CN109634731A - A kind of GPU resource packet scheduling implementation method and device based on AI cloud - Google Patents
A kind of GPU resource packet scheduling implementation method and device based on AI cloud Download PDFInfo
- Publication number
- CN109634731A CN109634731A CN201811442185.6A CN201811442185A CN109634731A CN 109634731 A CN109634731 A CN 109634731A CN 201811442185 A CN201811442185 A CN 201811442185A CN 109634731 A CN109634731 A CN 109634731A
- Authority
- CN
- China
- Prior art keywords
- gpu
- host
- type
- grouping
- business
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
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
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
-
- 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
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
-
- 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
- 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]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
-
- 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
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
- G06F2009/45562—Creating, deleting, cloning virtual machine instances
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/48—Indexing scheme relating to G06F9/48
- G06F2209/484—Precedence
-
- 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
- G06F2209/5021—Priority
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The embodiment of the invention discloses a kind of GPU resource packet scheduling implementation methods and device based on AI cloud, this method comprises: creating multiple groupings according to GPU type；According to the resource distribution situation of each host by each host assignment into corresponding grouping；The type of business being able to carry out to the GPU type of each grouping is classified, when obtaining a pending business, required GPU type is determined according to the type of business of the business, and according to the required GPU type host that acquisition GPU is in idle condition from corresponding grouping.By the example scheme, more convenient management resource is realized, whole resource utilization is improved, so that the task of high-performance GPU card is needed preferentially to obtain high-performance GPU resource, task is distributed more evenly, had both been improved GPU resource utilization rate, and had been also improved the training effectiveness of algorithm engineering teacher.
Description
Technical field
The present embodiments relate to AI cloud, espespecially a kind of GPU resource packet scheduling implementation method based on AI cloud and
Device.
Background technique
In the artificial intelligence AI epoch, algorithm engineering teacher needs to carry out a large amount of deep learning tasks, holds usually using docker
Device can significantly improve training speed as training environment, using expensive GPU card, a large amount of AI servers for being furnished with GPU card
(docker container host) passes through k8s (Kubernetes abbreviation, container cluster management system of increasing income) Lai Jinhang clustering money
Source control, then AI cloud platform can be managed collectively docker container cluster by k8s again.When algorithm engineering, teacher needs using GPU
When resource, it is necessary to which how the distribution of GPU resource makes the needing high-performance GPU card of the task preferentially to obtain GPU card to be exactly one
It is a need to face solve the problems, such as.
Summary of the invention
The embodiment of the invention provides a kind of GPU resource packet scheduling implementation method and device based on AI cloud, Neng Gougeng
Add convenient management resource, whole resource utilization is improved, so that the task of high-performance GPU card is needed preferentially to obtain high-performance
GPU resource, task distribute more evenly, had both improved GPU resource utilization rate, and also improved the training effectiveness of algorithm engineering teacher.
In order to reach purpose of the embodiment of the present invention, the embodiment of the invention provides a kind of figures based on artificial intelligence AI cloud
Processor GPU resource packet scheduling implementation method, the method may include:
Multiple groupings are created according to GPU type；
According to the resource distribution situation of each host by each host assignment into corresponding grouping；
The type of business being able to carry out to the GPU type of each grouping is classified, to obtain a pending industry
When business, required GPU type is determined according to the type of business of the business, and according to the required GPU type from corresponding point
The host that GPU is in idle condition is obtained in group.
In an exemplary embodiment of the present invention, the resource distribution situation may include: configured GPU type；Institute
Stating GPU type includes: P100, V100 or without GPU.
In an exemplary embodiment of the present invention, the resource distribution situation according to each host is by each host assignment
It may include: the GPU type configured according to each host into corresponding grouping, by calling the container cluster management system that increases income
The management interface of system k8s, stamps corresponding label to each host.
In an exemplary embodiment of the present invention, described that GPU is obtained from corresponding grouping according to the required GPU type
The host being in idle condition may include: that the k8s is determined and the required GPU class according to the label on each host
Whole hosts in the corresponding grouping of type, and the host that acquisition GPU is in idle condition from whole hosts.
In an exemplary embodiment of the present invention, the method can also include: when obtaining a pending business,
The corresponding container of the good pending business is configured, and after obtaining the host that the GPU is in idle condition, it will be described
Container is put into the host that the GPU is in idle condition.
In an exemplary embodiment of the present invention, the method can also include: to be performed in the pending business
Afterwards, container described in auto-destruct.
The embodiment of the invention also provides a kind of, and the graphics processor GPU resource packet scheduling based on artificial intelligence AI cloud is real
Existing device, the apparatus may include: grouping module, distribution module, categorization module and acquisition module；
The grouping module, for creating multiple groupings according to GPU type；
The distribution module, for being grouped each host assignment to corresponding according to the resource distribution situation of each host
In；
The categorization module, the type of business for being able to carry out to the GPU type of each grouping are classified, in institute
When stating acquisition module one pending business of acquisition, required GPU type, and root are determined according to the type of business of the business
The host that GPU is in idle condition is obtained from corresponding grouping according to the required GPU type.
In an exemplary embodiment of the present invention, the resource distribution situation may include: configured GPU type；Institute
Stating GPU type may include: P100, V100 or without GPU.
In an exemplary embodiment of the present invention, the distribution module will be each according to the resource distribution situation of each host
Host assignment may include: the GPU type configured according to each host into corresponding grouping, by calling sets of containers of increasing income
The management interface of cluster management system k8s stamps corresponding label to each host.
In an exemplary embodiment of the present invention, the acquisition module is grouped according to the required GPU type from corresponding
It is middle obtain the host that is in idle condition may include: by the k8s according to the label on each host determine with it is described
Whole hosts in the corresponding grouping of required GPU type, and obtain what GPU was in idle condition from whole hosts
Host.
In an exemplary embodiment of the present invention, described device can also include: configuration module；
The configuration module, it is good described wait hold for configuring when the acquisition module obtains a pending business
The corresponding container of capable business, and after the acquisition module obtains the host that the GPU is in idle condition, by the appearance
Device is put into the host that the GPU is in idle condition.
In an exemplary embodiment of the present invention, the configuration module can be also used for: in the pending business quilt
After execution, container described in auto-destruct.
The embodiment of the present invention includes: multiple groupings are created according to GPU type；It will according to the resource distribution situation of each host
Each host assignment is into corresponding grouping；The type of business being able to carry out to the GPU type of each grouping is classified, with
When obtaining a pending business, required GPU type is determined according to the type of business of the business, and required according to this
GPU type obtains the host that GPU is in idle condition from corresponding grouping.By the example scheme, realize more square
Just management resource, improves whole resource utilization, so that the task of high-performance GPU card is needed preferentially to obtain high-performance GPU
Resource, task distribute more evenly, had both improved GPU resource utilization rate, and also improved the training effectiveness of algorithm engineering teacher.
The other feature and advantage of the embodiment of the present invention will illustrate in the following description, also, partly from explanation
It is become apparent in book, or understand through the implementation of the invention.The purpose of the embodiment of the present invention and other advantages can pass through
Specifically noted structure is achieved and obtained in the specification, claims and drawings.
Detailed description of the invention
Attached drawing is used to provide one for further understanding technical solution of the embodiment of the present invention, and constituting specification
Point, it is used to explain the present invention the technical solution of embodiment together with embodiments herein, does not constitute to the embodiment of the present invention
The limitation of technical solution.
Fig. 1 is the GPU resource packet scheduling implementation method flow chart based on AI cloud of the embodiment of the present invention；
Fig. 2 is the GPU resource packet scheduling realization device composition block diagram based on AI cloud of the embodiment of the present invention.
Specific embodiment
Understand in order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing pair
The embodiment of the present invention is described in detail.It should be noted that in the absence of conflict, embodiment and reality in the application
The feature applied in example can mutual any combination.
Step shown in the flowchart of the accompanying drawings can be in a computer system such as a set of computer executable instructions
It executes.Also, although logical order is shown in flow charts, and it in some cases, can be to be different from herein suitable
Sequence executes shown or described step.
In order to reach purpose of the embodiment of the present invention, the embodiment of the invention provides a kind of figures based on artificial intelligence AI cloud
Processor GPU resource packet scheduling implementation method, as shown in Figure 1, the method may include S101-S102:
S101, multiple groupings are created according to GPU type.
In an exemplary embodiment of the present invention, the GPU type may include: P100, V100 or without GPU.
In an exemplary embodiment of the present invention, it can be created by preset grouping creating device and be marked with GPU first
Sign multiple groupings of (can be according to model such as P100, V100, nothing).Being grouped creating device creation (can be according to model with GPU label
Such as P100, V100, nothing) grouping when, rationally grouping can be created according to real resource situation, if nothing may be selected without GPU.
S102, according to the resource distribution situation of each host by each host assignment into corresponding grouping.
In an exemplary embodiment of the present invention, the resource distribution situation may include: configured GPU type；I.e.
Whether it is configured with GPU in the host, it, can be by the grouping of the host assignment to no GPU, if configuration if do not configured
There is GPU, then may further determine that GPU concrete type that the host is configured is, such as P100, V100, it can should
Host estimates that the specific GPU type is assigned in corresponding grouping.
In an exemplary embodiment of the present invention, each host can be assigned in each grouping by host packet device
(while labelling for host, in order to which k8s is scheduled by host label).
In an exemplary embodiment of the present invention, the resource distribution situation according to each host is by each host assignment
It may include: the GPU type configured according to each host into corresponding grouping, by calling the container cluster management system that increases income
The management interface of system k8s, stamps corresponding label to each host.
In an exemplary embodiment of the present invention, host packet device assigns to host in each grouping, the process of grouping
The process namely to label to host is labelled by calling k8s management interface to host, and k8s can pass through host mark
Label are to be scheduled.
S103, the type of business being able to carry out to the GPU type of each grouping are classified, with pending in acquisition one
Business when, determine required GPU type according to the type of business of the business, and according to the required GPU type from corresponding
Grouping in obtain the host that is in idle condition of GPU.
In an exemplary embodiment of the present invention, grouping resources can be applied to by certain business by grouping application apparatus
On, if P100, V100 grouping can be used for deep learning business, the grouping of GPU not can be used for not needing the business of GPU, for example,
There is no the grouping of GPU to can be used for creating visualization business (visualization business does not need GPU resource).
In an exemplary embodiment of the present invention, described that GPU is obtained from corresponding grouping according to the required GPU type
The host being in idle condition may include: that the k8s is determined and the required GPU class according to the label on each host
Whole hosts in the corresponding grouping of type, and the host that acquisition GPU is in idle condition from whole hosts.
In an exemplary embodiment of the present invention, the method can also include: when obtaining a pending business,
The corresponding container of the good pending business is configured, and after obtaining the host that the GPU is in idle condition, it will be described
Container is put into the host that the GPU is in idle condition.
In an exemplary embodiment of the present invention, host resource is reasonably grouped by host packet device, is provided
Source dispatching device according to business it needs to be determined that the particular requirement to host suitable host is selected as screening conditions in turn
Create container.
In an exemplary embodiment of the present invention, when user submits the deep learning of a GPU for needing P100 model to appoint
When business, resource management apparatus has configured containers size, and specifies and container is placed on the host with P100 label；K8s is logical
It crosses dispatching device and filters out the whole hosts for being equipped with P100 label, then choose idle GPU points further according to load on host computers situation
It is used to user.
In an exemplary embodiment of the present invention, the method can also include: to be performed in the pending business
Afterwards, container described in auto-destruct, to save host resource.
In an exemplary embodiment of the present invention, by designing reasonable resource grouping and application strategy, according to GPU type
(or model) is grouped resource, so that can not to occupy the host resource with GPU (CPU, interior for the business that does not need GPU
Deposit, GPU), prevent the host equipped with GPU because CPU, memory can not create the container with GPU due to the deficiencies of, and make GPU not busy
Set waste of resource；To improve GPU utilization rate.
The embodiment of the invention also provides a kind of, and the graphics processor GPU resource packet scheduling based on artificial intelligence AI cloud is real
Existing device 1, it should be noted that any embodiment in above-mentioned embodiment of the method can be applied in the Installation practice,
This is no longer going to repeat them.As shown in Fig. 2, the apparatus may include: grouping module 11, distribution module 12, categorization module 13
With acquisition module 14；
The grouping module 11, for creating multiple groupings according to GPU type；
The distribution module 12, for the resource distribution situation according to each host by each host assignment to corresponding point
In group；
The categorization module 13, the type of business for being able to carry out to the GPU type of each grouping are classified, with
When the acquisition module 14 obtains a pending business, required GPU type is determined according to the type of business of the business,
And the host that GPU is in idle condition is obtained from corresponding grouping according to the required GPU type.
In an exemplary embodiment of the present invention, the resource distribution situation may include: configured GPU type；Institute
Stating GPU type may include: P100, V100 or without GPU.
In an exemplary embodiment of the present invention, the distribution module 12 will be every according to the resource distribution situation of each host
A host assignment may include: the GPU type configured according to each host into corresponding grouping, by calling container of increasing income
The management interface of cluster management system k8s stamps corresponding label to each host.
In an exemplary embodiment of the present invention, the acquisition module 14 is according to the required GPU type from corresponding point
Obtained in group the host that is in idle condition of GPU may include: by the k8s according to the label on each host determine with
Whole hosts in the corresponding grouping of the required GPU type, and obtain GPU from whole hosts and be in idle shape
The host of state.
In an exemplary embodiment of the present invention, described device can also include: configuration module 15；
The configuration module 15, for when the acquisition module obtains a pending business, configure it is good described to
The corresponding container of the business of execution, and after the acquisition module obtains the host that the GPU is in idle condition, it will be described
Container is put into the host that the GPU is in idle condition.
In an exemplary embodiment of the present invention, the configuration module 15 can be also used for: in the pending business
After being performed, container described in auto-destruct.
The embodiment of the present invention includes: multiple groupings are created according to GPU type；It will according to the resource distribution situation of each host
Each host assignment is into corresponding grouping；The type of business being able to carry out to the GPU type of each grouping is classified, with
When obtaining a pending business, required GPU type is determined according to the type of business of the business, and required according to this
GPU type obtains the host that GPU is in idle condition from corresponding grouping.By the example scheme, realize more square
Just management resource, improves whole resource utilization, so that the task of high-performance GPU card is needed preferentially to obtain high-performance GPU
Resource, task distribute more evenly, had both improved GPU resource utilization rate, and also improved the training effectiveness of algorithm engineering teacher.
It will appreciated by the skilled person that whole or certain steps, system, dress in method disclosed hereinabove
Functional module/unit in setting may be implemented as software, firmware, hardware and its combination appropriate.In hardware embodiment,
Division between the functional module/unit referred in the above description not necessarily corresponds to the division of physical assemblies；For example, one
Physical assemblies can have multiple functions or a function or step and can be executed by several physical assemblies cooperations.Certain groups
Part or all components may be implemented as by processor, such as the software that digital signal processor or microprocessor execute, or by
It is embodied as hardware, or is implemented as integrated circuit, such as specific integrated circuit.Such software can be distributed in computer-readable
On medium, computer-readable medium may include computer storage medium (or non-transitory medium) and communication media (or temporarily
Property medium).As known to a person of ordinary skill in the art, term computer storage medium is included in for storing information (such as
Computer readable instructions, data structure, program module or other data) any method or technique in the volatibility implemented and non-
Volatibility, removable and nonremovable medium.Computer storage medium include but is not limited to RAM, ROM, EEPROM, flash memory or its
His memory technology, CD-ROM, digital versatile disc (DVD) or other optical disc storages, magnetic holder, tape, disk storage or other
Magnetic memory apparatus or any other medium that can be used for storing desired information and can be accessed by a computer.This
Outside, known to a person of ordinary skill in the art to be, communication media generally comprises computer readable instructions, data structure, program mould
Other data in the modulated data signal of block or such as carrier wave or other transmission mechanisms etc, and may include any information
Delivery media.
Claims (10)
1. a kind of graphics processor GPU resource packet scheduling implementation method based on artificial intelligence AI cloud, which is characterized in that described
Method includes:
Multiple groupings are created according to GPU type；
According to the resource distribution situation of each host by each host assignment into corresponding grouping；
The type of business being able to carry out to the GPU type of each grouping is classified, with when obtaining a pending business,
Required GPU type is determined according to the type of business of the business, and is obtained from corresponding grouping according to the required GPU type
Take the host that GPU is in idle condition.
2. the GPU resource packet scheduling implementation method according to claim 1 based on AI cloud, which is characterized in that the money
Source configuring condition includes: configured GPU type；The GPU type includes: P100, V100 or without GPU.
3. the GPU resource packet scheduling implementation method according to claim 1 or 2 based on AI cloud, which is characterized in that described
It by each host assignment include: to be configured according to each host into corresponding grouping according to the resource distribution situation of each host
GPU type, by call open source container cluster management system k8s management interface, stamp corresponding label to each host.
4. the GPU resource grouping scheduling method according to claim 3 based on AI cloud, which is characterized in that the basis should
It includes: the k8s according to each host that required GPU type, which obtains the host that GPU is in idle condition from corresponding grouping,
On label determine whole hosts in grouping corresponding with the required GPU type, and from whole hosts
Obtain the host that GPU is in idle condition.
5. the GPU resource grouping scheduling method according to claim 1 based on AI cloud, which is characterized in that the method is also
It include: the corresponding container of the good pending business of configuration when obtaining a pending business, and described in the acquisition
After the host that GPU is in idle condition, the container is put into the host that the GPU is in idle condition.
6. the GPU resource grouping scheduling method according to claim 5 based on AI cloud, which is characterized in that the method is also
It include: the container described in auto-destruct after the pending business is performed.
7. a kind of graphics processor GPU resource packet scheduling realization device based on artificial intelligence AI cloud, which is characterized in that described
Device includes: grouping module, distribution module, categorization module and obtains module；
The grouping module, for creating multiple groupings according to GPU type；
The distribution module, for according to the resource distribution situation of each host by each host assignment into corresponding grouping；
The categorization module, the type of business for being able to carry out to the GPU type of each grouping are classified, to obtain described
When modulus block obtains a pending business, required GPU type is determined according to the type of business of the business, and according to this
Required GPU type obtains the host that GPU is in idle condition from corresponding grouping.
8. the GPU resource packet scheduling realization device according to claim 7 based on AI cloud, which is characterized in that the money
Source configuring condition includes: configured GPU type；The GPU type includes: P100, V100 or without GPU.
9. the GPU resource packet scheduling realization device according to claim 7 or 8 based on AI cloud, which is characterized in that described
Each host assignment is included: according to each master into corresponding grouping according to the resource distribution situation of each host by distribution module
The GPU type that machine is configured stamps phase to each host by calling the management interface of open source container cluster management system k8s
The label answered.
10. the GPU resource packet scheduling apparatus according to claim 9 based on AI cloud, which is characterized in that the acquisition mould
It includes: by the k8s that root tuber, which obtains the host that GPU is in idle condition from corresponding grouping according to the required GPU type,
Whole hosts in grouping corresponding with the required GPU type are determined according to the label on each host, and from institute
It states and obtains the host that GPU is in idle condition in whole hosts.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811442185.6A CN109634731A (en) | 2018-11-29 | 2018-11-29 | A kind of GPU resource packet scheduling implementation method and device based on AI cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811442185.6A CN109634731A (en) | 2018-11-29 | 2018-11-29 | A kind of GPU resource packet scheduling implementation method and device based on AI cloud |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109634731A true CN109634731A (en) | 2019-04-16 |
Family
ID=66069818
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811442185.6A Pending CN109634731A (en) | 2018-11-29 | 2018-11-29 | A kind of GPU resource packet scheduling implementation method and device based on AI cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109634731A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111930525A (en) * | 2020-10-10 | 2020-11-13 | 北京世纪好未来教育科技有限公司 | GPU resource use method, electronic device and computer readable medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107135257A (en) * | 2017-04-28 | 2017-09-05 | 东方网力科技股份有限公司 | Task is distributed in a kind of node cluster method, node and system |
-
2018
- 2018-11-29 CN CN201811442185.6A patent/CN109634731A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107135257A (en) * | 2017-04-28 | 2017-09-05 | 东方网力科技股份有限公司 | Task is distributed in a kind of node cluster method, node and system |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111930525A (en) * | 2020-10-10 | 2020-11-13 | 北京世纪好未来教育科技有限公司 | GPU resource use method, electronic device and computer readable medium |
| CN111930525B (en) * | 2020-10-10 | 2021-02-02 | 北京世纪好未来教育科技有限公司 | GPU resource use method, electronic device and computer readable medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106656867B (en) | A kind of dynamic SDN configuration method of the application perception based on virtual network | |
| CN109684065B (en) | Resource scheduling method, device and system | |
| CN104965757B (en) | Method, virtual machine (vm) migration managing device and the system of live migration of virtual machine | |
| CN105512083B (en) | Method for managing resource, apparatus and system based on YARN | |
| WO2019055871A1 (en) | Systems and methods for a policy-driven orchestration of deployment of distributed applications | |
| CN110413412A (en) | A kind of method and apparatus based on GPU cluster resource allocation | |
| US20240111586A1 (en) | Multi-policy intelligent scheduling method and apparatus oriented to heterogeneous computing power | |
| CN105589731B (en) | A kind of virtual machine migration method and device | |
| CN109471725A (en) | Resource allocation methods, device and server | |
| CN109240825A (en) | Elastic method for scheduling task, device, equipment and computer readable storage medium | |
| CN110389843A (en) | A kind of business scheduling method, device, equipment and readable storage medium storing program for executing | |
| CN109445923A (en) | A kind of method and device of micro services task schedule | |
| CN113674131A (en) | Hardware accelerator equipment management method and device, electronic equipment and storage medium | |
| CN115048216A (en) | Resource management scheduling method, device and equipment for artificial intelligence cluster | |
| CN114374609A (en) | Deep learning operation running method and system based on RDMA (remote direct memory Access) equipment | |
| CN114943885A (en) | Synchronous cache acceleration method and system based on training task | |
| CN109634731A (en) | A kind of GPU resource packet scheduling implementation method and device based on AI cloud | |
| CN113835897A (en) | Method for allocating and using GPU resources on distributed computing cluster Kubernets | |
| CN108804202A (en) | A kind of enterprise-level container mirror image management method and system | |
| CN108170508A (en) | Batch creates method, apparatus, equipment and its storage medium of virtual machine | |
| CN111400021B (en) | Deep learning method, device and system | |
| CN109788061B (en) | Computing task deployment method and device | |
| CN109660575B (en) | Method and device for realizing NFV service deployment | |
| CN107025126A (en) | A kind of resource regulating method, NFVO and system | |
| WO2017125161A1 (en) | Resource allocation |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190416 |
|
| RJ01 | Rejection of invention patent application after publication |