CN110083449A - The method, apparatus and computing module of dynamic assigning memory and processor - Google Patents
The method, apparatus and computing module of dynamic assigning memory and processor Download PDFInfo
- Publication number
- CN110083449A CN110083449A CN201910276015.3A CN201910276015A CN110083449A CN 110083449 A CN110083449 A CN 110083449A CN 201910276015 A CN201910276015 A CN 201910276015A CN 110083449 A CN110083449 A CN 110083449A
- Authority
- CN
- China
- Prior art keywords
- memory
- processor
- communication connection
- connection relationship
- computing module
- 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.)
- Granted
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
- G06F9/5016—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 the resource being the memory
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5017—Task decomposition
Abstract
The application provides the method, apparatus and computing module of a kind of dynamic assigning memory and processor.The method of the dynamic assigning memory and processor obtains the first calculating task first.And first calculating task is divided to form multiple subtasks.It include multiple efficient parallel calculating subtasks and multiple non-effective parallel computation subtasks in the multiple subtask.Secondly, according to the multiple efficient parallel calculating subtask, and according to the multiple non-effective parallel computation subtask, the relationship of memory and processor is established, to form multiple first calculate nodes and multiple second calculate nodes.The method of the dynamic assigning memory and processor realizes dynamic assigning memory and processor.When carrying out parallel computation, the method for the dynamic assigning memory and processor can make full use of computing resource, improve calculating speed according to different the first calculate nodes of calculating task dynamic configuration and the second calculate node.
Description
Technical field
This application involves field of computer technology, more particularly to a kind of method of dynamic assigning memory and processor, dress
It sets and computing module.
Background technique
As data magnanimity increases severely, single computer cannot store mass data.Therefore high-performance computer is made
For one kind of parallel computer, have received widespread attention.The node for being used to be calculated in high-performance computer is known as calculating section
Point.Calculate node can be divided into the first calculate node and the second calculate node by computing capability.First calculate node can be used for
Calculate the high calculating task of parallel efficiency.Second calculate node can be used for calculate cannot effectively distributed parallel calculating appoint
Business.
The first calculate node and the second calculate node of traditional high-performance computer have been fixed in structure.But it is different
The parallelization degree of calculating task is different, and required calculate node is not also identical.Traditional high-performance computer is being calculated
When, computing resource cannot be made full use of.
Summary of the invention
Based on this, it is necessary to for traditional high-performance computer when being calculated, computing resource cannot be made full use of
The problem of, the method, apparatus and computing module of a kind of dynamic assigning memory and processor are provided.
A kind of method of method of dynamic assigning memory and processor, the dynamic assigning memory and processor is applied to dynamic
The device of the device of state storage allocation and processor, the dynamic assigning memory and processor include control module and with the control
The computing module of module bus connection processed；The computing module include at least one processor and with memory described at least one
Multiple processors of bus connection；
The described method includes:
S10 obtains the first calculating task by the control module, and is divided first calculating task with shape
Subtask and multiple non-effective parallel computations at multiple subtasks, in the multiple subtask including multiple efficient parallel calculating
Subtask；
S20 obtains the mission bit stream of the multiple subtask by the control module；
S30 is established in the computing module by the control module and is deposited described at least one according to the mission bit stream
Communication connection relationship between reservoir and the processor；
S40 is completed according to the communication connection relationship to the memory and the processor in the computing module
It dynamically distributes, to form multiple first calculate nodes and multiple second calculate nodes.
The computing module includes a memory and connect with the memory bus in one of the embodiments,
Multiple processors；
The S30 establishes at least one institute in the computing module by the control module according to the mission bit stream
The step of stating the communication connection relationship between memory and the processor include:
The control module establishes the first communication connection relationship, and the first communication connection relationship is multiple processors
Between communication connection relationship；
The S40, according to the communication connection relationship, complete in the computing module the memory and the place
It manages device to dynamically distribute, includes: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module obtains the first communication connection relationship, and will be described according to the first communication connection relationship
Multiple processors are divided into multiple first processing units and multiple the second processing unit, each described first processing units is for locating
Manage the efficient parallel calculating subtask, each described the second processing unit for handle one it is described it is non-effective simultaneously
The subtask that row calculates.
The memory includes multiple storage units in one of the embodiments,；
The S30 establishes at least one institute in the computing module by the control module according to the mission bit stream
The step of stating the communication connection relationship between memory and the processor further include:
The control module establishes the second communication connection relationship, and the second communication connection relationship is that the multiple storage is single
First communication connection relationship between the first processing units and described the second processing unit respectively；
The S40, according to the communication connection relationship, complete in the computing module the memory and the place
It manages device to dynamically distribute, includes: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module obtains the second communication connection relationship, and will be each according to the second communication connection relationship
It is communicated to connect between a first processing units and a storage unit, it is each to form multiple first calculate nodes
A first calculate node is used to calculate the subtask of the efficient parallel calculating；
The computing module also according to it is described second communication connection relationship by each described the second processing unit with it is described
It is communicated to connect between a storage unit in multiple storage units in remaining multiple storage units, to form multiple described
Two calculate nodes, each described second calculate node are used to calculate the subtask of a non-effective parallel computation.
The S20 in one of the embodiments, is believed by the task that the control module obtains the multiple subtask
After the step of breath further include:
The control module determines each of the multiple subtask according to the mission bit stream of the multiple subtask
The size of memory needed for subtask；
The size of memory needed for the computing module obtains each described described subtask, and according to each described institute
The memory is divided into multiple storage units by the size for stating memory needed for subtask.
The computing module includes multiple computing units in one of the embodiments, each described computing unit packet
Include a processor and a memory connecting with the processor communication；
The S30 establishes at least one institute in the computing module by the control module according to the mission bit stream
The step of stating the communication connection relationship between memory and the processor include:
The control module establishes the first communication connection relationship, and the first communication connection relationship is each described calculating
The communication connection relationship between memory and processor in unit；
The S40, according to the communication connection relationship, complete in the computing module the memory and the place
It manages device to dynamically distribute, includes: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module obtains the first communication connection relationship, and will be multiple according to the first communication connection relationship
The communication connection between memory and processor in the computing unit, and ensure the institute at least one described computing unit
Processor is stated not work.
The S30 in one of the embodiments, according to the mission bit stream, by described in control module foundation
The step of communication connection relationship in computing module between at least one described memory and the processor further include:
The control module establishes the second communication connection relationship, and the second communication connection relationship is that the multiple storage is single
Communication connection relationship between member；
The S40, according to the communication connection relationship, complete in the computing module the memory and the place
It manages device to dynamically distribute, includes: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module obtains the second communication connection relationship, and will be multiple according to the second communication connection relationship
The storage unit communication connection, to form multiple first calculate nodes and multiple second calculate nodes.
The S40 in one of the embodiments, is completed according to the communication connection relationship in the computing module
The memory and the processor dynamically distribute, to form the step of multiple first calculate nodes and multiple second calculate nodes
Include: after rapid
An efficient parallel calculating is calculated by the first calculate node of each of the multiple first calculate node
Subtask, and a non-effective parallel computation is calculated by the second calculate node of each of the multiple second calculate node
Subtask；
After the completion of first calculating task, the communication connection relationship in the computing module is released.
A kind of computing module, for obtaining communication connection relationship, the computing module includes: the computing module
One memory；
Multiple processors, each described processor are connect with the memory bus；
Wherein, the memory includes multiple storage units, and the computing module is completed according to the communication connection relationship
To in the computing module the memory and the processor dynamically distribute, to form multiple first calculate nodes and multiple
Second calculate node.
The communication connection relationship includes the first communication connection relationship and the second communication connection in one of the embodiments,
Relationship, communication connection relationship of the first communication connection relationship between multiple processors, second communication connection
Relationship is the communication connection between the first processing units and described the second processing unit respectively of the multiple storage unit
Relationship.
A kind of computing module, for obtaining communication connection relationship, the computing module includes: the computing module
Multiple computing units communicate to connect between the multiple computing unit；
Each described computing unit includes a processor and a memory connecting with the processor communication；
Wherein, the computing module is completed according to the communication connection relationship to the storage in the computing module
Device and the processor dynamically distribute, to form multiple first calculate nodes and multiple second calculate nodes.
The communication connection relationship includes the first communication connection relationship and the second communication connection in one of the embodiments,
Relationship, the first communication connection relationship are the communication connection between memory and processor in each described computing unit
Relationship, communication connection relationship of the second communication connection relationship between the multiple storage unit.
A kind of device of dynamic assigning memory and processor, comprising:
Computing module described in any one of above-described embodiment；And
Control module is connect with the computing module bus.
The application provides the method, apparatus and computing module of a kind of dynamic assigning memory and processor.The dynamic allocation
The method of memory and processor obtains the first calculating task first.And first calculating task is divided multiple to be formed
Subtask.It include multiple efficient parallel calculating subtasks and multiple non-effective parallel computation subtasks in the multiple subtask.
Secondly, according to the multiple efficient parallel calculating subtask, and according to the multiple non-effective parallel computation subtask, foundation is deposited
The relationship of reservoir and processor, to form multiple first calculate nodes and multiple second calculate nodes.The dynamic assigning memory
Dynamic assigning memory and processor are realized with the method for processor.When carrying out parallel computation, the dynamic assigning memory and
The method of processor can make full use of according to different the first calculate nodes of calculating task dynamic configuration and the second calculate node
Computing resource improves calculating speed.
Detailed description of the invention
Fig. 1 is the method flow diagram of dynamic assigning memory and processor that the application one embodiment provides；
Fig. 2 is a kind of computing module structure chart that the application one embodiment provides；
Fig. 3 is a kind of computing module structure chart that the application one embodiment provides；
Fig. 4 is the device figure of a kind of dynamic assigning memory and processor that the application one embodiment provides.
Main element drawing reference numeral explanation
The device 100 of dynamic assigning memory and processor
Control module 10
Computing module 20
Memory 21
Storage unit 211
Processor 22
First processing units 221
The second processing unit 222
Computing unit 23
Specific embodiment
In order to make the above objects, features, and advantages of the present application more apparent, with reference to the accompanying drawing to the application
Specific embodiment be described in detail.Many details are explained in the following description in order to fully understand this Shen
Please.But the application can be implemented with being much different from other way described herein, those skilled in the art can be not
Similar improvement is done in the case where violating the application intension, therefore the application is not limited by following public specific implementation.
It should be noted that it can directly on the other element when element is referred to as " being set to " another element
Or there may also be elements placed in the middle.When an element is considered as " connection " another element, it, which can be, is directly connected to
To another element or it may be simultaneously present centering elements.
Unless otherwise defined, all technical and scientific terms used herein and the technical field for belonging to the application
The normally understood meaning of technical staff is identical.The term used in the description of the present application is intended merely to description tool herein
The purpose of the embodiment of body, it is not intended that in limitation the application.Term " and or " used herein includes one or more phases
Any and all combinations of the listed item of pass.
Referring to Figure 1, the application one embodiment provides the method for a kind of dynamic assigning memory and processor.The dynamic
The method of storage allocation and processor is applied to the device 100 of dynamic assigning memory and processor.The dynamic assigning memory and
The device 100 of processor includes control module 10 and the computing module 20 connecting with 10 bus of control module.The calculating
Module 20 include at least one processor 21 and with 21 bus of memory is connect described at least one multiple processors 22.It is described
The method of dynamic assigning memory and processor includes:
S10 obtains the first calculating task by the control module 10.And by the control module 10 to described first
Calculating task is divided to form multiple subtasks, includes the subtask of multiple efficient parallel calculating in the multiple subtask
With the subtask of multiple non-effective parallel computations.
In step S10, the subtask of the efficient parallel calculating and described non-can be divided by the complexity of subtask
The subtask of efficient parallel calculating.The complexity of the subtask can be by logical between each subtask and other subtasks
Letter requires and the size of each subtask is judged.In an alternative embodiment, first calculating task can be drawn
Divide to form m subtask.It include the subtask and a non-effective parallel meters of p of n efficient parallel calculating in the m subtask
The subtask of calculation.Wherein, the m is the positive integer more than or equal to 1, and described n, p are the integer more than or equal to 0
S20 obtains the mission bit stream of the multiple subtask by the control module 10.
In step S20, the control module 10 can be management node in distributed computer.The multiple subtask
Mission bit stream may include the mission bit stream of the subtask of each efficient parallel calculating.The task of the multiple subtask
Information can also include the mission bit stream of the subtask of the non-effective parallel computation.The mission bit stream may include task institute
Need the size of memory and the complexity of task.
S30 establishes at least one institute in the computing module 20 by the control module 10 according to the mission bit stream
State the communication connection relationship between memory 21 and the processor 22.
In step S30, the communication connection relationship may include the processor 22 and other described processors 22
Between communication connection relationship.The communication connection relationship can also include between the processor 22 and the memory 21
Communication connection relationship.
S40, according to the communication connection relationship, complete in the computing module 20 the memory 21 and the place
It manages device 22 to dynamically distribute, to form multiple first calculate nodes and multiple second calculate nodes.
In step S40, first calculate node has less processor cores compared to second calculate node
With the memory of less capacity.The relationship of the memory 21 and processor 22 can be in one-to-one, one-to-many or multi-to-multi
It is one or more.First calculate node can be CPU calculate node.Second calculate node can be fat node.Institute
Stating the first calculate node may include two processors.Second calculate node may include 4 processors.First meter
Operator node is used to calculate the subtask of the efficient parallel calculating.Second calculate node is described non-effective parallel for calculating
The subtask of calculating.
In the present embodiment, the method for the dynamic assigning memory and processor obtains the first calculating task first.And to institute
The first calculating task is stated to be divided to form multiple subtasks.It include multiple efficient parallel calculating in the multiple subtask
Task and multiple non-effective parallel computation subtasks.Secondly, according to the multiple efficient parallel calculating subtask, and according to described
The relationship of the memory 21 and the processor 22 is established, to form multiple first in multiple non-effective parallel computation subtasks
Calculate node and multiple second calculate nodes.The method of the dynamic assigning memory and processor realize dynamic assigning memory and
Processor.When carrying out parallel computation, the method for the dynamic assigning memory and processor is according to different calculating task dynamics
The first calculate node and the second calculate node are configured, computing resource can be made full use of, improves calculating speed.
In one of the embodiments, the computing module 20 include a memory 21 and with 21 bus of memory
Multiple processors 21 of connection.The S30 establishes the calculating mould by the control module 10 according to the mission bit stream
The step of communication connection relationship in block 20 between at least one described memory 21 and the processor 22 can be the control
Molding block 10 establishes the first communication connection relationship, communication link of the first communication connection relationship between multiple processors
Connect relationship.
The S40, according to the communication connection relationship, complete in the computing module 20 the memory 21 and institute
The dynamic allocation of processor 22 are stated, it can be in terms of described by the step of forming multiple first calculate nodes and multiple second calculate nodes
It calculates module 20 and obtains the first communication connection relationship.And the computing module 20 will according to the first communication connection relationship
The multiple processor 22 divides for multiple first processing units 221 and multiple the second processing unit 222, each is at described first
Reason unit 221 is used to handle the subtask of the efficient parallel calculating, each described the second processing unit 222 is for locating
Manage the subtask of a non-effective parallel computation.
In an alternative embodiment, the computing module 20 includes a memory 21 and a processor 22.It is each
A processor 22 is connect with 21 bus of memory.The n processors are chosen from a processor 22
22, so that processor 22 described in each of described n processor 22 forms a first processing units 221.Described in one
First processing units 221 are with other first processing units 221 without communicating.One first processing units 221 is based on
Calculate the subtask of efficient parallel calculating.
In an alternative embodiment, according to the mission bit stream, confirm that the multiple non-effective parallel computation is appointed
The number of the processor 22 needed for subtask described in each of business.It can be appointed according to the son of each non-effective parallel computation
The size of business determines the number of the processor 22 needed for each described subtask.According to institute needed for each described subtask
The number for stating processor 22 establishes in the computing module 20 first between the processor 22 by the control module 10
Communication connection relationship.
In an alternative embodiment, when the number of the processor 22 needed for a subtask is x, foundation
The first communication connection relationship, chooses the x processors 22 from a processor 22, and makes the x places
It manages device 22 and establishes communication, to form a second processing unit 222.The x is the positive integer greater than 1.Each subtask
The number of the required processor 22 can be identical.The number of the processor 22 can not also phase needed for each subtask
Together.
In the present embodiment, the first communication connection between the processor 22 can establish by the mission bit stream and close
System.Multiple first processing units 221 and multiple the second processing unit 222 can be formed according to the first communication connection relationship.
The formation of the multiple first processing units 221 and the multiple the second processing unit 222 can be dynamic allocation memory 21
It lays a good foundation with processor 22.
The S20 in one of the embodiments, obtains the task of the multiple subtask by the control module 10
After the step of information further include:
The control module 10 determines each in the multiple subtask according to the mission bit stream of the multiple subtask
The size of a sub- required by task memory.The size of memory needed for the computing module 20 obtains each described described subtask,
And the size of the memory according to needed for each described described subtask, the memory 21 is divided into multiple storage units 211.
In an alternative embodiment, the memory 21 can be divided into m storage unit 211.
The memory 21 refers to the memory in server system in each server.The processor 22 refers to service
Processor 22 in device system in each server.It is communicated to connect between a processor 22.The memory 21 is drawn
The method for being divided into m storage unit 211 can deposit for the memory 21 is divided into m by logic chip or logical device
Storage unit 211.Each storage unit 211 is not associated with other storage units 211.
In the present embodiment, a memory 21 is divided into multiple storage units 211.The multiple storage unit 211
Between without communication.The multiple storage unit 211 is to divide calculate node according to calculating task to provide the foundation.
The S30 in one of the embodiments, establishes institute by the control module 10 according to the mission bit stream
The step of stating the communication connection relationship in computing module 20 between at least one described memory 21 and the processor 22 may be used also
Think that the control module 10 establishes the second communication connection relationship.The second communication connection relationship is the multiple storage unit
211 communication connection relationships between the first processing units 221 and described the second processing unit 222 respectively.
The S40, according to the communication connection relationship, complete in the computing module 20 the memory 21 and institute
The dynamic allocation of processor 22 are stated, the step of multiple first calculate nodes and multiple second calculate nodes can also be described to be formed
Computing module 20 obtains the second communication connection relationship.And the computing module 20 is according to the second communication connection relationship
It will be communicated to connect between each first processing units 221 and a storage unit 211.Each described first processing is single
Communication connection can form multiple first calculate nodes between member 221 and a storage unit 211.Each described first
Calculate node is used to calculate the subtask of the efficient parallel calculating.
The computing module 20 also according to the second communication connection relationship by each described the second processing unit 222 with
It is communicated to connect between a storage unit 211 in the multiple storage unit 211 in remaining multiple storage units 211.Often
One described the second processing unit 222 is deposited with one in multiple storage units 211 remaining in the multiple storage unit 211
Communication connection can form multiple second calculate nodes between storage unit 211.Each described second calculate node is used for
Calculate the subtask of a non-effective parallel computation.
In the present embodiment, by the communication established between the first processing units 221 and storage unit 211, to be formed
First calculate node.By the communication established between described the second processing unit 222 and storage unit 211, to be formed in terms of second
Operator node.The method of the dynamic assigning memory and processor according to different the first calculate node of calculating task dynamic configuration and
Second calculate node can make full use of computing resource, improve calculating speed.
The computing module 20 includes multiple computing units 23 in one of the embodiments,.Each described calculating is single
Member 23 includes a processor 22 and a memory 21 communicated to connect with the processor 22.In an alternative embodiment
In, the computing module 20 may include b computing unit 23.It is communicated to connect between the b computing unit 23.Each institute
Stating computing unit 23 includes a processor 22 and a memory 21 communicated to connect with the processor 22.
The S30 is established at least one in the computing module 20 according to the mission bit stream by the control module 10
The step of communication connection relationship between a memory 21 and the processor 22, can establish for the control module 10
First communication connection relationship, the first communication connection relationship are memory 21 and processing in each described computing unit 23
Communication connection relationship between device 22.
The S40, according to the communication connection relationship, complete in the computing module 20 the memory 21 and institute
The dynamic allocation of processor 22 are stated, it can be in terms of described by the step of forming multiple first calculate nodes and multiple second calculate nodes
It calculates module 20 and obtains the first communication connection relationship.And the computing module 20 will according to the first communication connection relationship
The communication connection between memory 21 and processor 22 in multiple computing units 23.And the computing module 20 ensure to
The processor 22 in a few computing unit 23 does not work.
In an alternative embodiment, according to the mission bit stream, each of the multiple subtask is determined
The size of required by task memory, and the number of processor 22 needed for the subtask of each of determining the multiple subtask.It can
The required size of memory and the number of the required processor 22 are determined with the size by the subtask.To establish
First communication connection relationship.
In the present embodiment, the first communication connection between the processor 22 can establish by the mission bit stream and close
System.It will be logical between the memory 21 and processor 22 in multiple computing units 23 according to the first communication connection relationship
Letter connection.And the computing module 20 ensures that the processor 22 at least one described computing unit 23 does not work.It is described
The communication connection between memory 21 and processor 22 in multiple computing units 23 can be dynamic allocation memory 21
It lays a good foundation with processor 22.
The S30 in one of the embodiments, establishes institute by the control module 10 according to the mission bit stream
The step of stating the communication connection relationship in computing module 20 between at least one described memory 21 and the processor 22 can be with
The second communication connection relationship is established for the control module 10.The second communication connection relationship is the multiple storage unit
Communication connection relationship between 211.
The S40, according to the communication connection relationship, complete in the computing module 20 the memory 21 and institute
The dynamic allocation of processor 22 are stated, it can be in terms of described by the step of forming multiple first calculate nodes and multiple second calculate nodes
It calculates module 20 and obtains the second communication connection relationship.And the computing module 20 will be more according to the second communication connection relationship
A storage unit 211 communicates to connect, to form multiple first calculate nodes and multiple second calculate nodes.
According to the place needed for the size according to memory needed for subtask described in each and each described subtask
The number of device 22 is managed, the control module 10 establishes the second communication connection relationship.It can be with according to the second communication connection relationship
The multiple computing unit 23 is set to form multiple zonings.The computing unit 23 for forming the multiple zoning does not include place
Manage the idle computing unit 23 of device 22.The quantity of multiple computing units 23 can not phase in the multiple zoning
Deng.According to the second communication connection relationship can also by the multiple zoning with do not work in the processor 22 connect
The memory 21 connect communicates to connect.
In the present embodiment, by controlling the correspondence between the computing unit 23, to be formed in terms of multiple described first
Operator node and multiple second calculate nodes.According to the first calculate node described in different calculating task dynamic configurations and described
Second calculate node can make full use of computing resource, improve calculating speed.
The S40 in one of the embodiments, is completed according to the communication connection relationship to the computing module 20
In the memory 21 and the processor 22 dynamically distribute, with formed multiple first calculate nodes and it is multiple second calculate section
Include: after the step of point
An efficient parallel calculating is calculated by the first calculate node of each of the multiple first calculate node
Subtask.And a non-effective parallel computation is calculated by the second calculate node of each of the multiple second calculate node
Subtask.After the completion of first calculating task, the communication connection relationship in the computing module 20 is released.Release institute
State the corresponding relationship of memory 21 and the processor 22.According to different calculating tasks, the different memories 21 is established
With the corresponding relationship of the processor 22, so that the first calculate node of dynamic configuration and the second calculate node, can make full use of
Computing resource improves calculating speed.
Fig. 2 is referred to, the application one embodiment provides a kind of computing module 20.The computing module 20 includes one and deposits
Reservoir 21 and a processor 22.
One memory 21 includes m storage unit 211.Determine each of the multiple subtask subtask
The size of required memory.According to the size of memory needed for each described subtask, the memory 21 is divided into m storage
Unit 211.The m is the positive integer more than or equal to 1.Processor 21 described in each of described a processor 22 with it is described
The connection of 21 bus of memory.
The memory 21 refers to the memory in server system in each server.The processor 22 refers to service
Processor 22 in device system in each server.It is communicated to connect between a processor 22.The memory 21 is drawn
The method for being divided into m storage unit 211 can deposit for the memory 21 is divided into m by logic chip or logical device
Storage unit 211.Each storage unit 211 is not associated with other storage units 211.
The n processors 22 are chosen from a processor 22, so that each of described n processor 22
The processor 22 forms a first processing units 221.One first processing units 221 and other first processing units
221 without communication.It establishes and communicates between each described first processing units 221 and a storage unit 211, to form one
A first calculate node.Each described first calculate node is used to calculate the subtask of the efficient parallel calculating.
Multiple processors 22 in a processor 22 establish communication, to form a second processing unit 222.Often
It establishes and communicates between one described the second processing unit 222 and a storage unit 211, to form second calculate node.
The number of processor 22 in each second node can be with the number phase of the processor 22 in second node described in other
Together.The number of processor 22 in each second node can also be with of the processor 22 in second node described in other
Number is not identical.Each described second calculate node is used to calculate the subtask of a non-effective parallel computation.
In the present embodiment, the computing module 20 includes a memory 21 and a processor 22.According to calculating task
The memory 21 can be divided into multiple storage units 211 by division.Described a can also be handled according to the division of task
Device 22 is divided into multiple first calculate nodes and multiple second calculate nodes.And then it is calculated by the multiple first
Node and the multiple second calculate node calculate the calculating task.The computing module 20 can be according to different
Calculating task is dynamically configured as the first calculate node and the second calculate node, can make full use of computing resource, improves and calculates
Speed.
Fig. 3 is referred to, the application one embodiment provides a kind of computing module 20.The computing module 20 includes b meter
Calculate unit 23.
It is communicated to connect between the b computing unit 23.The communication connection mode can connect for wireless communication.
Each described computing unit 23 includes a processor 22 and a memory 21 communicated to connect with the processor 22.From
The n computing units 23 are chosen in the b computing unit 23, so that each of described n computing unit 23 calculates
Unit 23 forms first calculate node.Each calculate node is not established with other calculate nodes to be communicated.It is each
A first calculate node is used to calculate the subtask of the efficient parallel calculating.
When processor 22 does not work at least one described computing unit 23, idle 22 communication link of processor
The memory 21 connect is communicated with multiple computing units 23 in the remaining computing unit 23, to form one the
Two calculate nodes.Each described second calculate node is used to calculate the subtask of a non-effective parallel computation.
In the present embodiment, by controlling the correspondence between the computing unit 23 in the computing module 20, with
Form n first calculate nodes and p second calculate nodes.The computing module 20 can be according to different calculating
Task is dynamically configured as first calculate node and second calculate node, can make full use of computing resource, improves
Calculating speed.
Fig. 4 is referred to, the application one embodiment provides the device 100 of a kind of dynamic assigning memory and processor.It is described
The device 100 of dynamic assigning memory and processor includes: computing module 20 and control module described in any of the above-described embodiment
10。
The control module 10 is connect with 20 bus of computing module.The control module 10 is for obtaining calculating task
And the calculating task is divided into multiple subtasks.The control module 10 is also used to control the computing module 20, with shape
At multiple first calculate nodes and how much second calculate nodes.The computing module 20 and the institute in above-described embodiment
The embodiment for stating computing module 20 is identical, and details are not described herein again.
In one embodiment, the device 100 of the dynamic assigning memory and processor is mainframe computer.In the control
Computer program is stored in molding block 10.The computing module 20 is realized in above-described embodiment when executing the computer program
The step of any one the method.
The application also provides a kind of computer readable storage medium, is stored thereon with computer program.The computer journey
The step of method described in any one of above-described embodiment is realized when sequence is executed by a controller.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
The limitation to claim therefore cannot be interpreted as.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (10)
1. a kind of method of dynamic assigning memory and processor, which is characterized in that the side of the dynamic assigning memory and processor
The device of device of the method applied to dynamic assigning memory and processor, the dynamic assigning memory and processor includes control module
(10) and with the control module (10) bus the computing module (20) connecting；The computing module (20) includes that at least one is deposited
Reservoir (21) and the multiple processors (22) being connect with memory described at least one (21) bus；
The described method includes:
S10 obtains the first calculating task by the control module (10), and is divided first calculating task with shape
Subtask and multiple non-effective parallel computations at multiple subtasks, in the multiple subtask including multiple efficient parallel calculating
Subtask；
S20 obtains the mission bit stream of the multiple subtask by the control module (10)；
S30 establishes at least one institute in the computing module (20) by the control module (10) according to the mission bit stream
State the communication connection relationship between memory (21) and the processor (22)；
S40, according to the communication connection relationship, complete in the computing module (20) the memory (21) and the place
It manages device (22) to dynamically distribute, to form multiple first calculate nodes and multiple second calculate nodes.
2. the method for dynamic assigning memory according to claim 1 and processor, which is characterized in that the computing module
(20) the multiple processors (21) for including a memory (21) and being connect with the memory (21) bus；
The S30 is established at least one in the computing module (20) according to the mission bit stream by the control module (10)
The step of communication connection relationship between a memory (21) and the processor (22) includes:
The control module (10) establishes the first communication connection relationship, and the first communication connection relationship is multiple processors
Between communication connection relationship；
The S40, according to the communication connection relationship, complete in the computing module (20) the memory (21) and institute
Processor (22) dynamic allocation are stated, include: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module (20) obtains the first communication connection relationship, and will be described according to the first communication connection relationship
Multiple processors (22) are divided into multiple first processing units (221) and multiple the second processing unit (222), each described first
Processing unit (221) is used to handle the subtask of the efficient parallel calculating, each described the second processing unit (222)
For handling the subtask of a non-effective parallel computation.
3. the method for dynamic assigning memory according to claim 2 and processor, which is characterized in that the memory (21)
Including multiple storage units (211)；
The S30 is established at least one in the computing module (20) according to the mission bit stream by the control module (10)
The step of communication connection relationship between a memory (21) and the processor (22) further include:
The control module (10) establishes the second communication connection relationship, and the second communication connection relationship is that the multiple storage is single
First (211) communication connection relationship between the first processing units (221) and described the second processing unit (222) respectively；
The S40, according to the communication connection relationship, complete in the computing module (20) the memory (21) and institute
Processor (22) dynamic allocation are stated, include: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module (20) obtains the second communication connection relationship, and will be each according to the second communication connection relationship
It communicates to connect between a first processing units (221) and a storage unit (211), is calculated with forming multiple described first
Node, each described first calculate node are used to calculate the subtask of the efficient parallel calculating；
The computing module (20) also according to it is described second communication connection relationship by each described the second processing unit (222) with
Communication link between a storage unit (211) in the multiple storage unit (211) in remaining multiple storage units (211)
Connect, to form multiple second calculate nodes, each described second calculate node for calculate one it is described it is non-effective simultaneously
The subtask that row calculates.
4. the method for dynamic assigning memory according to claim 3 and processor, which is characterized in that the S20 passes through institute
After the step of stating the mission bit stream of the multiple subtask of control module (10) acquisition further include:
The control module (10) determines each of the multiple subtask according to the mission bit stream of the multiple subtask
The size of memory needed for subtask；
The size of memory needed for the computing module (20) obtains each described described subtask, and according to each described institute
The memory (21) is divided into multiple storage units (211) by the size for stating memory needed for subtask.
5. the method for dynamic assigning memory according to claim 1 and processor, which is characterized in that the computing module
(20) include multiple computing units (23), each described computing unit (23) include a processor (22) and one with it is described
The memory (21) of processor (22) communication connection；
The S30 is established at least one in the computing module (20) according to the mission bit stream by the control module (10)
The step of communication connection relationship between a memory (21) and the processor (22) includes:
The control module (10) establishes the first communication connection relationship, and the first communication connection relationship is each described calculating
Communication connection relationship between memory (21) in unit (23) and processor (22)；
The S40, according to the communication connection relationship, complete in the computing module (20) the memory (21) and institute
Processor (22) dynamic allocation are stated, include: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module (20) obtains the first communication connection relationship, and will be multiple according to the first communication connection relationship
The communication connection between memory (21) and processor (22) in the computing unit (23), and ensure at least one described meter
The processor (22) calculated in unit (23) does not work.
6. the method for dynamic assigning memory according to claim 5 and processor, which is characterized in that the S30, according to institute
State mission bit stream, by the control module (10) establish in the computing module (20) at least one described memory (21) and
The step of communication connection relationship between the processor (22) further include:
The control module (10) establishes the second communication connection relationship, and the second communication connection relationship is that the multiple storage is single
Communication connection relationship between first (211)；
The S40, according to the communication connection relationship, complete in the computing module (20) the memory (21) and institute
Processor (22) dynamic allocation are stated, include: the step of multiple first calculate nodes and multiple second calculate nodes to be formed
The computing module (20) obtains the second communication connection relationship, and will be multiple according to the second communication connection relationship
Storage unit (211) communication connection, to form multiple first calculate nodes and multiple second calculate nodes.
7. the method for dynamic assigning memory according to claim 1-6 and processor, which is characterized in that described
S40 is completed according to the communication connection relationship to the memory (21) and the processor in the computing module (20)
(22) dynamically distribute, the step of to form multiple first calculate nodes and multiple second calculate nodes after include:
Appointed by the son that the first calculate node of each of the multiple first calculate node calculates an efficient parallel calculating
Business, and pass through the son of the second calculate node of each of the multiple second calculate node one non-effective parallel computation of calculating
Task；
After the completion of first calculating task, the communication connection relationship in the computing module (20) is released.
8. a kind of computing module (20), which is characterized in that the computing module (20) is for obtaining communication connection relationship, the meter
Calculating module (20) includes:
One memory (21)；
Multiple processors (22), each described processor (21) connect with the memory (21) bus；
Wherein, the memory (21) includes multiple storage units (211), and the computing module (20) is according to the communication connection
Relationship is completed to the memory (21) and the processor (22) dynamic allocation in the computing module (20), to be formed
Multiple first calculate nodes and multiple second calculate nodes.
9. a kind of computing module (20), which is characterized in that the computing module (20) is for obtaining communication connection relationship, the meter
Calculating module (20) includes:
Multiple computing units (23) communicate to connect between the multiple computing unit (23)；
Each described computing unit (23) includes that a processor (22) and one communicate to connect with the processor (22)
Memory (21)；
Wherein, the computing module (20) is completed according to the communication connection relationship to described in the computing module (20)
Memory (21) and the processor (22) dynamically distribute, to form multiple first calculate nodes and multiple second calculate nodes.
10. the device of a kind of dynamic assigning memory and processor characterized by comprising
Computing module (20) described in any one of claim 8 or 9；And
Control module (10) is connect with the computing module (20) bus.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910276015.3A CN110083449B (en) | 2019-04-08 | 2019-04-08 | Method and device for dynamically allocating memory and processor and computing module |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910276015.3A CN110083449B (en) | 2019-04-08 | 2019-04-08 | Method and device for dynamically allocating memory and processor and computing module |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110083449A true CN110083449A (en) | 2019-08-02 |
| CN110083449B CN110083449B (en) | 2020-04-28 |
Family
ID=67414444
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910276015.3A Active CN110083449B (en) | 2019-04-08 | 2019-04-08 | Method and device for dynamically allocating memory and processor and computing module |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110083449B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1494688A (en) * | 2001-02-24 | 2004-05-05 | �Ҵ���˾ | Novel massively parallel super computer |
| US20040139297A1 (en) * | 2003-01-10 | 2004-07-15 | Huppenthal Jon M. | System and method for scalable interconnection of adaptive processor nodes for clustered computer systems |
| CN101311917A (en) * | 2007-05-24 | 2008-11-26 | 中国科学院过程工程研究所 | Particle model faced multi-tier direct-connection cluster paralleling computing system |
| CN101620588A (en) * | 2008-07-03 | 2010-01-06 | 中国人民解放军信息工程大学 | Connection and management method of reconfigurable component in high performance computer |
| CN103336756A (en) * | 2013-07-19 | 2013-10-02 | 中国人民解放军信息工程大学 | Generating device for data computational node |
| US9465549B1 (en) * | 2013-12-19 | 2016-10-11 | Emc Corporation | Dynamic allocation of a high-speed memory pool between a cluster file system and a burst buffer appliance |
| CN108304261A (en) * | 2017-12-29 | 2018-07-20 | 曙光信息产业(北京)有限公司 | A kind of job scheduling method and device based on 6D-Torus networks |
| CN108897619A (en) * | 2018-06-27 | 2018-11-27 | 国家超级计算天津中心 | A kind of multi-layer resource flexibility configuration method for supercomputer |
-
2019
- 2019-04-08 CN CN201910276015.3A patent/CN110083449B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1494688A (en) * | 2001-02-24 | 2004-05-05 | �Ҵ���˾ | Novel massively parallel super computer |
| US20040139297A1 (en) * | 2003-01-10 | 2004-07-15 | Huppenthal Jon M. | System and method for scalable interconnection of adaptive processor nodes for clustered computer systems |
| CN101311917A (en) * | 2007-05-24 | 2008-11-26 | 中国科学院过程工程研究所 | Particle model faced multi-tier direct-connection cluster paralleling computing system |
| CN101620588A (en) * | 2008-07-03 | 2010-01-06 | 中国人民解放军信息工程大学 | Connection and management method of reconfigurable component in high performance computer |
| CN103336756A (en) * | 2013-07-19 | 2013-10-02 | 中国人民解放军信息工程大学 | Generating device for data computational node |
| US9465549B1 (en) * | 2013-12-19 | 2016-10-11 | Emc Corporation | Dynamic allocation of a high-speed memory pool between a cluster file system and a burst buffer appliance |
| CN108304261A (en) * | 2017-12-29 | 2018-07-20 | 曙光信息产业(北京)有限公司 | A kind of job scheduling method and device based on 6D-Torus networks |
| CN108897619A (en) * | 2018-06-27 | 2018-11-27 | 国家超级计算天津中心 | A kind of multi-layer resource flexibility configuration method for supercomputer |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110083449B (en) | 2020-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3738045B1 (en) | Distributed partitioned map reduce using a data fabric | |
| US10567213B1 (en) | Systems and methods for selecting specific code segments in conjunction with executing requested tasks | |
| CN102567086B (en) | Task scheduling method, equipment and system | |
| US20200134762A1 (en) | Order group allocation method and device | |
| CN102629219B (en) | Reduce in parallel computation frame holds adaptive load balancing method | |
| CN107688492B (en) | Resource control method and device and cluster resource management system | |
| WO2018095065A1 (en) | Method and apparatus for allocating data object, and electronic device | |
| CN104580396A (en) | Task scheduling method, node and system | |
| US20180246765A1 (en) | System and method for scheduling jobs in distributed datacenters | |
| CN102521051B (en) | Task scheduling method, device and system in Map Reduce system applied to nomography | |
| CN106503058B (en) | A kind of data load method, terminal and computing cluster | |
| CN108268317A (en) | A kind of resource allocation methods and device | |
| CN105980988A (en) | Methods and systems for dynamically allocating resources and tasks among database work agents in smp environment | |
| CN104199912B (en) | A kind of method and device of task processing | |
| CN106294233A (en) | The transfer control method of a kind of direct memory access and device | |
| US20200052885A1 (en) | File storage method and apparatus for blockchain network | |
| WO2017201127A1 (en) | Reconfigurable distributed processing | |
| CN103914556A (en) | Large-scale graph data processing method | |
| CN107562528A (en) | Support the blocking on-demand computing method and relevant apparatus of a variety of Computational frames | |
| WO2019000786A1 (en) | Order assigning method and device | |
| WO2016074130A1 (en) | Batch processing method and device for system invocation commands | |
| CN112988369A (en) | Computer resource allocation method and system based on computing element | |
| CN110083449A (en) | The method, apparatus and computing module of dynamic assigning memory and processor | |
| CN104239520A (en) | Historical-information-based HDFS (hadoop distributed file system) data block placement strategy | |
| CN104184685A (en) | Data center resource allocation method, device and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |