WO2021208915A1 - 算力共享方法及相关设备 - Google Patents

算力共享方法及相关设备 Download PDF

Info

Publication number
WO2021208915A1
WO2021208915A1 PCT/CN2021/086931 CN2021086931W WO2021208915A1 WO 2021208915 A1 WO2021208915 A1 WO 2021208915A1 CN 2021086931 W CN2021086931 W CN 2021086931W WO 2021208915 A1 WO2021208915 A1 WO 2021208915A1
Authority
WO
WIPO (PCT)
Prior art keywords
computing
computing power
terminal
demand
information
Prior art date
Application number
PCT/CN2021/086931
Other languages
English (en)
French (fr)
Inventor
刘莹莹
苗润泉
朱春晖
孙明
Original Assignee
展讯半导体(南京)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 展讯半导体(南京)有限公司 filed Critical 展讯半导体(南京)有限公司
Priority to JP2022562579A priority Critical patent/JP2023521222A/ja
Priority to KR1020227036077A priority patent/KR20230035516A/ko
Priority to US17/918,464 priority patent/US11876856B2/en
Publication of WO2021208915A1 publication Critical patent/WO2021208915A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • 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]
    • G06F9/5005—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
    • G06F9/505—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 considering the load
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
    • H04L65/40—Support for services or applications
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/50—Network services
    • H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • This application relates to the field of computing technology, and in particular to a computing power sharing method and related equipment.
  • the embodiments of the present application provide a computing power sharing method and related equipment, which can realize computing power sharing.
  • an embodiment of the present application provides a computing power sharing method, which is applied to a computing power sharing network element, and the computing power sharing network element is connected to the network for control plane transmission and data plane transmission, or, The computing power sharing network element is connected to the network for control plane transmission;
  • the method includes:
  • the computing power demand information and the available computing power information determine the computing collaboration terminal that provides computing power for the computing demand terminal, and directly or indirectly instruct the computing computing collaboration terminal that provides computing power to complete the computing demand End computing tasks.
  • the obtaining the computing power demand information sent by the computing demand terminal and the available computing power information sent by the computing cooperation terminal includes:
  • the computing power request includes the ID of the computing demand terminal and computing power demand information, and when the computing demand terminal expects to participate in computing power sharing, the computing power request also includes all The computing power information that can be participated in the computing demand side, and when the computing demand side expects to perform task allocation and information integration on the computing task during computing power sharing, the computing power request also includes the integration capability information with the computing demand side ;
  • the collaboration query includes a request to provide a computing power report;
  • the computing requirement is determined according to the integration capability information
  • the collaboration inquiry also includes an integration capability report request;
  • collaboration query response information returned by the computing collaboration terminal in response to the collaboration query
  • the collaboration query response information includes the computing collaboration terminal ID and computing power information returned in response to the computing power reporting request
  • the integration capability information of the computing collaboration terminal returned in response to the integration capability reporting request.
  • the obtaining the computing power demand information sent by the computing demand terminal and the available computing power information sent by the computing cooperation terminal includes:
  • the computing power providing request including the ID of the computing collaboration terminal and information that can provide computing power, and the computing collaboration terminal expects to perform the computing task on the computing power sharing
  • the computing power provision request also includes integration capability information
  • the demand query includes a computing power demand reporting request.
  • the demand query also includes a computing power reporting request, and the computing demand terminal is expected to
  • the demand query also includes an integration capability reporting request;
  • Receive demand query response information returned by the computing demand terminal in response to the demand query where the demand query response information includes the computing demand terminal ID and computing power demand information returned in response to the computing power demand reporting request, and respond to all The participating computing power information of the computing demand end returned when the computing power reporting request is requested, and the integration capability information of the computing demanding terminal returned in response to the integration capability reporting request.
  • the determining a computing cooperation terminal that provides computing power for the computing demand terminal according to the computing power demand information and the available computing power information includes:
  • the participating computing power information of the computing power sharing network element, the participating computing power information of the computing demand end, and the computing power available information of the computing collaboration terminal it is determined whether it is the computing demand. Provide computing power at the end;
  • the available computing power information and the available computing power information can meet the computing power demand determined by the computing power demand information, provide computing power for the computing demand side and determine to provide computing power for the computing demand side A first terminal of computing power, where the first terminal includes at least one of the computing power sharing network element, the computing demand terminal, and the computing collaboration terminal.
  • the directly or indirectly instructing the computing collaboration end that provides computing power to complete the computing task of the computing demand end includes:
  • the computing power sharing network element When the computing power sharing network element is connected to the network for control plane transmission and data plane transmission, according to the integration capability information of the computing demand side and the integration capability information of the computing cooperation terminal, the computing power sharing network element, The computing demand terminal and the computing collaboration terminal determine the second terminal for task allocation and information integration; or, when the computing power sharing network element is connected to the network for control plane transmission, according to the integration capability of the computing demand terminal Information and integration capability information of the computing collaboration terminal, and determining a second terminal for task allocation and information integration from the computing demand terminal and the computing collaboration terminal;
  • the method further includes:
  • the computing power distribution list is the ID of the first terminal and the assigned computing power.
  • the method further includes:
  • the network includes a core network
  • the computing power sharing network element interacts with the unified data management network element of the core network through a first reference point, and the computing power sharing network element interacts with the access and mobility management network of the core network through a second reference point Element interaction, where the computing power sharing network element interacts with the user plane network element of the core network through a third reference point;
  • the computing power sharing network element interacts with the unified data management network element of the core network through a first reference point, and the computing power sharing network element interacts with the access and mobility management network of the core network through a second reference point Meta interaction.
  • the computing power demand information includes the amount of computing required by the computing task.
  • the computing power requirement information further includes at least one of the maximum acceptable computing time for completing the computing task, the maximum task parallelism, the algorithm complexity of the computing task, and the specified collection data type.
  • the embodiments of the present application provide a computing power sharing network element that is connected to the network for control plane transmission and data plane transmission, or the computing power sharing network element is connected to the network For control plane transmission;
  • the computing power sharing network element includes:
  • the obtaining module is used to obtain the computing power demand information sent by the computing demand terminal and the available computing power information sent by the computing cooperation terminal;
  • the processing module is used to determine the computing collaboration terminal that provides computing power for the computing demand terminal based on the computing power demand information and the computing power information that can be provided, and directly or indirectly instruct the computing collaboration terminal that provides computing power Completing the computing task of the computing demand side.
  • the network includes a core network
  • the computing power sharing network element interacts with the unified data management network element of the core network through a first reference point, and the computing power sharing network element interacts with the access and mobility management network of the core network through a second reference point Element interaction, where the computing power sharing network element interacts with the user plane network element of the core network through a third reference point;
  • the computing power sharing network element interacts with the unified data management network element of the core network through a first reference point, and the computing power sharing network element interacts with the access and mobility management network of the core network through a second reference point Meta interaction.
  • an embodiment of the present application provides a computing power sharing network element, including: a transceiver, a processor, a memory, and a computer program stored on the memory and executed by the processor;
  • the processor is respectively connected to the transceiver and the memory, and is configured to implement the computing power sharing method described in the first aspect by executing the computer program.
  • an embodiment of the present application provides a computing power sharing system, including a computing demand terminal, a computing collaboration terminal, and the computing power sharing network element described in the second or third aspect.
  • an embodiment of the present application provides a computer storage medium that stores a computer program, and the computer program includes program instructions. When executed by a processor, the program instructions execute as in the first aspect.
  • the described computing power sharing method When executed by a processor, the program instructions execute as in the first aspect.
  • the computing power sharing network element is connected to the network for control plane transmission and data plane transmission, or the computing power sharing network element is connected to the network for control plane transmission; wherein the computing power sharing network element obtains computing requirements The computing power demand information sent by the terminal and the available computing power information sent by the computing collaboration terminal; then based on the computing power demand information and the available computing power information, determine the computing collaboration terminal that provides computing power for the computing demand side, and direct or indirect instructions
  • the computing collaboration end that provides computing power completes the computing tasks of the computing demand side, realizes the sharing of computing power, improves the calculation speed, and reduces the time required for calculation.
  • Figure 1a is a schematic diagram of a connection between a computing power sharing network element and a core network provided by an embodiment of the present application;
  • Figures 1b and 1c are schematic diagrams of a reference point connection between a computing power sharing network element and a core network provided by an embodiment of the present application;
  • FIG. 2 is a schematic flowchart of a computing power sharing method provided by an embodiment of the present application
  • FIG. 3 is a schematic diagram of an application authorization process of a computing power sharing method provided by an embodiment of the present application
  • FIG. 4 is a schematic diagram of an application authorization process of a computing power sharing method provided by an embodiment of the present application
  • FIG. 5 is a schematic diagram of a scheduling allocation process of a computing power sharing method provided by an embodiment of the present application.
  • FIG. 6 is a schematic diagram of a scheduling allocation flow of a computing power sharing method provided by an embodiment of the present application.
  • FIG. 7 is a schematic diagram of a scheduling allocation process of a computing power sharing method provided by an embodiment of the present application.
  • FIG. 8 is a schematic structural diagram of a computing power sharing network element provided by an embodiment of the present application.
  • FIG. 9 is a schematic structural diagram of a computing power sharing network element provided by an embodiment of the present application.
  • FIG. 10 is a schematic structural diagram of a computing power sharing system provided by an embodiment of the present application.
  • this application proposes a computing power sharing method that can realize computing power sharing so as to quickly complete computing tasks.
  • the computing demand end can be an agent, a user equipment (UE), a computing server, a base station, etc., any carrier that needs or can provide computing power services
  • the computing collaboration end can be an agent, and the user Equipment (User Equipment, UE), computing server, base station, etc., any carrier that requires or can provide computing power services.
  • Both the computing demand terminal and the computing cooperation terminal have a unique identification ID, and the ID may be a sequence consisting of at least one of numbers, letters, and special characters.
  • a 3GPP the 3rd Generation partnership project
  • This application provides a computing power sharing method, which is applied to computing power sharing network elements.
  • the computing power sharing network element is connected to the network.
  • the network includes a core network.
  • the computing power sharing network element is a new core network for computing power sharing.
  • the network element is an entity deployed by the operator.
  • the computing power sharing network element is connected to the core network.
  • Figure 1a is a schematic diagram of the connection between the computing power sharing network element and the core network provided by the embodiment of this application.
  • the network element passes through the newly added fourth interface.
  • the Nai interface is connected to the interfaces of other network elements in the core network to realize interaction with other network elements in the core network.
  • an embodiment of the present application provides a schematic diagram of the connection of the reference point of the computing power sharing network element and the core network; in Figure 1b, the computing power sharing network element is connected to the core network
  • the computing power sharing network element interacts with the Unified Data Management (UDM) network element through the newly added first reference point N60, and the computing power sharing network element interacts with the unified data management (UDM) network element through the new
  • the added second reference point N59 interacts with Access and Mobility Management Function (AMF) network elements to achieve control plane transmission, and the computing power sharing network element has scheduling capabilities.
  • UDM Unified Data Management
  • AMF Access and Mobility Management Function
  • the computing power sharing network element interacts with the User Plane Function (UPF) network element through the newly added third reference point N61, so as to realize data plane transmission. That is, the computing power sharing network element can realize large-scale communication with other terminal devices. Data transmission, such as transmission of training models or training sets, etc.
  • the computing power sharing network element has the ability to integrate information.
  • the computing power sharing network element is connected to the core network for control plane transmission; specifically, the computing power sharing network element is connected to unified data management through the newly added first reference point N60 ( Unified Data Management (UDM) network elements interact, and computing power sharing network elements interact with Access and Mobility Management Function (AMF) network elements through the newly added second reference point N59 to achieve control Plane transmission, computing power sharing network elements have scheduling capabilities.
  • UDM Unified Data Management
  • AMF Access and Mobility Management Function
  • FIG. 2 is a schematic flowchart of a method for sharing computing power provided by an embodiment of the present application; the method for sharing computing power includes:
  • the computing power sharing network element obtains the computing power demand information of the computing demand side, and the computing power information that can be provided by the computing collaboration terminal.
  • the computing power demand information includes the amount of calculation required for its computing task, and can also include the acceptable completion of the computing task. At least one of the longest calculation time of the calculation time, the maximum task parallelism N (specify that the calculation task can be allocated to N terminals for processing at most), the algorithm complexity of the calculation task, and the specified collection data type. Specifying the type of collected data is used to specify the type of data that needs to be collected by the computing partner during the calculation process.
  • the computing power information that can be provided includes the amount of calculation that can be provided by the computing collaboration terminal.
  • the computing power information that can be provided can also include the time required to provide the above calculation amount, the maximum acceptable task parallelism, and the complexity of the algorithm that can be processed. At least one of.
  • the computing power demand information and the computing power information that can be provided determine the computing collaboration terminal that provides computing power for the computing demand side, and directly or indirectly instruct the computing collaboration terminal that provides computing power to complete the computing task of the computing demand side.
  • the number of computing collaboration terminals that provide computing power for computing demand terminals may be more than one.
  • the computing power sharing network element can directly instruct the computing collaboration end that provides computing power to complete the computing tasks of the computing demand side, or indirectly instruct the computing collaboration end that provides computing power to complete the computing tasks of the computing demand side, so as to realize the computing power of the computing collaboration side. shared.
  • the computing power sharing network element may instruct the computing demand side or the computing collaboration terminal to control the computing collaboration terminal that provides computing power to complete the computing tasks of the computing demand side.
  • the network is no longer just a communication channel, but has a scheduling function.
  • the computing power sharing network element can coordinate the scheduling of the computing demand side and the computing cooperation side to achieve the purpose of sharing computing power, increase computing speed, and reduce computing
  • the calculation time required by the task improves the calculation efficiency, and the calculation power sharing method based on the calculation power sharing network element has a higher degree of trust.
  • the computing power sharing network element directly instructing the computing collaboration terminal that provides computing power to complete the computing task of the computing demand side as an example.
  • the computing power sharing network element assigns the computing task to the computing collaboration terminal.
  • the sub-processing results of the computing task can be obtained, and the sub-processing results of the computing collaboration end can be integrated by the computing power sharing network element with information integration capability to obtain the total processing result of the computing task.
  • the calculation The power sharing network element sends the new task to be processed to the computing collaboration end, and the computing collaboration end re-processes the task until the total processing result meets the requirements, then the computing power sharing network element returns the total processing result to the computing demand end.
  • the computing demand side or the computing collaboration side has the information integration capability, they can also replace the computing power sharing network element for task allocation and information integration to obtain the total processing result of the computing task.
  • the computing demand side or the computing power sharing network element has computing capabilities, they are equivalent to the computing collaboration terminal and can also participate in computing power sharing.
  • step 201 includes:
  • the computing power request includes the ID of the computing demand side and computing power demand information.
  • the computing power request also includes the available computing power information of the computing demand side.
  • the computing power request also includes the integration capability information of the computing demand side;
  • the computing power sharing network element receives the computing power request sent by the computing demand terminal, where the computing power request must include the unique identification ID of the computing demand terminal and computing power demand information.
  • the computing power request must include the unique identification ID of the computing demand terminal and computing power demand information.
  • participating computing power information and integration capability information whether it is included in the computing power request is optional.
  • the available computing power information is the computing power information used by the computing demand side for computing power sharing. It can include the amount of calculation that the computing demand side can provide, as well as the time required to provide the above calculation amount and acceptable task parallelism. , And at least one of the algorithm complexity that can be processed.
  • the computing demand side When the computing demand side does not have redundant computing power, it can directly not carry the participating computing power information, or the participating computing power information can be replaced with "0". In the same way, when the computing demand side has information integration capabilities, it can choose to carry the integration capability information in the computing power request to indicate that the computing demand side expects to perform task allocation and information integration for computing tasks when computing power is shared, or choose not to carry it .
  • the ability to integrate information can use "1" to indicate the ability to integrate information.
  • the computing power requirement information includes the required calculation amount, and may also include at least one of the maximum acceptable computing time, task parallelism, and algorithm complexity.
  • the computing power request can also carry contract information, such as the price paid for the computing power demand information.
  • the price can be cash or computing power, or other valuable rewards.
  • Send a collaborative inquiry to the computing collaboration terminal which includes a request to provide a computing power report; when the computing power sharing network element is connected to the network for control plane transmission, it is determined based on the integration capability information that the computing demand terminal does not have information integration capability or computing When the demand side does not send its integration capability information, the collaborative inquiry also includes an integration capability report request;
  • the computing power sharing network element After the computing power sharing network element receives the computing power request, it then sends a collaboration inquiry to the computing collaboration terminal (for example, computing collaboration terminal-1, computing collaboration terminal-2, computing collaboration terminal-3) according to the computing power request. Including the available computing power report request, so that the computing collaboration terminal will return its ID and available computing power information according to the available computing power reporting request.
  • the computing collaboration terminal for example, computing collaboration terminal-1, computing collaboration terminal-2, computing collaboration terminal-3
  • the collaborative inquiry when it is determined based on the integration capability information that the computing demand end does not have the information integration capability, or the computing power request does not carry the integration capability information, the collaborative inquiry also includes an integration capability report request to request the computing collaboration terminal to send the computing power sharing network element Report its integration capability information.
  • the collaboration query response information includes the computing collaboration terminal ID and computing power information returned when responding to the computing power reporting request, and it is returned when responding to the integration capability reporting request. Information on the integration capabilities of the computing collaboration end.
  • the computing collaboration end has the computing power that can be provided
  • the ID of the computing collaboration terminal and the computing power information that can be provided in response to the reporting request for providing computing power are returned. Integration capability information at the end.
  • the computing power sharing process can be initiated in response to a computing power request initiated by the computing demand side, is initiated actively, and is suitable for actual demand scenarios, where the demand side initiates the request independently.
  • step 201 includes:
  • the computing power provision request includes the ID of the computing collaboration terminal and information that can provide computing power.
  • the request for force provision also includes information on the integration capability;
  • the computing power sharing network element receives the computing power provision request sent by the computing collaboration end, where whether the computing power provision request carries the integration capability information is optional.
  • the computing collaboration end has the information integration capability, it can choose to carry or Do not carry.
  • the computing collaboration terminal expects to perform task allocation and information integration on computing tasks during computing power sharing, it will carry its integration capability information in the computing power provision request.
  • the computing demand end which includes a computing power demand report request.
  • the demand query also includes a computing power report request.
  • the demand inquiry also includes the integration capability report request;
  • the computing power sharing network element sends a demand query to the computing demand terminal (for example, computing demand terminal-1, computing demand terminal-2, and computing demand terminal-3) according to the computing power provision request.
  • the demand query includes a computing power demand reporting request,
  • the demand query can also include a computing power report request.
  • the computing demand side is expected to perform task assignment and information integration of computing tasks during computing power sharing, the demand query can also include integration capabilities.
  • the report request that is, whether the computing power report request and the integration capability report request are carried is optional.
  • the computing power demand report request is used to request the computing demand side to return its ID and computing power demand information
  • the computing power report request is used to request the computing demand side to return its participating computing power information
  • the integrated capacity reporting request is used to request the computing demand side to return it. Integrate capability information.
  • the demand query response information includes the computing demand terminal ID and computing power demand information returned when responding to the computing power demand report, and the computing demand terminal returned when responding to the computing power reporting request. Can participate in computing power information, and the integration capability information of the computing demand side returned when responding to the integration capability reporting request.
  • the computing demand side returns its ID and computing power demand information when responding to the computing power demand report, or it can choose not to respond; the computing demand side returns its participating computing power information when responding to the computing power demand reporting request, or it can choose not to Respond; the computing demand side returns its integration capability information when responding to the integration capability reporting request, or it can choose not to respond.
  • the computing cooperation terminal actively initiates a request for computing power to start the computing power sharing process.
  • step 202 includes:
  • 501 Determine whether to provide computing power for the computing demand side according to the computing power demand information, the participating computing power information of the computing power sharing network element, the computing power information of the computing demand side, and the computing power information that can be provided on the computing collaboration terminal; specifically, When it can participate in computing power information and can provide computing power information to meet the computing power demand determined by the computing power demand information, provide computing power for the computing demand side and determine the first terminal that provides computing power for the computing demand side, the first terminal It includes at least one of a computing power sharing network element, a computing demand terminal, and a computing collaboration terminal.
  • the computing power sharing network element determines whether to provide computing power for the computing demand side, that is, determining the available computing power (the computing power that the computing power sharing network element can participate in, and the computing collaboration terminal can provide Whether the computing power and the computing power that can participate in the computing demand side) can meet the computing power demand of the computing demand side, and provide computing power assistance for it.
  • the participating computing power information of the computing power sharing network element may include the amount of calculation that the computing power sharing network element can provide, the time required to provide the above computing amount, the maximum acceptable task parallelism, and the complexity of the algorithm that can be processed Spend.
  • the computing power sharing network element determines whether it can meet the computing power demand determined by the computing power demand information by matching computing power demand information, participating computing power information, and providing computing power information.
  • the computing power sharing network element receives a computing power demand information, where the required calculation size is 20M, the maximum acceptable calculation time is 20 minutes, the task parallelism is 3, and the algorithm complexity is Take the time complexity as an example.
  • the time complexity is O(n), which means that the amount of data is increased by several times, and the time consumption is also increased by several times.
  • the computing capacity that the computing power sharing network element can provide is 1M.
  • the required time is 10 minutes, the task parallelism is 3, and the algorithm complexity is O(n).
  • the computing power information that the computing demand side can participate in received by the computing power sharing network element is: the calculation amount is 15M.
  • the required time is 5 minutes, the task parallelism is 3, and the algorithm complexity is O(n).
  • the computing power information received by the computing power sharing network element from the computing collaboration terminal is: the calculation amount of the first computing collaboration terminal is 50M, the required time is 20 minutes, the task parallelism is 3, and the algorithm complexity is O(n) ;
  • the calculation amount of the second computing collaboration end is 5M, the required time is 20 minutes, the task parallelism is 2, and the algorithm complexity is O(n);
  • the calculation amount of the third computing collaboration end is 15M, and the required time is 20 minutes,
  • the task parallelism is 2, and the algorithm complexity is O(n).
  • the computing power sharing network element calculates various possible computing power sharing solutions that meet the computing power requirements, and selects one of them as the final computing power sharing solution according to the preset selection conditions.
  • the preset selection condition can be the maximum task parallelism (ie Choose the solution with the largest number of terminals participating in the computing power sharing process), the smallest task parallelism, the least time, etc.; the selection conditions can be determined according to specific needs. For example, taking the selection condition as the maximum power of task parallelism, two computing power sharing schemes can be determined according to Table 1. The first is that the computing power sharing network element, the computing demand side and the first computing cooperation terminal provide the computing demand side. In the second type of computing power, the second computing collaboration terminal and the third computing collaboration terminal provide computing power for the computing demand side.
  • the first scheme Since the first scheme has three parties participating in computing power sharing, while the second scheme only has two parties participating in computing power sharing, it can be determined that the final computing power sharing solution is the first one, which provides computing power terminals for the computing demand side.
  • the first terminal is a computing power sharing network element, a computing demand terminal, and a first computing collaboration terminal. Assuming that the solution that meets the requirements of computing power demand information cannot be calculated using Table 1, and it is determined that computing power assistance cannot be provided for the computing demand side, it will not respond to the computing power request of the computing demand side or not respond to the computing power supply request of the computing collaboration terminal. .
  • the computing power sharing network element When the computing power sharing network element is connected to the network for control plane transmission and data plane transmission, according to the integration capability information of the computing demand side and the integration capability information of the computing collaboration terminal, from the computing power sharing network element, the computing demand side and the computing collaboration terminal
  • the second terminal for task distribution and information integration is determined in the process; or, when the computing power sharing network element is connected to the network for control plane transmission, according to the integration capability information of the computing demand side and the integration capability information of the computing collaboration terminal, the computing demand side and Determine the second terminal for task allocation and information integration in the computing cooperation terminal;
  • step 501 After it is determined in step 501 that the computing power can be provided for the computing demand side, it is necessary to determine the second terminal for task allocation and information integration in the computing power sharing process according to the integration capability information.
  • the computing power sharing network element When the computing power sharing network element is connected to the network for control plane transmission and data plane transmission, that is, the computing power sharing network element has the information integration capability, the computing power sharing is based on the integration capability information of the computing demand side and the integration capability information of the computing collaboration side.
  • the network element, the computing demand side, and the computing collaboration side determine the second terminal for task allocation and information integration.
  • the integration capability information of the computing demand side indicates that the computing demand side does not have the information integration capability
  • the computing power sharing network element is preferred as the first terminal.
  • Two terminals when the computing power sharing network element cannot be used as the second terminal due to heavy load or other reasons, a suitable computing collaboration terminal is selected as the second terminal according to the integration capability information of the computing collaboration terminal.
  • the computing demand side has information integration capabilities, you can set rules in advance to set the priority to select the computing demand side as the second terminal; in order to reduce the amount of calculation on the computing demand side, you can also set the priority to select the computing power sharing network element as the second terminal.
  • Second terminal choose according to the actual situation.
  • the computing power sharing network element When the computing power sharing network element is connected to the network for control plane transmission, that is, the computing power sharing network element does not have the information integration capability, according to the integration capability information of the computing demand end and the integration capability information of the computing collaboration end, from the computing demand end and The computing collaboration terminal determines the second terminal for task allocation and information integration. Assuming that the computing demand terminal has information integration capability, the computing collaboration terminal does not report the integration capability information, and the second terminal at this time is the computing demand terminal. When the computing demand terminal does not have the information integration capability, it is necessary to select one (at least one) of the computing collaboration terminals that have reported the integration capability information as the second terminal.
  • the computing collaboration terminal that first reports the integration capability information may be selected as the second terminal, or the computing collaboration terminal that is more suitable based on the performance of the computing collaboration terminal may be selected as the second terminal.
  • the method further includes:
  • the computing power distribution list is the correspondence between the ID of the first terminal and the assigned computing power. ;
  • the computing power sharing network element performs computing power scheduling according to the computing power demand information and the corresponding available computing power information of the first terminal to obtain the computing power distribution list.
  • the computing task is 20M calculations
  • the first terminal is a computing power sharing network element, a computing demand side, and a computing collaboration terminal
  • the computing power sharing network element is a computing power sharing network element based on the available computing power information of the first terminal.
  • the calculation amount allocated by the computing demand side and the computing cooperation side are 2M, 10M and 8M respectively.
  • the first terminal is authorized at this time (assuming the computing cooperation terminal-1, computing power Collaboration terminal-2) Provide computing services.
  • the computing power sharing network element sends authorization information to the first terminal.
  • the authorization information includes the ID of the computing demand terminal and computing power demand information.
  • the authorization information also includes Signing information.
  • the first terminal returns the ID of the first terminal to the computing power sharing network element as an authorization acceptance response.
  • the second terminal is not a computing power sharing network element, and the second terminal is a computing demand terminal or a computing collaboration terminal, it is also necessary to authorize the second terminal to provide integrated services.
  • the computing collaboration terminal is a second terminal, authorize The information includes the ID of the computing demand terminal, the computing power demand information, and the computing power distribution list.
  • the computing collaboration terminal is also the first terminal, it is authorized to provide integrated services when it is authorized to provide computing services.
  • the authorization information includes the ID of the computing demand terminal, computing power demand information, and computing power distribution list.
  • the authorization information includes the computing power distribution list.
  • the computing collaboration end or the computing demand end returns their respective ID to the computing power sharing network element as an authorization acceptance response.
  • the computing power request of the computing demand side has been processed.
  • the computing power sharing network element returns consent information to the computing demand side.
  • the consent information includes the ID of the first terminal and the ID of the second terminal.
  • the agreement information does not need to include the ID of the second terminal. If it cannot provide computing power assistance for the computing demand side, it will not respond to computing power requests.
  • the computing demand terminal is now authorized to apply for computing power, and the computing power sharing network element sends the calculation
  • the demand end sends authorization information, and the authorization information includes the ID of the second terminal, the ID of the first terminal, and the available computing power information.
  • the first terminal returns its ID to the computing power sharing network element as an authorization acceptance response.
  • the second terminal is not a computing power sharing network element, and the second terminal is a computing demand terminal or a computing collaboration terminal, it is also necessary to authorize the second terminal to provide integrated services.
  • the computing collaboration terminal is a second terminal
  • authorize The information includes the ID of the computing demand terminal, the computing power demand information, and the computing power distribution list.
  • the authorization information includes the computing power distribution list.
  • the computing collaboration end or the computing demand end returns their respective ID to the computing power sharing network element as an authorization acceptance response.
  • the computing power providing request of the computing collaboration terminal has been processed, and when the computing power assistance can be provided to the computing demand terminal, the computing power sharing network element returns consent information to the computing collaboration terminal, and the consent information includes the ID of the computing demand terminal. If it is unable to provide computing power assistance to the computing demand side, it will not respond to computing power provision requests.
  • the second terminal After completing the application for authorization, the second terminal obtains the computing task sent by the computing demand terminal, distributes the computing task to the first terminal according to the computing power distribution list, and receives the sub-processing results returned by the first terminal, and integrates the sub-processing results to obtain the computing task And return the total processing result to the computing demand side.
  • the second terminal is the computing demand terminal.
  • FIG. 5 is a schematic diagram of the scheduling and allocation process of a computing power sharing method provided by an embodiment of the present application; in FIG. 5, the computing task is taken as an example of a training model.
  • the computing demand end processes according to the training model and the computing power distribution list, and completes the task assignment. For example, the computing demand end sends the training model and their respective training tasks to the computing cooperation end-1 and the computing cooperation end-2.
  • the computing collaboration terminal-1 and computing collaboration terminal-2 After the computing collaboration terminal-1 and computing collaboration terminal-2 complete the training, they upload the sub-processing results, such as the local model, to the computing demand terminal.
  • the computing demand side After the computing demand side updates the training model based on the weights uploaded by the computing collaboration terminal-1 and computing collaboration terminal-2, it determines whether the training result has converged. If there is no convergence, the computing collaboration terminal-1 and computing collaboration terminal-2 continue to train and upload Sub-processing results, until the training result is convergent, the computing demand side sends a training end notification to the computing collaboration terminal-1 and computing collaboration terminal-2, and the computing power provided by each first terminal for the computing task and the amount spent The time is sent to the computing power sharing network element.
  • the second terminal is a computing cooperative terminal.
  • FIG. 6 is a schematic diagram of the scheduling and allocation process of a computing power sharing method provided by an embodiment of the present application; in FIG. 6, it is assumed that the second terminal is a computing power Collaboration terminal-1, the computing demand terminal sends the computing task to the computing collaboration terminal-1.
  • the computing task takes the training model as an example.
  • the computing collaboration terminal-1 performs processing according to the training model and the computing power distribution list to complete the task assignment, for example, computing collaboration Terminal-1 sends the training model and their respective training tasks to the computing demand terminal and the computing collaboration terminal-2.
  • the computing cooperation terminal-1 updates the training model based on the weights uploaded by the computing demand terminal and the computing cooperation terminal-2, it is judged whether the training result has converged. If there is no convergence, the computing demand terminal and the computing cooperation terminal-2 continue to perform training and upload sub-processing As a result, until the training result is convergent, the computing cooperation terminal-1 sends a training end notification to the computing demand terminal and the computing cooperation terminal-2.
  • the computing power provided by each first terminal for the computing task and the time spent are sent to the computing power sharing network element, and the training result is also sent to the computing demand terminal, and the training result is the trained model.
  • the computing power distribution list is obtained by scheduling in the computing power sharing network element, so as to reduce the calculation amount of the second terminal.
  • the computing power requirement information includes the type of data that needs to be collected
  • the first terminal returns the data that needs to be collected as part of the sub-processing result to the second terminal.
  • the method further includes:
  • the available computing power information of the first terminal includes at least one of the computing power that can be provided by the computing power sharing network element, the computing power that can be provided by the computing demand side, and the computing power that can be provided by the computing cooperation terminal. Since the computing power sharing network element has the ID and computing power information of all terminals, that is, it has the ID of the first terminal and the available computing power information. Therefore, when the computing power sharing network element is the second terminal, there is no need to send the first terminal The ID and available computing power information to the second terminal.
  • FIG. 7 is a schematic diagram of the scheduling and allocation flow of a computing power sharing method provided by an embodiment of the present application; the computing demand side sends computing tasks to the computing power sharing network element
  • the computing task takes the training model as an example.
  • the computing power sharing network element processes the training model and the available computing power information of the first terminal to complete the task allocation. For example, the computing amount required for the training model is 20M.
  • the first terminal at this time is the computing cooperation terminal-1 and the computing cooperation terminal-2, and their available computing power information is 3M and 20M respectively.
  • the computing power sharing network element delivers the training model and their respective training tasks to the computing collaboration terminal-1 and computing collaboration terminal-2.
  • the computing collaboration terminal-1 and computing collaboration terminal-2 complete the training, they upload the sub-processing results, such as the local area model, to the computing power sharing network element.
  • the computing power sharing network element updates the training model based on the weights uploaded by the computing collaboration terminal-1 and computing collaboration terminal-2, it determines whether the training result has converged. If there is no convergence, the computing collaboration terminal-1 and computing collaboration terminal-2 continue training And upload the sub-processing result, until the training result is convergent, the computing power sharing network element sends the training result to the computing demand end, that is, the trained model.
  • the second terminal is a computing demand terminal or a computing collaboration terminal
  • the method of sending the ID of the first terminal and the available computing power information to the second terminal and the sending method of the aforementioned computing power distribution list Similar, not repeat them here.
  • the computing demand terminal and the computing cooperation terminal After receiving the ID of the first terminal and the available computing power information, the computing demand terminal and the computing cooperation terminal perform task allocation and information integration according to the available computing power information.
  • the computing power sharing method further includes:
  • the computing power sharing network element obtains the computing power provided by the first terminal for the computing task and the time spent, where the second terminal can start timing when the task is issued to the first terminal, and the first terminal receives The returned sub-processing result stops timing. If the calculation task only needs to be executed once, the obtained statistical time is taken as the time spent by the first terminal for the calculation task, and the calculation amount provided by the first terminal for completing the assigned task As the computing power provided by the first terminal for the computing task. If the calculation task is executed X times repeatedly, the sum of X counts of time needs to be taken as the time spent by the first terminal for the calculation task. Similarly, the sum of X calculations is taken as the first terminal for the calculation The amount of computing power provided by the task.
  • FIG. 8 is a schematic structural diagram of a computing power sharing network element provided by an embodiment of the present application.
  • the computing power sharing network element is connected to the network for control plane transmission and data plane transmission, or the computing power sharing network element is connected to the network for control plane transmission; the computing power sharing network element includes:
  • the obtaining module 801 is used to obtain the computing power demand information sent by the computing demand terminal and the available computing power information sent by the computing cooperation terminal;
  • the processing module 802 is configured to determine the computing collaboration terminal that provides computing power for the computing demand terminal based on computing power demand information and the computing power information that can be provided, and directly or indirectly instruct the computing collaboration terminal that provides computing power to complete the computing task of the computing demand terminal.
  • the network includes a core network
  • the computing power sharing network element interacts with the core network unified data management network element through the first reference point, the computing power sharing network element interacts with the core network access and mobility management network element through the second reference point, and the computing power sharing network element Interact with the user plane network element of the core network through the third reference point;
  • the computing power sharing network element interacts with the unified data management network element of the core network through the first reference point, and the computing power sharing network element interacts with the access and mobility management network element of the core network through the second reference point.
  • the acquisition module includes:
  • the first receiving sub-module is used to receive the computing power request sent by the computing demand side.
  • the computing power request includes the ID of the computing demand side and computing power demand information.
  • the computing power request also includes the computing power request. Can participate in computing power information.
  • the computing power request also includes the integration capability information of the computing demand side;
  • the first sending sub-module is used to send a collaborative inquiry to the computing collaboration end.
  • the collaborative inquiry includes a request to provide a computing power report; when the computing power sharing network element is connected to the network for control plane transmission, it is determined based on the integration capability information to determine whether the computing demand end is not.
  • the collaborative inquiry also includes the integration capability reporting request;
  • the first receiving sub-module is also used to receive the collaboration query response information returned by the computing collaboration terminal in response to the collaboration query.
  • the collaboration query response information includes the computing collaboration terminal ID and computing power information returned in response to the computing power reporting request, and The integration capability information of the computing collaboration terminal returned in response to the integration capability reporting request.
  • the acquiring module includes:
  • the second receiving sub-module is used to receive the computing power providing request sent by the computing cooperation terminal.
  • the computing power providing request includes the ID of the computing collaboration terminal and the available computing power information.
  • the computing collaboration terminal expects to allocate computing tasks when computing power is shared.
  • the request for the provision of computing power also includes the integration capability information;
  • the second sending sub-module is used to send a demand query to the computing demand side.
  • the demand query includes a computing power demand reporting request.
  • the computing demand side is expected to participate in computing power sharing
  • the demand query also includes a computing power reporting request.
  • demand inquiries also include integration capabilities reporting requests;
  • the second receiving sub-module is also used to receive the demand query response information returned by the computing demand terminal in response to the demand query.
  • the demand query response information includes the computing demand terminal ID and computing power demand information returned in response to the computing power demand reporting request, and the response computing power The participating computing power information of the computing demand side returned when the report request, and the integration capability information of the computing demand side returned in response to the integration capability reporting request.
  • the processing module includes:
  • the first processing sub-module is used to determine whether it is provided for the computing demand side according to the computing power demand information, the participating computing power information of the computing power sharing network element, the computing power participating computing power information of the computing demand side, and the computing power providing information of the computing collaboration terminal Computing power; specifically, when it can participate in computing power information and can provide computing power information to meet the computing power demand determined by the computing power demand information, provide computing power for the computing demand side and determine the first computing power to provide computing power for the computing demand side A terminal, where the first terminal includes at least one of a computing power sharing network element, a computing demand terminal, and a computing cooperation terminal.
  • the processing module includes:
  • the second processing sub-module is used when the computing power sharing network element is connected to the network for control plane transmission and data plane transmission, and task allocation is determined from the computing power sharing network element and the computing demand side according to the integration capability information of the computing demand side
  • the second terminal integrated with the information; or, when the computing power sharing network element is connected to the network for control plane transmission, according to the integration capability information of the computing demand side and the integration capability information of the computing cooperation terminal, from the computing demand side and the computing cooperation terminal Determine the second terminal for task allocation and information integration in the
  • the third processing sub-module is used to send the ID of the first terminal and the corresponding usable computing power information to the second terminal, so that the second terminal can obtain the computing task sent by the computing demand terminal, and calculate according to the usable computing power information
  • the task is allocated to the first terminal, and the sub-processing result returned by the first terminal is received, and the sub-processing result is integrated to obtain the total processing result of the calculation task.
  • the processing module when the second terminal is a computing demand terminal or a computing cooperation terminal, the processing module includes:
  • the fourth processing sub-module is used to perform computing power scheduling according to the computing power demand information and the corresponding available computing power information of the first terminal to obtain a computing power distribution list, which is the ID of the first terminal and the assigned computing power Correspondence of force;
  • the fifth processing sub-module is used to send the computing power distribution list to the second terminal, so that the second terminal obtains the computing tasks sent by the computing demand terminal, and distributes the computing tasks to the first terminal according to the computing power distribution list, and receives the first terminal.
  • the sub-processing result returned by a terminal is integrated to obtain the total processing result of the computing task.
  • the computing power sharing network element further includes:
  • the storage module is used to obtain and store the computing power provided by each first terminal for the computing task and the time spent.
  • the computing power requirement information includes the amount of computing required by the computing task.
  • the computing power requirement information also includes at least one of the maximum acceptable computing time for completing the computing task, the maximum task parallelism, the algorithm complexity of the computing task, and the specified collection data type.
  • each unit or module in the said computing power sharing network element can be separately or completely combined into one or several other units or modules, or some of the units or modules can also be split into functions. It is composed of multiple smaller units or modules, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application.
  • the above-mentioned units or modules are divided based on logical functions. In practical applications, the function of one unit (or module) can also be realized by multiple units (or modules), or the function of multiple units (or modules) can be implemented by one unit. (Or module) implementation.
  • FIG. 9 is a schematic structural diagram of a computing power sharing network element provided by an embodiment of the present application.
  • An embodiment of the present application also provides a computing power sharing network element, including: a transceiver 903, a processor 902, and a computer program stored in the memory 901 and executed by the memory 901; the processor 902 is connected to the transceiver 903 and the memory respectively.
  • the 901 connection is used to implement the aforementioned computing power sharing method by executing a computer program.
  • the transceiver 903 may include a transceiving antenna, or a network card, and other physical results with information transceiving functions.
  • the memory 901 may include storage devices of various storage media.
  • the processor 902 may be various processing devices such as a central processing unit, a microprocessor, a digital signal processor, an application processor, a programmable array, and an application specific integrated circuit.
  • the processor 902 may be connected to the transceiver 903 and the memory 901 through an integrated circuit bus or the like, and may be executed by a computer program to realize the foregoing computing power sharing method. It should be understood that the computing power sharing network element described in the embodiment of the present application can execute the computing power sharing method described above, and will not be repeated here. In addition, the description of the beneficial effects of using the same method will not be repeated.
  • FIG. 10 is a schematic structural diagram of a computing power sharing system provided by an embodiment of the present application.
  • a computing power sharing system includes a computing demand terminal 101, a computing cooperation terminal 103, and the foregoing The computing power sharing network element 102.
  • the embodiment of the present application also provides a computer storage medium, and the computer storage medium stores the computer program executed by the aforementioned computing power sharing network element, and the computer program Including program instructions.
  • the processor executes the program instructions, it can execute the foregoing description of the computing power sharing method. Therefore, it will not be repeated here.
  • the description of the beneficial effects of using the same method will not be repeated.
  • the program can be stored in a computer readable storage medium. During execution, it may include the procedures of the above-mentioned method embodiments.
  • the storage medium can be a magnetic disk, an optical disc, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请实施例公开一种算力共享方法及相关设备，本申请实施例中，算力共享网元与网络连接以进行控制面传输和数据面传输，或者，算力共享网元与网络连接以进行控制面传输；其中，算力共享网元获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；再根据算力需求信息和可提供算力信息，确定为计算需求端提供算力的计算协作端，并直接或间接指示提供算力的计算协作端完成计算需求端的计算任务，实现算力共享，提高计算速度，减少计算所需时间。

Description

算力共享方法及相关设备
本申请要求于2020年4月13日递交的申请号为“202010288252.4”、发明名称为“算力共享方法及相关设备”的发明申请的优先权，上述在先申请的内容通过引用结合在本申请中。
技术领域
本申请涉及计算技术领域，尤其涉及一种算力共享方法及相关设备。
背景技术
随着大数据时代的到来，算力在数字经济时代成为一种重要的生产力。在新兴科技不断突破的过程中，算力作为驱动技术进步的“燃料”，成为一种新兴“能源”。
现有技术中，有些应用需要使用巨大的计算能力才能完成，如果采用集中式计算，需要耗费相当长的时间，亟需解决这一技术问题。
发明内容
本申请实施例提供了一种算力共享方法及相关设备，可以实现算力共享。
第一方面，本申请实施例提供了一种算力共享方法，应用于算力共享网元中，所述算力共享网元与网络连接以进行控制面传输和数据面传输，或者，所述算力共享网元与网络连接以进行控制面传输；
所述方法包括：
获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；
根据所述算力需求信息和所述可提供算力信息，确定为所述计算需求端提供算力的计算协作端，并直接或间接指示所述提供算力的计算协作端完成所述计算需求端的计算任务。
可选地，所述获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息，包括：
接收所述计算需求端发送的算力请求，所述算力请求包括所述计算需求端 的ID、算力需求信息，所述计算需求端期望参与算力共享时，所述算力请求还包括所述计算需求端的可参与算力信息，所述计算需求端期望在算力共享时对所述计算任务进行任务分配和信息整合时，所述算力请求还包括和所述计算需求端的整合能力信息；
向所述计算协作端发送协作询问，所述协作询问包括可提供算力上报请求；所述算力共享网元与网络连接以进行控制面传输时，根据所述整合能力信息确定所述计算需求端不具备信息整合能力或所述计算需求端未发送所述整合能力信息时，所述协作询问还包括整合能力上报请求；
接收所述计算协作端响应所述协作询问返回的协作询问响应信息，所述协作询问响应信息包括响应所述可提供算力上报请求时返回的所述计算协作端的ID和可提供算力信息，以及响应所述整合能力上报请求时返回的所述计算协作端的整合能力信息。
可选地，所述获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息，包括：
接收所述计算协作端发送的算力提供请求，所述算力提供请求包括所述计算协作端的ID、可提供算力信息，所述计算协作端期望在算力共享时对所述计算任务进行任务分配和信息整合时，所述算力提供请求还包括整合能力信息；
向所述计算需求端发送需求询问，所述需求询问包括算力需求上报请求，期望所述计算需求端参与算力共享时，所述需求询问还包括算力上报请求，期望所述计算需求端在算力共享时对所述计算任务进行任务分配和信息整合时，所述需求询问还包括整合能力上报请求；
接收所述计算需求端响应所述需求询问返回的需求询问响应信息，所述需求询问响应信息包括响应所述算力需求上报请求时返回的所述计算需求端的ID和算力需求信息，响应所述算力上报请求时返回的所述计算需求端的可参与算力信息，以及响应所述整合能力上报请求时返回的所述计算需求端的整合能力信息。
可选地，所述根据所述算力需求信息和所述可提供算力信息，确定为所述计算需求端提供算力的计算协作端，包括：
根据所述算力需求信息、所述算力共享网元的可参与算力信息、所述计算需求端的可参与算力信息和所述计算协作端的可提供算力信息确定是否为所述计算需求端提供算力；
当所述可参与算力信息、所述可提供算力信息可满足由所述算力需求信息确定的算力需求时，为所述计算需求端提供算力并确定为所述计算需求端提供算力的第一终端，所述第一终端包括所述算力共享网元、所述计算需求端和所述计算协作端中至少一种。
可选地，所述直接或间接指示所述提供算力的计算协作端完成所述计算需求端的计算任务，包括：
所述算力共享网元与网络连接以进行控制面传输和数据面传输时，根据所述计算需求端的整合能力信息和所述计算协作端的整合能力信息，从所述算力共享网元、所述计算需求端、所述计算协作端中确定进行任务分配和信息整合的第二终端；或者，所述算力共享网元与网络连接以进行控制面传输时，根据所述计算需求端的整合能力信息和所述计算协作端的整合能力信息，从所述计算需求端和所述计算协作端中确定进行任务分配和信息整合的第二终端；
发送所述第一终端的ID和对应的可使用算力信息至所述第二终端，以使所述第二终端获取所述计算需求端发送的计算任务，并根据所述可使用算力信息将所述计算任务分配至所述第一终端，以及接收所述第一终端返回的子处理结果，整合所述子处理结果得到所述计算任务的总处理结果。
可选地，确定进行任务分配和信息整合的第二终端后，所述第二终端为计算需求端或计算协作端时，所述方法还包括：
根据所述算力需求信息和对应的所述第一终端的可使用算力信息进行算力调度，得到算力分配清单，所述算力分配清单为所述第一终端的ID和分配的算力的对应关系；
发送所述算力分配清单至所述第二终端，以使所述第二终端获取所述计算需求端发送的计算任务，并根据所述算力分配清单将所述计算任务分配至所述第一终端，以及接收所述第一终端返回的子处理结果，整合所述子处理结果得到所述计算任务的总处理结果。
可选地，所述方法还包括：
获取并存储各所述第一终端为所述计算任务所提供的算力大小和所花费的时间。
可选地，所述网络包括核心网；
所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互，所述算力共享网元通过第三参考点与所述核心网的用户平面网元交互；
或者，
所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互。
可选地，所述算力需求信息包括所述计算任务所需的计算量。
可选地，所述算力需求信息还包括完成所述计算任务可接受的最长计算时间，最大任务并行度，计算任务的算法复杂度、指定采集数据类型中的至少一种。
第二方面，本申请实施例提供了一种算力共享网元，所述算力共享网元与网络连接以进行控制面传输和数据面传输，或者，所述算力共享网元与网络连接以进行控制面传输；
所述算力共享网元包括：
获取模块，用于获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；
处理模块，用于根据所述算力需求信息和所述可提供算力信息，确定为所述计算需求端提供算力的计算协作端，并直接或间接指示所述提供算力的计算协作端完成所述计算需求端的计算任务。
可选地，所述网络包括核心网；
所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互，所述算力共享网元通过第三参考点与所述核心网的用户平面网元交互；
或者，
所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互。
第三方面，本申请实施例提供了一种算力共享网元，包括：收发器、处理器、存储器以及存储在存储器上并由所述处理器执行的计算机程序；
所述处理器分别与所述收发器、所述存储器相连，用于通过执行所述计算机程序，实现第一方面所述的算力共享方法。
第四方面，本申请实施例提供了一种算力共享系统，包括计算需求端、计算协作端和第二方面或第三方面所述的算力共享网元。
第五方面，本申请实施例提供了一种计算机存储介质，所述计算机存储介质存储有计算机程序，所述计算机程序包括程序指令，所述程序指令当被处理器执行时，执行如第一方面所述的算力共享方法。
本申请实施例中，算力共享网元与网络连接以进行控制面传输和数据面传输，或者，算力共享网元与网络连接以进行控制面传输；其中，算力共享网元获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；再根据算力需求信息和可提供算力信息，确定为计算需求端提供算力的计算协作端，并直接或间接指示提供算力的计算协作端完成计算需求端的计算任务，实现算力共享，提高计算速度，减少计算所需时间。
附图说明
为了更清楚地说明本申请实施例技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1a是本申请实施例提供的一种算力共享网元和核心网的连接示意图；
图1b、图1c是本申请实施例提供的一种算力共享网元和核心网的参考点连接示意图；
图2是本申请实施例提供的一种算力共享方法的流程示意图；
图3是本申请实施例提供的一种算力共享方法的申请授权流程示意图；
图4是本申请实施例提供的一种算力共享方法的申请授权流程示意图；
图5是本申请实施例提供的一种算力共享方法的调度分配流程示意图；
图6是本申请实施例提供的一种算力共享方法的调度分配流程示意图；
图7是本申请实施例提供的一种算力共享方法的调度分配流程示意图；
图8是本申请实施例提供的一种算力共享网元的结构示意图；
图9是本申请实施例提供的一种算力共享网元的结构示意图；
图10是本申请实施例提供的一种算力共享系统的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述。
应当理解，本申请的说明书和权利要求书及附图中的术语“第一”、“第二”等是用于区别不同对象，而不是用于描述特定顺序。此外，术语“包括”和“具有”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。
在本申请中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本申请所描述的实施例可以与其它实施例相结合。
现有技术中，由于单一计算体的算力有限，无法快速完成计算量巨大的计算任务，因此，本申请提出一种算力共享方法，可以实现算力共享，以便快速完成计算任务。
本申请中，计算需求端可以是智能体，用户设备(User Equipment，UE)， 计算服务器，基站等，任何需要算力服务或者能够提供算力服务的载体，计算协作端可以是智能体，用户设备(User Equipment，UE)，计算服务器，基站等，任何需要算力服务或者能够提供算力服务的载体。计算需求端和计算协作端均具有唯一识别ID，ID可以是由数字、字母、特殊字符中的至少一种组成的序列。其中，计算需求端和计算协作端之间可以建立有3GPP(the 3rd Generation partnership project,第三代合作伙伴计划)连接以进行数据传输。
本申请提供一种算力共享方法，应用于算力共享网元中，算力共享网元与网络连接，网络包括核心网，算力共享网元是核心网的一个用于算力共享的新的网元，是属于运营商部署的实体，算力共享网元和核心网连接，参考图1a，是本申请实施例提供的一种算力共享网元和核心网的连接示意图，算力共享网元通过新增的第四接口。例如Nai接口连接核心网中其他网元的接口，以实现和核心网其他网元的交互。
在一个可能的实施例中，参考图1b、图1c，本申请实施例提供的一种算力共享网元和核心网的参考点连接示意图；图1b中，算力共享网元与核心网连接以进行控制面传输和数据面传输，具体地，算力共享网元通过新增的第一参考点N60与统一数据管理(Unified Data Management，UDM)网元实现交互，算力共享网元通过新增的第二参考点N59与接入和移动性管理(Access and Mobility Management Function，AMF)网元实现交互，以此实现控制面传输，算力共享网元具有调度能力。而算力共享网元通过新增的第三参考点N61与用户平面(User Plane Function，UPF)网元实现交互，以此实现数据面传输，即算力共享网元可以与其他终端设备实现大数据量传输，例如传输训练模型或训练集等，算力共享网元具有信息整合能力。
在另一个可能的实施例中，参考图1c，算力共享网元与核心网连接以进行控制面传输；具体地，算力共享网元通过新增的第一参考点N60与统一数据管理(Unified Data Management，UDM)网元实现交互，算力共享网元通过新增的第二参考点N59与接入和移动性管理(Access and Mobility Management Function，AMF)网元实现交互，以此实现控制面传输，算力共享网元具有调度能力。
请参见图2，是本申请实施例提供的一种算力共享方法的流程示意图；所述算力共享方法，包括：
201、获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；
具体地，算力共享网元获取计算需求端的算力需求信息，以及计算协作端的可提供算力信息，算力需求信息包括其计算任务所需的计算量大小，还可以包括完成计算任务可接受的最长计算时间，最大任务并行度N(指定最多可将计算任务分配给N个终端去处理)，计算任务的算法复杂度、指定采集数据类型中的至少一种。指定采集数据类型用于指定计算过程中，计算协作端需要采集的数据类型。以模型训练场景为例，如果指定的采集数据类型为速度，则计算协作端在计算过程中，需要收集自身的速度，以辅助其完成本地的训练任务。而可提供算力信息包括计算协作端可提供的计算量大小，可提供算力信息还可以包括提供上述计算量所需的时间、可以接受的最大任务并行度，以及可以处理的算法复杂度中的至少一种。
202、根据算力需求信息和可提供算力信息，确定为计算需求端提供算力的计算协作端，并直接或间接指示提供算力的计算协作端完成计算需求端的计算任务。
具体地，为计算需求端提供算力的计算协作端的数目可以是一个以上。算力共享网元可以直接指示提供算力的计算协作端去完成计算需求端的计算任务，也可以是间接指示提供算力的计算协作端去完成计算需求端的计算任务，以实现计算协作端的算力共享。间接指示时，算力共享网元可以指示计算需求端或计算协作端去控制提供算力的计算协作端完成计算需求端的计算任务。利用图2的方法，网络不再只是作为通信管道，而是具有调度功能，算力共享网元可以统筹调度计算需求端和计算协作端，以实现共享算力的目的，提高计算速度，减少计算任务所需的计算时间，提高计算效率，而且基于算力共享网元的算力共享方法，其信任度更高。
特别地，以算力共享网元直接指示提供算力的计算协作端完成计算需求端的计算任务为例，算力共享网元将计算任务分配给计算协作端，当计算协作端 完成分配的任务后，可以得到计算任务的子处理结果，可以由具有信息整合能力的算力共享网元来整合计算协作端的子处理结果，以得到计算任务的总处理结果，当总处理结果不满足要求时，算力共享网元将新的需要处理的任务发送给计算协作端，计算协作端重新进行任务处理，直到总处理结果符合规定为止，则算力共享网元将总处理结果返回给计算需求端。实际上，当计算需求端或计算协作端具有信息整合能力时，它们也能代替算力共享网元进行任务分配，以及进行信息整合得到计算任务的总处理结果。同理，当计算需求端或算力共享网元具有计算能力时，它们相当于计算协作端，也能参与算力共享。
在一个可能的实施例中，参考图3，图3是本申请实施例提供的一种算力共享方法的申请授权流程示意图；步骤201包括：
301、接收计算需求端发送的算力请求，算力请求包括计算需求端的ID、算力需求信息，计算需求端期望参与算力共享时，算力请求还包括计算需求端的可参与算力信息，计算需求端期望在算力共享时对计算任务进行任务分配和信息整合时，算力请求还包括计算需求端的整合能力信息；
具体地，算力共享网元接收计算需求端发送的算力请求，其中，算力请求必须包括计算需求端的唯一识别ID以及算力需求信息。而对于可参与算力信息、整合能力信息而言，是否在算力请求中携带是可选的。当计算需求端具有多余的计算力时，可以选择在算力请求中携带可参与算力信息以表示计算需求端期望参与算力共享，也可以选择不携带。可参与算力信息为计算需求端用于算力共享的算力大小信息，可以包括计算需求端可提供的计算量大小，还可以包括提供上述计算量所需的时间、可以接受的任务并行度，以及可以处理的算法复杂度中的至少一种。当计算需求端不具有多余计算力时，可以直接不携带可参与算力信息，或者是可参与算力信息用“0”来代替。同理，当计算需求端具有信息整合能力时，可以选择在算力请求中携带整合能力信息以表示计算需求端期望在算力共享时对计算任务进行任务分配和信息整合，也可以选择不携带。整合能力信息可以用“1”来表示具有信息整合能力。本实施例中，算力需求信息包括所需的计算量大小，还可以包括可接受的最长计算时间，任务并行度和算法复杂度中的至少一种。
特别地，为了激励算力共享，算力请求中还可以携带签约信息，例如为该算力需求信息所支付的代价，代价可以是现金或者计算能力，或者是其他具有价值的奖励。
302、向计算协作端发送协作询问，协作询问包括可提供算力上报请求；算力共享网元与网络连接以进行控制面传输时，根据整合能力信息确定计算需求端不具备信息整合能力或计算需求端未发送其整合能力信息时，协作询问还包括整合能力上报请求；
具体地，算力共享网元接收到算力请求后，接着根据算力请求向计算协作端(例如计算协作端-1、计算协作端-2、计算协作端-3)发送协作询问，协作询问包括可提供算力上报请求，以使计算协作端根据可提供算力上报请求返回其ID和可提供算力信息。
特别地，当根据整合能力信息确定计算需求端不具备信息整合能力时，或者算力请求未携带整合能力信息时，协作询问还包括整合能力上报请求，以请求计算协作端向算力共享网元上报其整合能力信息。
由于在计算需求端进行信息整合，可以提高计算任务的信息保密性，还可以减小算力共享网元的计算压力，因此，不管算力共享网元是否具备信息整合能力，当计算需求端具备信息整合能力时，都不需要计算协作端进行信息整合，也即不需要其上报自身的整合能力信息。
303、接收计算协作端响应协作询问返回的协作询问响应信息，协作询问响应信息包括响应可提供算力上报请求时返回的计算协作端的ID和可提供算力信息，以及响应整合能力上报请求时返回的计算协作端的整合能力信息。
具体地，计算协作端具有可提供的算力时，响应可提供算力上报请求返回计算协作端的ID和可提供算力信息，计算协作端具有信息整合能力时，响应整合能力上报请求返回计算协作端的整合能力信息。
可见，算力共享过程可以是响应计算需求端发起的算力请求而启动的，是主动发起的，适用于实际的需求场景，需求方自主发起请求。
在一个可能的实施例中，参考图4，图4是本申请实施例提供的一种算力共享方法的申请授权流程示意图；步骤201包括：
401、接收计算协作端发送的算力提供请求，算力提供请求包括计算协作端的ID、可提供算力信息，计算协作端期望在算力共享时对计算任务进行任务分配和信息整合时，算力提供请求还包括整合能力信息；
具体地，算力共享网元接收计算协作端发送的算力提供请求，其中，算力提供请求中是否携带整合能力信息是可选的，当计算协作端具有信息整合能力时，可以选择携带或者不携带。计算协作端期望在算力共享时对计算任务进行任务分配和信息整合时，则在算力提供请求中携带其整合能力信息。
402、向计算需求端发送需求询问，需求询问包括算力需求上报请求，期望计算需求端参与算力共享时，需求询问还包括算力上报请求，期望计算需求端在算力共享时对计算任务进行任务分配和信息整合时，需求询问还包括整合能力上报请求；
具体地，算力共享网元根据算力提供请求向计算需求端(例如计算需求端-1、计算需求端-2、计算需求端-3)发送需求询问，需求询问包括算力需求上报请求，当期望计算需求端参与算力共享时，需求询问中还可以包括算力上报请求，当期望计算需求端在算力共享时对计算任务进行任务分配和信息整合时，需求询问还可以包括整合能力上报请求，即算力上报请求和整合能力上报请求是否携带是可选的。算力需求上报请求用于请求计算需求端返回其ID和算力需求信息，算力上报请求用于请求计算需求端返回其可参与算力信息，整合能力上报请求用于请求计算需求端返回其整合能力信息。
403、接收计算需求端响应需求询问返回的需求询问响应信息，需求询问响应信息包括响应算力需求上报时返回的计算需求端的ID和算力需求信息，响应算力上报请求时返回的计算需求端的可参与算力信息，以及响应整合能力上报请求时返回的计算需求端的整合能力信息。
具体地，计算需求端在响应算力需求上报时返回其ID和算力需求信息，也可以选择不响应；计算需求端在响应算力上报请求时返回其可参与算力信息，也可以选择不响应；计算需求端在响应整合能力上报请求时返回其整合能力信息，也可以选择不响应。
可见，利用上述方法，通过计算协作端主动发起算力提供请求，以启动算 力共享过程。
在一个可能的实施例中，步骤202包括：
501、根据算力需求信息、算力共享网元的可参与算力信息、计算需求端的可参与算力信息和计算协作端的可提供算力信息确定是否为计算需求端提供算力；具体地，当可参与算力信息、可提供算力信息可满足由算力需求信息确定的算力需求时，为计算需求端提供算力并确定为计算需求端提供算力的第一终端，第一终端包括算力共享网元、计算需求端和计算协作端中至少一种。
具体地，参考图3和图4，算力共享网元确定是否为计算需求端提供算力，即确定可利用的算力(算力共享网元可参与的算力、计算协作端可提供的算力以及计算需求端可参与的算力)是否可以满足计算需求端的算力需求，为其提供算力帮助。其中，算力共享网元的可参与算力信息可以包括算力共享网元可提供的计算量大小、提供上述计算量所需的时间、可以接受的最大任务并行度，以及可以处理的算法复杂度。
因此，算力共享网元通过匹配算力需求信息、可参与算力信息和可提供算力信息，来确定是否能满足算力需求信息所确定的算力需求。
例如，假设当前时刻，算力共享网元接收到一个算力需求信息，其中，要求的计算量大小为20M，可接受的最长计算时间为20分钟，任务并行度为3，算法复杂度以时间复杂度为例，时间复杂度为O(n)，就代表数据量增大几倍，耗时也增大几倍。
而算力共享网元可提供的计算量为1M。所需时间为10分钟，任务并行度为3，算法复杂度为O(n)。算力共享网元接收到的计算需求端的可参与算力信息为：计算量为15M。所需时间为5分钟，任务并行度为3，算法复杂度为O(n)。而算力共享网元接收到的计算协作端的可提供算力信息为：第一计算协作端的计算量为50M，所需时间为20分钟，任务并行度为3，算法复杂度为O(n)；第二计算协作端的计算量为5M，所需时间为20分钟，任务并行度为2，算法复杂度为O(n)；第三计算协作端的计算量为15M，所需时间为20分钟，任务并行度为2，算法复杂度为O(n)。如表1所示。
表1
  计算量/M 所需时间/分钟 任务并行度 算法复杂度
算力共享网元 1 10 3 O(n)
计算需求端 15 5 3 O(n)
第一计算协作端 50 20 3 O(n)
第二计算协作端 5 20 2 O(n)
第三计算协作端 15 20 2 O(n)
算力共享网元通过计算各种符合算力需求的可能的算力共享方案，根据预设选择条件从中选择一个方案作为最终的算力共享方案，预设选择条件可以是任务并行度最大(即选择参与算力共享过程的终端数目最大的那个方案)、任务并行度最小、花费时间最少等；可以根据具体需要确定选择条件。例如，以选择条件为任务并行度最大威力，根据表1可以确定两种算力共享方案，第一种是由算力共享网元、计算需求端和第一计算协作端来为计算需求端提供算力，第二种是由第二计算协作端和第三计算协作端来为计算需求端提供算力。由于第一种方案有三方参与算力共享，而第二种方案只有两方参与算力共享，因此，可以确定最终的算力共享方案为第一种方案，为计算需求端提供算力的终端，即第一终端为算力共享网元、计算需求端和第一计算协作端。而假设利用表1无法计算得到符合算力需求信息的要求的方案时，确定不能为计算需求端提供算力帮助，则不响应计算需求端的算力请求，或者不响应计算协作端的算力提供请求。
502、算力共享网元与网络连接以进行控制面传输和数据面传输时，根据计算需求端的整合能力信息和计算协作端的整合能力信息，从算力共享网元、计算需求端和计算协作端中确定进行任务分配和信息整合的第二终端；或者，算力共享网元与网络连接以进行控制面传输时，根据计算需求端的整合能力信息和计算协作端的整合能力信息，从计算需求端和计算协作端中确定进行任务分配和信息整合的第二终端；
具体地，利用步骤501确定了能为计算需求端提供算力后，需要根据整合能力信息确定算力共享过程中进行任务分配和信息整合的第二终端。算力共享 网元与网络连接以进行控制面传输和数据面传输时，即算力共享网元具有信息整合能力，则根据计算需求端的整合能力信息和计算协作端的整合能力信息，从算力共享网元、计算需求端和计算协作端中确定进行任务分配和信息整合的第二终端，当计算需求端的整合能力信息表明计算需求端不具备信息整合能力时，优先选择算力共享网元为第二终端，当算力共享网元因为负载过重等原因不能作为第二终端时，根据计算协作端的整合能力信息选择合适的计算协作端为第二终端。而当计算需求端有信息整合能力时，可以预先设置规则，设定优先选择计算需求端作为第二终端；为了减少计算需求端的计算量，也可以是设定优先选择算力共享网元作为第二终端，根据实际情况进行选择。
而当算力共享网元与网络连接以进行控制面传输时，即算力共享网元不具备信息整合能力，则根据计算需求端的整合能力信息和计算协作端的整合能力信息，从计算需求端和计算协作端中确定进行任务分配和信息整合的第二终端，假设计算需求端具有信息整合能力，则计算协作端不上报整合能力信息，此时的第二终端为计算需求端。而当计算需求端不具备信息整合能力时，则需要从上报了整合能力信息的计算协作端中(至少一个)选择一个作为第二终端。可以是选择最先上报整合能力信息的那个计算协作端作为第二终端，也可以是根据计算协作端的性能选择更适合的那个计算协作端作为第二终端。
在一个可能的实施例中，确定了进行任务分配和信息整合的第二终端后，第二终端为计算需求端或计算协作端时，方法还包括：
601、根据算力需求信息和对应的第一终端的可使用算力信息进行算力调度，得到算力分配清单，算力分配清单为所述第一终端的ID和分配的算力的对应关系；
具体地，算力共享网元根据算力需求信息和对应的第一终端的可使用算力信息进行算力调度，得到算力分配清单。例如，计算任务为20M的计算量，第一终端为算力共享网元、计算需求端和计算协作端，算力共享网元根据第一终端的可使用算力信息，为算力共享网元、计算需求端和计算协作端分配的计算量分别为2M、10M和8M。
602、发送算力分配清单至第二终端，以使第二终端获取计算需求端发送 的计算任务，并根据算力分配清单将计算任务分配至第一终端，以及接收第一终端返回的子处理结果，整合子处理结果得到计算任务的总处理结果。
具体地，在确定了第二终端和得到算力分配清单后，参考图3，对于计算需求端主动发起的算力请求而言，此时授权第一终端(假设为计算协作端-1、计算协作端-2)提供计算服务，算力共享网元向第一终端发送授权信息，授权信息包括计算需求端的ID和算力需求信息，当有算力请求中有签约信息时，授权信息还包括签约信息。第一终端返回第一终端的ID至算力共享网元作为授权接受响应。
另外，当第二终端不为算力共享网元时，第二终端为计算需求端或者计算协作端，此时还需要授权第二终端提供整合服务，当计算协作端为第二终端时，授权信息包括计算需求端的ID、算力需求信息以及算力分配清单。当该计算协作端同时也作为第一终端时，在授权其提供计算服务时，同时授权其提供整合服务，授权信息包括计算需求端的ID、算力需求信息以及算力分配清单。当计算需求端为第二终端时，授权信息包括算力分配清单。计算协作端或计算需求端返回各自的ID至算力共享网元作为授权接受响应。
至此，计算需求端的算力请求处理完毕，当能为计算需求端提供算力帮助时，算力共享网元返回同意信息至计算需求端，同意信息包括第一终端的ID和第二终端的ID，当计算需求端作为第二终端时，则同意信息不需要包括第二终端的ID。如果不能为计算需求端提供算力帮助，则不响应算力请求。
参考图4，对于计算协作端发起的提供算力请求而言，在确定了第二终端和得到算力分配清单后，此时授权计算需求端进行算力申请服务，算力共享网元向计算需求端发送授权信息，授权信息包括第二终端的ID、第一终端的ID和可使用算力信息。第一终端返回其ID至算力共享网元作为授权接受响应。
另外，当第二终端不为算力共享网元时，第二终端为计算需求端或者计算协作端，此时还需要授权第二终端提供整合服务，当计算协作端为第二终端时，授权信息包括计算需求端的ID、算力需求信息以及算力分配清单。当计算需求端为第二终端时，授权信息包括算力分配清单。计算协作端或计算需求端返回各自的ID至算力共享网元作为授权接受响应。
至此，计算协作端的算力提供请求处理完毕，当能为计算需求端提供算力帮助时，算力共享网元返回同意信息至计算协作端，同意信息包括计算需求端的ID。如果不能为计算需求端提供算力帮助，则不响应算力提供请求。
完成申请授权后，第二终端获取计算需求端发送的计算任务，并根据算力分配清单将计算任务分配至第一终端，以及接收第一终端返回的子处理结果，整合子处理结果得到计算任务的总处理结果，并将总处理结果返回给计算需求端。
第一种情况，第二终端为计算需求端，参考图5，图5是本申请实施例提供的一种算力共享方法的调度分配流程示意图；图5中，计算任务以训练模型为例，计算需求端根据训练模型和算力分配清单进行处理，完成任务分配，例如计算需求端将训练模型和各自的训练任务下发给计算协作端-1和计算协作端-2。计算协作端-1和计算协作端-2完成训练后，上传子处理结果，例如局域模型至计算需求端。计算需求端基于计算协作端-1和计算协作端-2上传的权重更新训练模型后，判断训练结果是否收敛，如果没有收敛，则计算协作端-1和计算协作端-2继续进行训练和上传子处理结果，直到训练结果为收敛，则计算需求端发送训练结束通知给计算协作端-1和计算协作端-2，并将各个第一终端为该计算任务所提供的算力和所花费的时间发送给算力共享网元。
第二种情况，第二终端为某一计算协作端，参考图6，图6是本申请实施例提供的一种算力共享方法的调度分配流程示意图；图6中，假设第二终端为计算协作端-1，计算需求端发送计算任务至计算协作端-1，计算任务以训练模型为例，计算协作端-1根据训练模型和算力分配清单进行处理，完成任务分配，例如，计算协作端-1将训练模型和各自的训练任务下发给计算需求端和计算协作端-2。计算需求端和计算协作端-2完成训练后，上传子处理结果，例如局域模型至计算协作端-1。计算协作端-1基于计算需求端和计算协作端-2上传的权重更新训练模型后，判断训练结果是否收敛，如果没有收敛，则计算需求端和计算协作端-2继续进行训练和上传子处理结果，直到训练结果为收敛，则计算协作端-1发送训练结束通知给计算需求端和计算协作端-2。并将各个第一终端为该计算任务所提供的算力和所花费的时间发送给算力共享网元，还将发送 训练结果给计算需求端，训练结果即为训练好的模型。
通过在算力共享网元中进行调度得到算力分配清单，以减轻第二终端的计算量。上述调度分配过程中，以训练模型场景为例，假设算力需求信息中包括需要采集的数据类型，则第一终端将需要采集的数据作为子处理结果的一部分返回第二终端。
特别指出的是，除了第二终端根据算力分配清单进行任务分配的方法之外，在一个可能的实施例中，确定进行任务分配和信息整合的第二终端后，方法还包括：
701、发送第一终端的ID和对应的可使用算力信息至第二终端，以使第二终端获取计算需求端发送的计算任务，并根据可使用算力信息将计算任务分配至第一终端，以及接收第一终端返回的子处理结果，整合子处理结果得到计算任务的总处理结果。
具体地，第一终端的可使用算力信息包括算力共享网元可提供的计算量大小、计算需求端可提供的计算量大小和计算协作端可提供的计算量大小中的至少一个。由于算力共享网元具有所有终端的ID和算力信息，即具有第一终端的ID和可使用算力信息，因此，当算力共享网元为第二终端时，不需要发送第一终端的ID和可使用算力信息给第二终端。
以第二终端为算力共享网元为例，参考图7，图7是本申请实施例提供的一种算力共享方法的调度分配流程示意图；计算需求端发送计算任务至算力共享网元，图7中，计算任务以训练模型为例，算力共享网元根据训练模型和第一终端的可使用算力信息进行处理，完成任务分配，例如，训练模型所需的计算量为20M，此时的第一终端为计算协作端-1和计算协作端-2，它们的可使用算力信息分别为3M和20M，此时，可以分配计算量为2M的任务给计算协作端-1，分配计算量为18M的任务给计算协作端-2。算力共享网元将训练模型和各自的训练任务下发给计算协作端-1和计算协作端-2。计算协作端-1和计算协作端-2完成训练后，上传子处理结果，例如局域模型至算力共享网元。算力共享网元基于计算协作端-1和计算协作端-2上传的权重更新训练模型后，判断训练结果是否收敛，如果没有收敛，则计算协作端-1和计算协作端-2继续进行 训练和上传子处理结果，直到训练结果为收敛，则算力共享网元发送训练结果给计算需求端，即训练好的模型。
第二终端为计算需求端或计算协作端时，参考图5和图6，此时将第一终端的ID和可使用算力信息发送给第二终端的方法和前述算力分配清单的发送方法类似，此处不再赘述。计算需求端和计算协作端接收到第一终端的ID和可使用算力信息后，根据可使用算力信息进行任务分配以及信息整合。
在一个可能的实施例中，算力共享方法还包括：
获取并存储各第一终端为计算任务所提供的算力大小和所花费的时间。
具体地，算力共享网元获取第一终端为计算任务所提供的算力大小和所花费的时间，其中，第二终端可以从下发任务给第一终端时开始计时，接收到第一终端返回的子处理结果停止计时，若计算任务只需执行一次即可，则将得到的统计时间作为第一终端为计算任务所花费的时间，将第一终端为完成分配的任务所提供的计算量作为该第一终端为计算任务所提供的算力大小。若计算任务重复执行了X次，则需要将X个统计时间的时间之和作为该第一终端为计算任务所花费的时间，同理，将X个计算量之和作为该第一终端为计算任务所提供的算力大小。如图7中第5点，需要重复执行时，假设重复执行了3次，则需要累计3次统计的时间之和作为计算协作端-2为计算任务所花费的时间，3次所花费的计算量之和为计算协作端-2为计算任务所提供的总的算力大小。
基于上述算力共享方法实施例的描述，本申请实施例还公开了一种算力共享网元，参考图8，图8是本申请实施例提供的一种算力共享网元的结构示意图，算力共享网元与网络连接以进行控制面传输和数据面传输，或者，算力共享网元与网络连接以进行控制面传输；算力共享网元包括：
获取模块801，用于获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；
处理模块802，用于根据算力需求信息和可提供算力信息，确定为计算需求端提供算力的计算协作端，并直接或间接指示提供算力的计算协作端完成计算需求端的计算任务。
在一个可能的实施例中，网络包括核心网；
算力共享网元通过第一参考点与核心网的统一数据管理网元交互，算力共享网元通过第二参考点与核心网的接入和移动性管理网元交互，算力共享网元通过第三参考点与核心网的用户平面网元交互；
或者，
算力共享网元通过第一参考点与核心网的统一数据管理网元交互，算力共享网元通过第二参考点与核心网的接入和移动性管理网元交互。
在一个可能的实施例中，获取模块包括：
第一接收子模块，用于接收计算需求端发送的算力请求，算力请求包括计算需求端的ID、算力需求信息，计算需求端期望参与算力共享时，算力请求还包括计算需求端的可参与算力信息，计算需求端期望在算力共享时对计算任务进行任务分配和信息整合时，算力请求还包括计算需求端的整合能力信息；
第一发送子模块，用于向计算协作端发送协作询问，协作询问包括可提供算力上报请求；算力共享网元与网络连接以进行控制面传输时，根据整合能力信息确定计算需求端不具备信息整合能力时，协作询问还包括整合能力上报请求；
第一接收子模块，还用于接收计算协作端响应协作询问返回的协作询问响应信息，协作询问响应信息包括响应可提供算力上报请求时返回的计算协作端的ID和可提供算力信息，以及响应整合能力上报请求时返回的计算协作端的整合能力信息。
在另一个可能的实施例中，获取模块包括：
第二接收子模块，用于接收计算协作端发送的算力提供请求，算力提供请求包括计算协作端的ID、可提供算力信息，计算协作端期望在算力共享时对计算任务进行任务分配和信息整合时，算力提供请求还包括整合能力信息；
第二发送子模块，用于向计算需求端发送需求询问，需求询问包括算力需求上报请求，期望计算需求端参与算力共享时，需求询问还包括算力上报请求，期望计算需求端在算力共享时对计算任务进行任务分配和信息整合时，需求询问还包括整合能力上报请求；
第二接收子模块，还用于接收计算需求端响应需求询问返回的需求询问响应信息，需求询问响应信息包括响应算力需求上报请求时返回的计算需求端的ID和算力需求信息，响应算力上报请求时返回的计算需求端的可参与算力信息，以及响应整合能力上报请求时返回的计算需求端的整合能力信息。
在一个可能的实施例中，处理模块包括：
第一处理子模块，用于根据算力需求信息、算力共享网元的可参与算力信息、计算需求端的可参与算力信息和计算协作端的可提供算力信息确定是否为计算需求端提供算力；具体地，当可参与算力信息、可提供算力信息可满足由算力需求信息确定的算力需求时，为计算需求端提供算力并确定为计算需求端提供算力的第一终端，第一终端包括算力共享网元、计算需求端和计算协作端中至少一种。
在一个可能的实施例中，处理模块包括：
第二处理子模块，用于算力共享网元与网络连接以进行控制面传输和数据面传输时，根据计算需求端的整合能力信息，从算力共享网元、计算需求端中确定进行任务分配和信息整合的第二终端；或者，用于算力共享网元与网络连接以进行控制面传输时，根据计算需求端的整合能力信息和计算协作端的整合能力信息，从计算需求端和计算协作端中确定进行任务分配和信息整合的第二终端；
第三处理子模块，用于发送第一终端的ID和对应的可使用算力信息至第二终端，以使第二终端获取计算需求端发送的计算任务，并根据可使用算力信息将计算任务分配至第一终端，以及接收第一终端返回的子处理结果，整合子处理结果得到计算任务的总处理结果。
在一个可能的实施例中，第二终端为计算需求端或计算协作端时，处理模块包括：
第四处理子模块，用于根据算力需求信息和对应的第一终端的可使用算力信息进行算力调度，得到算力分配清单，算力分配清单为第一终端的ID和分配的算力的对应关系；
第五处理子模块，用于发送算力分配清单至第二终端，以使第二终端获取 计算需求端发送的计算任务，并根据算力分配清单将计算任务分配至第一终端，以及接收第一终端返回的子处理结果，整合子处理结果得到计算任务的总处理结果。
在一个可能的实施例中，算力共享网元还包括：
存储模块，用于获取并存储各第一终端为计算任务所提供的算力大小和所花费的时间。
在一个可能的实施例中，算力需求信息包括计算任务所需的计算量。算力需求信息还包括完成计算任务可接受的最长计算时间，最大任务并行度，计算任务的算法复杂度、指定采集数据类型中的至少一种。
值得指出的是，其中，算力共享网元的具体功能实现方式可以参见上述算力共享方法的描述，这里不再进行赘述。所述的算力共享网元中的各个单元或模块可以分别或全部合并为一个或若干个另外的单元或模块来构成，或者其中的某个(些)单元或模块还可以再拆分为功能上更小的多个单元或模块来构成，这可以实现同样的操作，而不影响本申请的实施例的技术效果的实现。上述单元或模块是基于逻辑功能划分的，在实际应用中，一个单元(或模块)的功能也可以由多个单元(或模块)来实现，或者多个单元(或模块)的功能由一个单元(或模块)实现。
基于上述方法实施例的描述，参考图9，图9是本申请实施例提供的一种算力共享网元的结构示意图。本申请实施例还提供一种算力共享网元，包括：收发器903、处理器902、存储器901存储在存储器901上并由处理器执行的计算机程序；处理器902分别与收发器903、存储器901相连，用于通过执行计算机程序，实现上述的算力共享方法。
收发器903可包括收发天线，或者，网卡等各种具有信息收发功能的物理结果。存储器901可包括各种存储介质的存储器件。处理器902可为中央处理器、微处理器、数字信号处理器、应用处理器、可编程阵列、专用集成电路等各种处理器件。处理器902可以通过集成电路总线等，与收发器903及存储器901连接，可以通过计算机程序执行，实现前述的算力共享方法。应当理解， 本申请实施例中所描述的算力共享网元可执行前文所述算力共享方法，在此不再赘述。另外，对采用相同方法的有益效果描述，也不再进行赘述。
基于上述方法实施例的描述，参考图10，图10是本申请实施例提供的一种算力共享系统的结构示意图，一种算力共享系统，包括计算需求端101、计算协作端103和前述的算力共享网元102。
此外，这里需要指出的是：本申请实施例还提供了一种计算机存储介质，且所述计算机存储介质中存储有前文提及的算力共享网元所执行的计算机程序，且所述计算机程序包括程序指令，当所述处理器执行所述程序指令时，能够执行前文所述算力共享方法的描述，因此，这里将不再进行赘述。另外，对采用相同方法的有益效果描述，也不再进行赘述。对于本申请所涉及的计算机存储介质实施例中未披露的技术细节，请参照本申请方法实施例的描述。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的程序可存储于一计算机可读取存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory，ROM)或随机存储记忆体(Random Access Memory，RAM)等。
以上所揭露的仅为本申请较佳实施例而已，当然不能以此来限定本申请之权利范围，因此依本申请权利要求所作的等同变化，仍属本申请所涵盖的范围。

Claims (15)

  1. 一种算力共享方法，其特征在于，应用于算力共享网元中，所述算力共享网元与网络连接以进行控制面传输和数据面传输，或者，所述算力共享网元与网络连接以进行控制面传输；
    所述方法包括：
    获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；
    根据所述算力需求信息和所述可提供算力信息，确定为所述计算需求端提供算力的计算协作端，并直接或间接指示所述提供算力的计算协作端完成所述计算需求端的计算任务。
  2. 根据权利要求1所述的方法，其特征在于，所述获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息，包括：
    接收所述计算需求端发送的算力请求，所述算力请求包括所述计算需求端的ID、算力需求信息，所述计算需求端期望参与算力共享时，所述算力请求还包括所述计算需求端的可参与算力信息，所述计算需求端期望在算力共享时对所述计算任务进行任务分配和信息整合时，所述算力请求还包括所述计算需求端的整合能力信息；
    向所述计算协作端发送协作询问，所述协作询问包括可提供算力上报请求；所述算力共享网元与网络连接以进行控制面传输时，根据所述整合能力信息确定所述计算需求端不具备信息整合能力或所述计算需求端未发送所述整合能力信息时，所述协作询问还包括整合能力上报请求；
    接收所述计算协作端响应所述协作询问返回的协作询问响应信息，所述协作询问响应信息包括响应所述可提供算力上报请求时返回的所述计算协作端的ID和可提供算力信息，以及响应所述整合能力上报请求时返回的所述计算协作端的整合能力信息。
  3. 根据权利要求1所述的方法，其特征在于，所述获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息，包括：
    接收所述计算协作端发送的算力提供请求，所述算力提供请求包括所述计 算协作端的ID、可提供算力信息，所述计算协作端期望在算力共享时对所述计算任务进行任务分配和信息整合时，所述算力提供请求还包括整合能力信息；
    向所述计算需求端发送需求询问，所述需求询问包括算力需求上报请求，期望所述计算需求端参与算力共享时，所述需求询问还包括算力上报请求，期望所述计算需求端在算力共享时对所述计算任务进行任务分配和信息整合时，所述需求询问还包括整合能力上报请求；
    接收所述计算需求端响应所述需求询问返回的需求询问响应信息，所述需求询问响应信息包括响应所述算力需求上报请求时返回的所述计算需求端的ID和算力需求信息，响应所述算力上报请求时返回的所述计算需求端的可参与算力信息，以及响应所述整合能力上报请求时返回的所述计算需求端的整合能力信息。
  4. 根据权利要求2或3所述的方法，其特征在于，所述根据所述算力需求信息和所述可提供算力信息，确定为所述计算需求端提供算力的计算协作端，包括：
    根据所述算力需求信息、所述算力共享网元的可参与算力信息、所述计算需求端的可参与算力信息和所述计算协作端的可提供算力信息确定是否为所述计算需求端提供算力；
    当所述可参与算力信息、所述可提供算力信息可满足由所述算力需求信息确定的算力需求时，为所述计算需求端提供算力并确定为所述计算需求端提供算力的第一终端，所述第一终端包括所述算力共享网元、所述计算需求端和所述计算协作端中至少一种。
  5. 根据权利要求4所述的方法，其特征在于，所述直接或间接指示所述提供算力的计算协作端完成所述计算需求端的计算任务，包括：
    所述算力共享网元与网络连接以进行控制面传输和数据面传输时，根据所述计算需求端的整合能力信息和计算协作端的整合能力信息，从所述算力共享网元、所述计算需求端和所述计算协作端中确定进行任务分配和信息整合的第二终端；或者，所述算力共享网元与网络连接以进行控制面传输时，根据所述计算需求端的整合能力信息和所述计算协作端的整合能力信息，从所述计算需 求端和所述计算协作端中确定进行任务分配和信息整合的第二终端；
    发送所述第一终端的ID和对应的可使用算力信息至所述第二终端，以使所述第二终端获取所述计算需求端发送的计算任务，并根据所述可使用算力信息将所述计算任务分配至所述第一终端，以及接收所述第一终端返回的子处理结果，整合所述子处理结果得到所述计算任务的总处理结果。
  6. 根据权利要求5所述的方法，其特征在于，确定进行任务分配和信息整合的第二终端后，所述第二终端为计算需求端或计算协作端时，所述方法还包括：
    根据所述算力需求信息和对应的所述第一终端的可使用算力信息进行算力调度，得到算力分配清单，所述算力分配清单为所述第一终端的ID和分配的算力的对应关系；
    发送所述算力分配清单至所述第二终端，以使所述第二终端获取所述计算需求端发送的计算任务，并根据所述算力分配清单将所述计算任务分配至所述第一终端，以及接收所述第一终端返回的子处理结果，整合所述子处理结果得到所述计算任务的总处理结果。
  7. 根据权利要求5或6所述的方法，其特征在于，所述方法还包括：
    获取并存储各所述第一终端为所述计算任务所提供的算力大小和所花费的时间。
  8. 根据权利要求1、2、3、5或6所述的方法，其特征在于，所述网络包括核心网；
    所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互，所述算力共享网元通过第三参考点与所述核心网的用户平面网元交互；
    或者，
    所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互。
  9. 根据权利要求1、2、3、5或6所述的方法，其特征在于，所述算力需 求信息包括所述计算任务所需的计算量。
  10. 根据权利要求9所述的方法，其特征在于，所述算力需求信息还包括完成所述计算任务可接受的最长计算时间，最大任务并行度，计算任务的算法复杂度、指定采集数据类型中的至少一种。
  11. 一种算力共享网元，其特征在于，所述算力共享网元与网络连接以进行控制面传输和数据面传输，或者，所述算力共享网元与网络连接以进行控制面传输；
    所述算力共享网元包括：
    获取模块，用于获取计算需求端发送的算力需求信息以及计算协作端发送的可提供算力信息；
    处理模块，用于根据所述算力需求信息和所述可提供算力信息，确定为所述计算需求端提供算力的计算协作端，并直接或间接指示所述提供算力的计算协作端完成所述计算需求端的计算任务。
  12. 根据权利要求11所述的算力共享网元，其特征在于，所述网络包括核心网；
    所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互，所述算力共享网元通过第三参考点与所述核心网的用户平面网元交互；
    或者，
    所述算力共享网元通过第一参考点与所述核心网的统一数据管理网元交互，所述算力共享网元通过第二参考点与所述核心网的接入和移动性管理网元交互。
  13. 一种算力共享网元，其特征在于，包括：收发器、处理器、存储器以及存储在存储器上并由所述处理器执行的计算机程序；
    所述处理器分别与所述收发器、所述存储器相连，用于通过执行所述计算机程序，实现权利要求1-10任一项所述的算力共享方法。
  14. 一种算力共享系统，其特征在于，包括计算需求端、计算协作端和权利要求11、12或13所述的算力共享网元。
  15. 一种计算机存储介质，其特征在于，所述计算机存储介质存储有计算机程序，所述计算机程序包括程序指令，所述程序指令当被处理器执行时，执行如权利要求1-10任一项所述的算力共享方法。
PCT/CN2021/086931 2020-04-13 2021-04-13 算力共享方法及相关设备 WO2021208915A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
JP2022562579A JP2023521222A (ja) 2020-04-13 2021-04-13 計算力共有方法及び関連デバイス
KR1020227036077A KR20230035516A (ko) 2020-04-13 2021-04-13 컴퓨팅 파워 공유 방법 및 관련 장치
US17/918,464 US11876856B2 (en) 2020-04-13 2021-04-13 Method for computational-power sharing and related device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010288252.4A CN113542316B (zh) 2020-04-13 2020-04-13 算力共享方法及相关设备
CN202010288252.4 2020-04-13

Publications (1)

Publication Number Publication Date
WO2021208915A1 true WO2021208915A1 (zh) 2021-10-21

Family

ID=78084679

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/086931 WO2021208915A1 (zh) 2020-04-13 2021-04-13 算力共享方法及相关设备

Country Status (5)

Country Link
US (1) US11876856B2 (zh)
JP (1) JP2023521222A (zh)
KR (1) KR20230035516A (zh)
CN (1) CN113542316B (zh)
WO (1) WO2021208915A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114265703A (zh) * 2022-03-02 2022-04-01 梯度云科技(北京)有限公司 云服务器跨区域算力调度方法、系统及设备
WO2023155655A1 (zh) * 2022-02-21 2023-08-24 华为技术有限公司 算力能力感知方法和装置
WO2023178567A1 (zh) * 2022-03-23 2023-09-28 北京小米移动软件有限公司 一种上报方法/装置/设备及存储介质
WO2024016143A1 (zh) * 2022-07-19 2024-01-25 华为技术有限公司 一种计算协作方法及相关装置

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117097618A (zh) * 2022-05-12 2023-11-21 华为技术有限公司 信息处理的方法和装置
CN117480764A (zh) * 2022-05-27 2024-01-30 北京小米移动软件有限公司 一种多设备协同的控制方法、装置及可读存储介质
CN117278409A (zh) * 2022-06-14 2023-12-22 大唐移动通信设备有限公司 算网融合方法和相关设备
CN117320081A (zh) * 2022-06-21 2023-12-29 维沃移动通信有限公司 算力处理方法、装置及通信设备
CN117499995A (zh) * 2022-07-21 2024-02-02 中兴通讯股份有限公司 基站算力编排方法、装置、电子设备及存储介质
CN117998518A (zh) * 2022-11-07 2024-05-07 中国移动通信有限公司研究院 一种通信方法、装置、相关设备和存储介质
CN116074541B (zh) * 2023-03-09 2023-06-02 新华三工业互联网有限公司 一种资源处理方法、系统、装置及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108874526A (zh) * 2018-07-05 2018-11-23 天使翼(武汉)科技创业发展有限公司 一种基于网络通讯的算力共享平台及方法
CN108985937A (zh) * 2018-07-10 2018-12-11 陈曦 一种基于区块链技术的计算资源共享方法及区块链系统
CN109242637A (zh) * 2018-09-26 2019-01-18 亿邦客科技有限公司 一种算力共享智能系统
US20200076884A1 (en) * 2018-08-31 2020-03-05 Nakamoto & Turing Labs Inc Methods and apparatus for performing distributed computing using blockchain
CN110928685A (zh) * 2019-11-20 2020-03-27 圆通速递有限公司 一种智能协同计算系统和方法

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6463457B1 (en) * 1999-08-26 2002-10-08 Parabon Computation, Inc. System and method for the establishment and the utilization of networked idle computational processing power
CN103733194A (zh) * 2011-01-27 2014-04-16 康姆普特奈科斯特有限公司 动态组织云计算资源以便于发现
US11327811B2 (en) * 2017-07-06 2022-05-10 Centurylink Intellectual Property Llc Distributed computing mesh
CN109818868B (zh) * 2017-11-20 2021-06-22 中兴通讯股份有限公司 一种实现边缘网络能力开放的方法、装置、设备及存储介质
US10742728B2 (en) * 2018-06-07 2020-08-11 At&T Intellectual Property I, L.P. Edge sharing orchestration system
CN110782343B (zh) * 2018-07-30 2022-12-16 中移(苏州)软件技术有限公司 分布式网络中基于区块链的算力流通方法及系统
WO2020133450A1 (zh) * 2018-12-29 2020-07-02 驭势科技（北京）有限公司 一种移动设备动态组网分享算力的系统与方法
CN109960575B (zh) * 2019-03-26 2023-09-15 深圳市网心科技有限公司 一种计算能力共享方法、系统及相关设备
CN110225129A (zh) * 2019-06-18 2019-09-10 余俊龙 基于区块链应用扩展控制方法及智能终端、私有云服务器
CN110231987A (zh) * 2019-06-21 2019-09-13 深圳市网心科技有限公司 一种数据处理方法及相关装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108874526A (zh) * 2018-07-05 2018-11-23 天使翼(武汉)科技创业发展有限公司 一种基于网络通讯的算力共享平台及方法
CN108985937A (zh) * 2018-07-10 2018-12-11 陈曦 一种基于区块链技术的计算资源共享方法及区块链系统
US20200076884A1 (en) * 2018-08-31 2020-03-05 Nakamoto & Turing Labs Inc Methods and apparatus for performing distributed computing using blockchain
CN109242637A (zh) * 2018-09-26 2019-01-18 亿邦客科技有限公司 一种算力共享智能系统
CN110928685A (zh) * 2019-11-20 2020-03-27 圆通速递有限公司 一种智能协同计算系统和方法

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023155655A1 (zh) * 2022-02-21 2023-08-24 华为技术有限公司 算力能力感知方法和装置
CN114265703A (zh) * 2022-03-02 2022-04-01 梯度云科技(北京)有限公司 云服务器跨区域算力调度方法、系统及设备
CN114265703B (zh) * 2022-03-02 2022-05-20 梯度云科技(北京)有限公司 云服务器跨区域算力调度方法、系统及设备
WO2023178567A1 (zh) * 2022-03-23 2023-09-28 北京小米移动软件有限公司 一种上报方法/装置/设备及存储介质
WO2024016143A1 (zh) * 2022-07-19 2024-01-25 华为技术有限公司 一种计算协作方法及相关装置

Also Published As

Publication number Publication date
US11876856B2 (en) 2024-01-16
CN113542316A (zh) 2021-10-22
CN113542316B (zh) 2022-11-25
US20230144062A1 (en) 2023-05-11
KR20230035516A (ko) 2023-03-14
JP2023521222A (ja) 2023-05-23

Similar Documents

Publication Publication Date Title
WO2021208915A1 (zh) 算力共享方法及相关设备
CN108540568B (zh) 计算能力共享方法及智能设备
JP2023536711A (ja) 計算力共有方法及び関連デバイス
CN108650667A (zh) 终端调度方法和装置
CN110166524A (zh) 数据中心的切换方法、装置、设备及存储介质
TW202121274A (zh) 雲端資源管理方法、裝置、電子設備及電腦可讀儲存媒體
US20200358715A1 (en) Resource sharing method, computer device, and storage medium
TW202131171A (zh) 程式設計平台的使用者代碼運行方法、電子設備和電腦可讀儲存介質
Chen et al. DDPG-based computation offloading and service caching in mobile edge computing
CN108111337A (zh) 分布式系统仲裁主节点的方法及设备
CN114595167A (zh) 分布式缓存系统、方法以及装置
CN109348486A (zh) 一种异构无线网络资源分配方法
WO2021027842A1 (zh) 一种实现边缘计算的方法、装置和系统
CN105593814A (zh) 归属管理设备虚拟化方法、装置及系统
Krishna Priya et al. Crossover-based improved sine cosine algorithm for multimedia content distribution in cloud environment
Liu et al. Distributed Computation Offloading with Low Latency for Artificial Intelligence in Vehicular Networking
CN106453596A (zh) 一种分享图片的方法、装置和系统
CN112561301A (zh) 工单分配方法、装置、设备和计算机可读介质
CN109587641B (zh) 移动智能设备中基于用户匹配的数据流量共享方法
CN115065727B (zh) 一种基于边缘计算场景的任务卸载方法
CN115695187B (zh) 通讯资源获取方法、装置、计算机设备和存储介质
CN105992360B (zh) 一种资源分配方法及装置
WO2024036456A1 (zh) 一种基于人工智能ai提供服务的方法、装置、设备及存储介质
CN108762929B (zh) 在sql数据库下管理处理器核心数的方法和装置
WO2023243102A1 (ja) 無線通信システム、無線通信方法、及び基地局装置

Legal Events

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

Ref document number: 21787734

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2022562579

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21787734

Country of ref document: EP

Kind code of ref document: A1