution condition (Y (0), Z (0)) of initial time is sent to user interactive module.Wherein Y (0), Z (0) are respectively initial The normal resource at moment configures and copes with the anticipating risk resource distribution since resource random fluctuation is influenced and generated by environment.
Resource distribution model above-mentioned based on backward stochastic differential equation has the feature that
(1) the resource distribution model based on backward stochastic differential equation established meets following equation
Wherein, [0, T] t ∈, t are the time, and at the time of T is following determines, W is the Brownian movement of 3 dimensions.
(2) resource configuration amount can be expressed as the processing capacity CPU (t) of central processing unit, memory within [0, the T] period Capacity MEM (t), bandwidth capacity BW (t).
(3) Y (t) is t moment resource distribution, i.e. Y (t)=(CPU (t), MEM (t), BW (t)).
(4) Z (t) is set as coping with the resource distribution since the anticipating risk of resource random fluctuation is influenced and generated by environment (the hereinafter referred to as resource distribution of anticipating risk), and wherein environment influences mainly Z (t)=(CPU (t) ', MEM (t) ', BW (t) ') Refer to the factors such as pop-up mission request, the temperature surge of calculate node, main board power supply deficiency.Z (t) can be disturbed at random with coping resources It is dynamic, redundancy is provided for resource distribution.
(5) generating function G (*) is the relation function of Y (t), Z (t), t, according to the resource distribution of each node of cloud computing system Historical data processing result or the setting of industry experience value, can be provided by historical data processing module.
(6) ζ is the terminal condition of backward stochastic differential equation, is one group of measurable stochastic variable, and ζ is handled by user's request Module provides.
4, user interactive module receives the resource distribution condition of the initial time from backward inference module, and resource Allocation plan feeds back to user.
This example can cover the present invention.By the description to this example it can be found that need to only input the following a certain determination Time cloud computing system resource distribution demand to be achieved, the cloud computing system resource that can reversely release initial time are matched It sets.It is provided by the invention to have the beneficial effect that: (1) can be determined according to the following computational resource requirements situation for determining the moment The computing resource currently needed, and guarantee that the computing resource currently prepared is " most saving "；(2) the following money can be responded actively The stochastic volatility of source configuration, improves the stability and availability of cloud computing system.

Claims (2)

1.基于倒向随机微分方程的云计算资源配置系统，由用户处理请求模块、历史数据处理模块、逆向推理模块和用户交互模块组成，其特征在于：用户处理请求模块首先按照指定格式接收用户对云计算服务器的访问请求，以及对资源配置的约束条件，并根据云计算系统计算节点的配置和网络条件，将前述访问请求和约束条件解析为中央处理器需求、带宽需求、内存需求的参数，形式为一个三元组ζ＝(CPU(T),MEM(T),BW(T))，其中T为用户与云服务提供商在服务等级协议SLA中约定的未来确定时刻，CPU(T)为T时刻的中央处理器需求，MEM(t)为内存的容量、BW(t)为带宽的容量，把三元组ζ传递给逆向推理模块；历史数据处理模块一方面收集云计算系统运行的资源配置历史数据；另一方面历史数据处理模块根据收集的云计算系统运行的资源配置历史数据，确定倒向随机微分方程的生成函数G(*)；历史数据处理模块具有两个布尔值型标志位：G(*)预设置标志位PRECONFIG_G和更新标志位UPDATE_FLAG，根据倒向随机微分方程的生成函数设定方法，确定G(*)解析式；生成函数G(*)为Y(t)、Z(t)、t的关系函数，根据云计算系统各节点的资源配置历史数据处理结果或行业经验值设定，可以由历史数据处理模块提供；Y(t)为t时刻资源配置，即Y(t)＝(CPU(t),MEM(t),BW(t))；Z(t)设为应对由于受到环境影响而产生资源随机波动的预防风险的资源配置；且Z(t)＝(CPU(t)’,MEM(t)’,BW(t)’)其中环境影响主要是指突发任务请求、计算节点的温度激增、主板供电不足，Z(t)可以应对资源随机扰动，为资源配置提供冗余；1. The cloud computing resource allocation system based on the backward stochastic differential equation is composed of a user processing request module, a historical data processing module, a reverse reasoning module and a user interaction module, and it is characterized in that: the user processing request module first receives the user's request according to a specified format. The access request of the cloud computing server, as well as the constraints on resource configuration, and according to the configuration and network conditions of the computing nodes of the cloud computing system, the aforementioned access requests and constraints are parsed into parameters of CPU requirements, bandwidth requirements, and memory requirements. The form is a triplet ζ=(CPU(T), MEM(T), BW(T)), where T is the future definite time agreed by the user and the cloud service provider in the service level agreement SLA, CPU(T) For the CPU demand at time T, MEM(t) is the memory capacity, BW(t) is the bandwidth capacity, and the triplet ζ is passed to the reverse reasoning module; The historical data of resource allocation; on the other hand, the historical data processing module determines the generation function G(*) of the backward stochastic differential equation according to the historical data of the resource allocation of the collected cloud computing system operation; the historical data processing module has two Boolean value type flags Bit: G(*) preset flag bit PRECONFIG_G and update flag bit UPDATE_FLAG, determine the analytical formula of G(*) according to the generating function setting method of the reverse stochastic differential equation; the generating function G(*) is Y(t), The relationship function of Z(t) and t can be provided by the historical data processing module according to the historical data processing results or industry experience values of resource allocation of each node of the cloud computing system; Y(t) is the resource allocation at time t, that is, Y (t)=(CPU(t), MEM(t), BW(t)); Z(t) is set as the resource allocation for preventing the risk of random fluctuation of resources due to environmental impact; and Z(t)= (CPU(t)', MEM(t)', BW(t)') The environmental impact mainly refers to the sudden task request, the temperature surge of the computing node, the insufficient power supply of the motherboard, Z(t) can deal with random disturbance of resources, Provide redundancy for resource allocation; 将生成函数G(*)传递给逆向推理模块；逆向推理模块接收来自于用户处理请求模块的参数ζ，以及来自于历史数据处理模块的生成函数G(*)；将ζ作为终端条件，采用生成函数G(*)建立基于倒向随机微分方程的资源配置模型；根据前述基于倒向随机微分方程的资源配置模型，由倒向随机微分方程的数值解法得出一组唯一确定的(Y(0),Z(0))，进而得出云计算系统初始资源配置；把初始时刻的资源配置条件(Y(0),Z(0))发送给用户交互模块；其中Y(0)、Z(0)分别为初始时刻的正常资源配置和应对由于受到环境影响而产生资源随机波动的预防风险资源配置；用户交互模块接收来自于逆向推理模块的初始时刻的资源配置条件，并把资源配置方案反馈给用户。Pass the generating function G(*) to the reverse reasoning module; the reverse reasoning module receives the parameter ζ from the user processing request module, and the generating function G(*) from the historical data processing module; The function G(*) establishes a resource allocation model based on the backward stochastic differential equation; according to the aforementioned resource allocation model based on the backward stochastic differential equation, a set of uniquely determined (Y(0) is obtained by the numerical solution of the backward stochastic differential equation. ), Z(0)), and then obtain the initial resource configuration of the cloud computing system; send the resource configuration conditions (Y(0), Z(0)) at the initial moment to the user interaction module; where Y(0), Z( 0) are the normal resource allocation at the initial moment and the risk-prevention resource allocation to deal with random fluctuations of resources due to environmental influence; the user interaction module receives the resource allocation conditions at the initial moment from the reverse reasoning module, and feeds back the resource allocation plan to users. 2.基于倒向随机微分方程的云计算资源配置方法，其特征在于，包括如下步骤：2. the cloud computing resource allocation method based on the backward stochastic differential equation, is characterized in that, comprises the following steps: (1)用户处理请求模块首先按照指定格式接收用户对云计算服务器的访问请求，以及对资源配置的约束条件；并根据云计算系统计算节点的配置和网络条件，将前述访问请求和约束条件解析为中央处理器需求、带宽需求、内存需求的参数，形式为一个三元组ζ＝(CPU(T),MEM(T),BW(T))，其中T为用户与云服务提供商在服务等级协议SLA中约定的未来确定时刻，并把ζ传递给逆向推理模块；(1) The user processing request module firstly receives the user's access request to the cloud computing server according to the specified format, as well as the constraints on the resource configuration; It is the parameters of CPU demand, bandwidth demand, and memory demand, in the form of a triple ζ=(CPU(T), MEM(T), BW(T)), where T is the service between the user and the cloud service provider. The future determination time agreed in the level agreement SLA, and pass ζ to the reverse reasoning module; (2)历史数据处理模块一方面收集云计算系统运行的资源配置历史数据；另一方面历史数据处理模块根据行业经验值或者收集的云计算系统运行的资源配置历史数据，确定倒向随机微分方程的生成函数G(*)；历史数据处理模块具有两个布尔值型标志位：G(*)预设置标志位PRECONFIG_G和更新标志位UPDATE_FLAG，根据倒向随机微分方程的生成函数设定方法，确定G(*)解析式；并将生成函数G(*)传递给逆向推理模块；(2) On the one hand, the historical data processing module collects the historical data of the resource allocation of the operation of the cloud computing system; on the other hand, the historical data processing module determines the backward stochastic differential equation according to the industry experience value or the collected historical data of the resource allocation of the operation of the cloud computing system The generation function G(*) of G(*) analytic expression; pass the generating function G(*) to the reverse reasoning module; 前述的倒向随机微分方程的生成函数设定方法具体还包括：The aforementioned method for setting the generating function of the inverse stochastic differential equation further includes: (2.1)历史数据处理模块首先检查G(*)预设置标志位PRECONFIG_G，若为1则生成函数G(*)已由行业经验值设定，转(2.7)，否则转(2.2)；(2.1) The historical data processing module first checks the G(*) preset flag bit PRECONFIG_G, if it is 1, the generation function G(*) has been set by the industry experience value, go to (2.7), otherwise go to (2.2); (2.2)检查更新标志位UPDATE_FLAG是否为0，若UPDATE_FLAG为0，转(2.3)，否则转(2.4)；(2.2) Check whether the update flag bit UPDATE_FLAG is 0, if UPDATE_FLAG is 0, go to (2.3), otherwise go to (2.4); (2.3)历史数据处理模块根据收集的云计算系统运行的资源配置历史数据，利用一个预设的函数集合拟合倒向随机微分方程的生成函数G(*)，并把UPDATE_FLAG设置为1，同时记录此时的时间Lasttime，转(2.7)；(2.3) The historical data processing module uses a preset function set to fit the generation function G(*) of the backward stochastic differential equation according to the collected historical data of the resource configuration of the cloud computing system, and sets UPDATE_FLAG to 1, and at the same time Record the time Lasttime at this time, turn (2.7); (2.4)此时UPDATE_FLAG为1，读取当前时刻系统时间Nowtime,若Nowtime-Lasttime>＝Interval，其中Interval>0为预设更新周期常数，转(2.5)，否则转(2.6)；(2.4) At this time, UPDATE_FLAG is 1, read the system time Nowtime at the current moment, if Nowtime-Lasttime>=Interval, where Interval>0 is the preset update cycle constant, go to (2.5), otherwise go to (2.6); (2.5)此时重新对生成函数G(*)进行拟合，完成G(*)的更新；同时记录此时的时间Lasttime，转(2.7)；(2.5) At this time, re-fit the generating function G(*) to complete the update of G(*); at the same time, record the time Lasttime at this time, and go to (2.7); (2.6)此时，Nowtime-Lasttime<Interval,记录此时的时间Lasttime，转(2.7)；(2.6) At this time, Nowtime-Lasttime<Interval, record the time Lasttime at this time, go to (2.7); (2.7)历史数据处理模块把生成函数G(*)传递给逆向推理模块；(2.7) The historical data processing module passes the generating function G(*) to the reverse reasoning module; (3)逆向推理模块接收来自于用户处理请求模块的参数ζ，以及来自于历史数据处理模块的生成函数G(*)；并将ζ作为终端条件，采用生成函数G(*)建立基于倒向随机微分方程的资源配置模型；并根据前述基于倒向随机微分方程的资源配置模型，由倒向随机微分方程的数值解法得出一组唯一确定的(Y(0),Z(0))，进而得出云计算系统初始资源配置；并把初始时刻的资源配置条件(Y(0),Z(0))发送给用户交互模块；其中Y(0)、Z(0)分别为初始时刻的正常资源配置和应对由于受到环境影响而产生资源随机波动的预防风险资源配置；(3) The reverse reasoning module receives the parameter ζ from the user processing request module, and the generation function G(*) from the historical data processing module; takes ζ as the terminal condition, and uses the generation function G(*) to establish a reverse-based A resource allocation model for stochastic differential equations; and according to the aforementioned resource allocation model based on backward stochastic differential equations, a set of uniquely determined (Y(0), Z(0)) is obtained by the numerical solution of backward stochastic differential equations, Then, the initial resource configuration of the cloud computing system is obtained; and the resource configuration conditions (Y(0), Z(0)) at the initial moment are sent to the user interaction module; where Y(0) and Z(0) are respectively the initial moment. Normal resource allocation and risk prevention resource allocation to deal with random fluctuations of resources due to environmental impacts; 前述的基于倒向随机微分方程的资源配置模型具有如下特征：The aforementioned resource allocation model based on inverse stochastic differential equations has the following characteristics: (3.1)建立的基于倒向随机微分方程的资源配置模型满足如下方程(3.1) The established resource allocation model based on inverse stochastic differential equation satisfies the following equation 其中，t∈[0,T]，t为时间，T为未来确定的时刻，W是d维的布朗运动；Among them, t∈[0,T], t is the time, T is the time determined in the future, and W is the d-dimensional Brownian motion; (3.2)在[0,T]时间段内资源配置量可以表示为中央处理器的处理容量CPU(t)、内存的容量MEM(t)、带宽的容量BW(t)；(3.2) In the [0, T] time period, the resource allocation amount can be expressed as the processing capacity CPU(t) of the central processing unit, the capacity MEM(t) of the memory, and the capacity BW(t) of the bandwidth; (3.3)Y(t)为t时刻资源配置，即Y(t)＝(CPU(t),MEM(t),BW(t))；(3.3) Y(t) is the resource configuration at time t, that is, Y(t)=(CPU(t), MEM(t), BW(t)); (3.4)Z(t)设为应对由于受到环境影响而产生资源随机波动的预防风险的资源配置，且Z(t)＝(CPU(t)’,MEM(t)’,BW(t)’)其中环境影响指突发任务请求、计算节点的温度激增、主板供电不足；Z(t)可以应对资源随机扰动，为资源配置提供冗余；(3.4) Z(t) is set as the resource allocation for preventing the risk of random fluctuation of resources due to environmental impact, and Z(t)=(CPU(t)', MEM(t)', BW(t)' ) where environmental impact refers to sudden task requests, temperature surges of computing nodes, and insufficient power supply of the motherboard; Z(t) can cope with random disturbances of resources and provide redundancy for resource allocation; (3.5)生成函数G(*)为Y(t)、Z(t)、t的关系函数，根据云计算系统各节点的资源配置历史数据处理结果或行业经验值设定，可以由历史数据处理模块提供；(3.5) The generating function G(*) is a relational function of Y(t), Z(t), and t. It can be set according to the historical data processing results or industry experience values of resource allocation of each node of the cloud computing system, and can be processed by historical data. module provides; (3.6)ζ为倒向随机微分方程的终端条件，为一组可测的随机变量，ζ由用户处理请求模块给出；(3.6) ζ is the terminal condition of the backward stochastic differential equation, which is a set of measurable random variables, ζ is given by the user processing request module; (4)用户交互模块接收来自于逆向推理模块的初始时刻的资源配置条件，并把资源配置方案反馈给用户。(4) The user interaction module receives the resource allocation conditions at the initial moment from the reverse reasoning module, and feeds back the resource allocation scheme to the user.
CN201610363496.8A 2016-05-27 2016-05-27 Cloud computing system resource distribution backward inference system and configuration method based on backward stochastic differential equation Active CN106095565B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610363496.8A CN106095565B (en) 2016-05-27 2016-05-27 Cloud computing system resource distribution backward inference system and configuration method based on backward stochastic differential equation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610363496.8A CN106095565B (en) 2016-05-27 2016-05-27 Cloud computing system resource distribution backward inference system and configuration method based on backward stochastic differential equation

Publications (2)

Publication Number Publication Date
CN106095565A CN106095565A (en) 2016-11-09
CN106095565B true CN106095565B (en) 2019-05-21

Family

ID=57230145

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610363496.8A Active CN106095565B (en) 2016-05-27 2016-05-27 Cloud computing system resource distribution backward inference system and configuration method based on backward stochastic differential equation

Country Status (1)

Country Link
CN (1) CN106095565B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10903894B1 (en) 2017-06-27 2021-01-26 Quantenna Communications, Inc. Mobile wireless repeater

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104266657A (en) * 2014-09-12 2015-01-07 海华电子企业(中国)有限公司 Shortest path planning parallelization method based on cooperative computing of CPU and MIC

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104266657A (en) * 2014-09-12 2015-01-07 海华电子企业(中国)有限公司 Shortest path planning parallelization method based on cooperative computing of CPU and MIC

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Parallel Option Pricing with BSDE method on GPU;Bin Dai 等;《2010 Ninth International Conference on Grid and Cloud Computing》;20101105;191-195
倒向随机微分方程及其应用;彭实戈;《数学进展》;19970430;第26卷(第2期);97-112
基于倒向随机微分方程的研发决策研究;刘宝生;《中国优秀硕士学位论文全文数据库 基础科学辑》;20120215(第2期);A002-355

Also Published As

Publication number Publication date
CN106095565A (en) 2016-11-09

Similar Documents

Publication Publication Date Title
Chaczko et al. Availability and load balancing in cloud computing
Kim et al. CometCloud: An autonomic cloud engine
US10783002B1 (en) Cost determination of a service call
US9342370B2 (en) Server migration
CN106453146B (en) Method, system, device and readable storage medium for allocating private cloud computing resources
WO2016148963A1 (en) Intelligent placement within a data center
CN105915535B (en) A kind of virtual resources access control method based on user identity
EP3547634A1 (en) Method and apparatus for determining access permission, and terminal
CN103281374B (en) A kind of method of data fast dispatch during cloud stores
CN101923615B (en) Grey fuzzy comprehensive evaluation-based trust quantization method
CN109711986A (en) A kind of token application and award method based on cloud storage plateform system
Perumal et al. Power‐conservative server consolidation based resource management in cloud
Qiu et al. Probabilistic demand allocation for cloud service brokerage
CN105991596A (en) Access control method and system
Cowdrey et al. Applying queueing theory for the optimization of a banking model
CN106095565B (en) Cloud computing system resource distribution backward inference system and configuration method based on backward stochastic differential equation
CN103399799B (en) Computational physics resource node load evaluation method and device in cloud operating system
CN109087209A (en) Community&#39;s dynamic adjusting method, device and server
Dechouniotis et al. A control‐theoretic approach towards joint admission control and resource allocation of cloud computing services
CN119065848B (en) Resource scheduling method and device based on cloud service platform
CN113885794B (en) Data access method and device based on multi-cloud storage, computer equipment and medium
Fan et al. An adaptive feedback load balancing algorithm in HDFS
CN106533839A (en) Application node monitoring method and apparatus of hybrid cloud
CN105335362B (en) The processing method and system of real time data, instant disposal system for treating
Tchernykh et al. Toward digital twins' workload allocation on clouds with low-cost microservices streaming interaction

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant