Node selection method, communication device and storage medium

Abstract

The invention provides a node selection method, communication equipment and a storage medium. The node selection method applied to a first network element node comprises the following steps: receiving a task operation request sent by a second network element node; performing node selection according to a node selection criterion; sending a task operation report to the second network element node; wherein the task operation report comprises node selection information.

Description

Technical Field

[0001] This application relates to the field of communication technology, specifically to a node selection method, communication equipment, and storage medium. Background Technology

[0002] In complex network environments, such as edge computing networks or mobile networks, nodes often have limited computing resources, storage capacity, and bandwidth. Existing node selection methods may fail to efficiently utilize these limited resources, leading to inefficient inference or slow training processes. Simultaneously, model training and inference typically consume significant amounts of electricity, increasing operating costs and burdening the environment. Therefore, prioritizing the use of nodes powered by green energy to reduce the environmental impact of Artificial Intelligence (AI) technology is crucial when selecting training or inference nodes. Furthermore, certain application scenarios (such as edge intelligence applications) have extremely high real-time requirements, while traditional methods often cannot meet the low-latency inference demands. Summary of the Invention

[0003] In view of this, embodiments of this application provide a node selection method, a communication device, and a storage medium, which improve the overall efficiency and performance of task operations in the network.

[0004] This application provides a node selection method applied to a first network element node, including:

[0005] Receive task operation requests sent by the second network element node;

[0006] Nodes are selected according to node selection criteria;

[0007] Send a task operation report to the second network element node; wherein the task operation report includes node selection information.

[0008] This application provides a node selection method applied to a second network element node, including:

[0009] Send a task operation request to the first network element node;

[0010] The system receives a task operation report sent by the first network element node; wherein the task operation report includes node selection information; the node selection information is obtained by the first network element node through node selection according to node selection criteria.

[0011] This application provides a node selection device applied to a first network element node, comprising:

[0012] The receiving module is configured to receive task operation requests sent by the second network element node;

[0013] The selection module is configured to select nodes based on node selection criteria.

[0014] The sending module is configured to send a task operation report to the second network element node; wherein the task operation report includes node selection information.

[0015] This application provides a node selection device applied to a second network element node, comprising:

[0016] The sending module is configured to send task operation requests to the first network element node;

[0017] The receiving module is configured to receive a task operation report sent by the first network element node; wherein the task operation report includes node selection information; the node selection information is obtained by the first network element node through node selection according to node selection criteria.

[0018] This application provides a communication device, including: a memory, and one or more processors;

[0019] The memory is configured to store one or more programs;

[0020] When the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any of the above embodiments.

[0021] This application provides a storage medium storing a computer program, which, when executed by a processor, implements the methods described in any of the above embodiments. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of a network management architecture provided by existing technology;

[0023] Figure 2 This is a flowchart of a node selection method provided in an embodiment of this application;

[0024] Figure 3 This is a flowchart of another node selection method provided in an embodiment of this application;

[0025] Figure 4 This is a schematic diagram illustrating an interactive implementation of node selection provided in an embodiment of this application;

[0026] Figure 5 This is a schematic diagram illustrating another interactive implementation of node selection provided in an embodiment of this application;

[0027] Figure 6 This is a schematic diagram illustrating another interactive implementation of node selection provided in an embodiment of this application;

[0028] Figure 7 This is a schematic diagram illustrating another interactive implementation of node selection provided in an embodiment of this application;

[0029] Figure 8 This is a schematic diagram illustrating another interactive implementation of node selection provided in an embodiment of this application;

[0030] Figure 9 This is a schematic diagram illustrating another interactive implementation of node selection provided in an embodiment of this application;

[0031] Figure 10 This is a structural block diagram of a node selection device provided in an embodiment of this application;

[0032] Figure 11 This is a structural block diagram of another node selection device provided in the embodiments of this application;

[0033] Figure 12 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application. Detailed Implementation

[0034] The embodiments of this application will be described below with reference to the accompanying drawings. The examples given are for illustrative purposes only and are not intended to limit the scope of this application.

[0035] Figure 1 This is a schematic diagram of a network management architecture provided by existing technology, such as... Figure 1 As shown, the network management architecture may include a Business Support System (BSS), a Cross-Domain Management Function (CD-MnF), a Domain Management Function (Domain-MnF), and Network Elements (NEs). The CD-MnF manages one or more Domain Management Functions. Each Domain Management Function can manage one or more Network Elements.

[0036] A business support system (BSS) is oriented towards communication services and provides functions and management services such as billing, settlement, accounting, customer service, sales, network monitoring, communication service lifecycle management, and service intent translation. The BSS can be an operator's operating system or a vertical industry operating system (vertical OTSystem).

[0037] The Cross-Domain Management Function (NMF), also known as the Network Management Function (NMF), can be a network management entity such as a Network Management System (NMS), a Network Management Service Producer (MnS Producer), a Network Management Service Consumer (MnS Consumer), and a Network Function Management Service Consumer (NFMS_C). The NMF provides one or more of the following management functions or services: network lifecycle management, network deployment, network fault management, network performance management, network configuration management, network assurance, network optimization, and translation of network intents from Communication Service Providers (Intent-CSPs). The network referred to in the above management functions or services can include one or more network elements or subnetworks, or it can be a network slice. In other words, a network management function unit can be a Network Slice Management Function (NSMF), a Management Data Analytical Function (MDAF), a Self-Organization Network Function (SON), an Intent-Driven Management Service (Intent Driven MnS), or a Close Control Loop Management Service.

[0038] The domain management function unit, also known as the network subnet management function unit (NSMF) or network element management function unit, can be a wireless automation engine (MBB automationengine, MAE), a network element management system (EMS), a network function management service provider (NFMS_P), a network slice subnet management function (NSSMF), a domain management data analysis function (Domain MDAF), a self-organization network function (SON Function), a domain intent management function unit, or a close control loop management service, an MnS producer, an MnS consumer, or other network element management entities. Domain management function units can be classified in the following ways: By network type, they can be divided into: Radio Access Network (RAN) Domain Management Function (RAN domain MnF), Core Network Domain Management Function (CN domain MnF), and Transport Network Domain Management Function (TN domain MnF), etc. It should be noted that a domain management function unit can also be a domain network management system, managing one or more of the access network, core network, or transport network. By administrative region, they can be divided into: domain management function units for a specific region, such as the domain management function unit for city A, the domain management function unit for city B, etc.The domain management functional unit provides one or more of the following functions or management services: lifecycle management of subnetworks or network elements, deployment of subnetworks or network elements, fault management of subnetworks or network elements, performance management of subnetworks or network elements, assurance of subnetworks or network elements, optimization functions of subnetworks or network elements, and translation of subnetwork or network element intents (Intent from Network Operator, Intent-NOP), etc. Here, a subnetwork includes one or more network elements. A subnetwork can also include subnetworks, that is, one or more subnetworks forming a larger subnetwork. Here, a subnetwork can also be a network slice subnetwork.

[0039] A network element is an entity that provides network services, including core network elements, radio access network elements, or transport network elements. Specifically, core network elements may include, but are not limited to, Access and Mobility Management Function (AMF) entities, Session Management Function (SMF) entities, Policy Control Function (PCF) entities, Network Data Analysis Function (NWDAF) entities, Network Repository Function (NRF) entities, and gateways. Wireless access network elements may include, but are not limited to: various base stations (e.g., Generation NodeB (gNB), Evolved NodeB (eNB), Central Unit Control Panel (CUCP), Central Unit (CU), Distributed Unit (DU), and Central Unit User Panel (CUUP), etc.). In this application, network function (NF) is also referred to as network element (NE). A network element can provide one or more of the following management functions or services: network element lifecycle management, deployment, fault management, performance management, assurance, optimization functions, and translation of network element intent, etc.

[0040] It should be noted that in this application, any consumer / producer can be a UE or Figure 1Any entity in the list can be a UE, gNB, NF (such as NWDAF), Operation Administration and Management (OAM) (such as MDAF), cross-domain OAM, or Operations Support System (OSS) / BSS.

[0041] In one embodiment, Figure 2 This is a flowchart illustrating a node selection method provided in an embodiment of this application. This embodiment is applied to the selection of various operating nodes in a network. This embodiment can be executed by a first network element node. Figure 2 As shown, this embodiment includes: S210-S230.

[0042] S210, Receive the task operation request sent by the second network element node.

[0043] In one example, a task operation request refers to a request message that performs different types of task operations. For example, task operations may include training, inference, and computation. Correspondingly, a task operation request may include at least one of the following: a training request, an inference request, and a computation request. In one example, a training request may include at least one of the following: a distributed learning training request and a federated learning training request.

[0044] S220. Select nodes according to the node selection criteria.

[0045] In one example, the process of the first network element node selecting a node based on the node selection criteria may include: the first network element node selecting a candidate node based on the node selection criteria, or selecting a confirmed node based on the node selection criteria.

[0046] S230. Send a task operation report to the second network element node; wherein, the task operation report includes node selection information.

[0047] In one example, when the first network element node selects candidate nodes based on node selection criteria, it receives candidate node information and sends a task operation report containing this information to the second network element node. In another example, when the first network element node selects a confirmed node based on the same criteria, it receives the confirmed node information and sends a task operation report containing this information to the second network element node. By comprehensively analyzing node selection criteria that include factors such as resource management, real-time performance, performance differences, and dynamic adaptability, and selecting nodes based on these criteria, the first network element node ensures that the selected nodes meet network requirements. This improves the overall efficiency and performance of various operations within the network and promotes the sustainable development of artificial intelligence technology.

[0048] In one embodiment, the node selection criteria are obtained through one of the following methods: sent by the second network element node; pre-configured; or proprietary. In one example, the first network element node can select nodes based on the node selection criteria sent by the second network element, or it can use pre-configured node selection criteria, or it can use its own proprietary node selection criteria. In one example, the first network element node can select candidate nodes based on the node selection criteria to form candidate node information. In this case, the first network element node can send a task operation report containing the candidate node information to the second network element node, so that the second network element node can select a confirmation node from the candidate node information based on the node selection criteria, thus obtaining confirmation node information. In one example, the first network element node can select a confirmation node based on the node selection criteria to form confirmation node information. In this case, the first network element node can send a task operation report containing the confirmation node information to the second network element node.

[0049] In one embodiment, the node selection information includes at least one of the following: candidate node information; confirmed node information; and node change information. In one example, the candidate node information can be multiple node sets; the confirmed node information can be a single node set; each node set contains identifiers for one or more nodes. In one example, the node change information refers to the changes in each node during task execution. In one example, the candidate node information can also be understood as information about multiple available nodes.

[0050] In one embodiment, the node selection information is candidate node information; the node selection method applied to the first network element node further includes: receiving confirmation node information fed back by the second network element node. In one example, when the node selection information contained in the task operation report is candidate node information, the second network element node receives the task operation report containing candidate node information sent by the first network element node, and then the second network element node selects one or more nodes from the candidate node information based on the node selection criteria as confirmation node information, and sends all the nodes contained in the confirmation node information to the first network element node; then, after the first network element node receives the confirmation node information fed back by the second network element node, the first network element node uses one or more nodes contained in the confirmation node information to perform the current task operation. In one example, when the second network element node selects a confirmation node from the candidate node information based on node selection criteria, the second network element node can select a confirmation node from the candidate node information based on its own private node selection criteria, or it can select a confirmation node from the candidate node information based on pre-configured node selection criteria, obtain confirmation node information, and send all nodes contained in the confirmation node information to the first network element node; then, after the first network element node receives the confirmation node information fed back by the second network element node, the first network element node uses one or more nodes contained in the confirmation node information to perform the task operation.

[0051] In one embodiment, the node selection method applied to the first network element node further includes: performing a task operation based on the node selected by the first network element node or the node confirmed by the second network element node; correspondingly, the task operation report also includes the task execution result. In one example, after the first network element node obtains the confirmed node information based on the node selection criteria, the first network element node can perform a task operation based on one or more nodes in the confirmed node information determined by the second network element node, obtain the task execution result, and send the task execution result in the task operation report to the second network element node. In another example, after the first network element node obtains the confirmed node information based on the node selection criteria sent by the second network element node, either pre-configured or its own private criteria, the first network element node can perform a task operation based on one or more nodes in the confirmed node information, obtain the task execution result, and send the task execution result in the task operation report to the second network element node.

[0052] In one embodiment, the task operation request includes at least one of the following: federated learning training request; distributed learning training request; inference request; computation request.

[0053] In one embodiment, the federated learning training request, distributed learning training request, inference request, and computation request include node selection criteria. In one example, a second network element node can send a task operation request containing node selection criteria to a first network element node, allowing the first network element node to directly use the node selection criteria sent by the second network element node for node selection. In another example, when the second network element node sends a task operation request containing node selection criteria to the first network element node, the first network element node can also select nodes based on pre-configured or its own proprietary node selection criteria.

[0054] In one embodiment, the node selection criteria include at least one of the following: node green energy information; node infrastructure information; node selection strategy.

[0055] In one embodiment, the node green energy information includes at least one of the following: renewable energy information; carbon emission information; and green energy ratio information. In one example, the renewable energy information is used at the single-node granularity to indicate the requirements a single node needs to meet in terms of renewable energy indicators. For example, the renewable energy information may include at least one of the following: a renewable energy indicator (indicating that the node needs to use renewable energy) and a renewable energy category (e.g., one or more of wind, hydro, and solar energy). In one example, the carbon emission information is used at the single-node granularity to indicate the requirements a single node needs to meet in terms of carbon emission indicators. For example, the carbon emission information may include at least one of the following: maximum carbon emissions (e.g., energy consumption (kWh) relative to kilograms of carbon dioxide). In one example, the green energy ratio information is used at the granularity of the entire task or multiple nodes within the entire task. For example, the green energy ratio information may be characterized by indicators such as the percentage of green energy consumed in the entire task, the proportion of nodes using green energy or renewable energy, or the minimum number of nodes using green energy or renewable energy.

[0056] In one embodiment, the node infrastructure information includes at least one of the following: hardware information; computing performance indicators; memory and storage performance indicators; parallel processing capability indicators; energy efficiency ratio indicators; and comprehensive computing power indicators.

[0057] In one example, hardware information may include, but is not limited to, one of the following: hardware categories such as graphics processing units (GPUs), central processing units (CPUs), and data processing units (DPUs), and their corresponding models, manufacturers, and performance parameters.

[0058] In one example, computational performance metrics may include at least one of the following: Floating Point Operations Per Second (FLOPS), Million Instructions Per Second (MIPS), and Tera Operations Per Second (TOPS). In one example, FLOPS is an important metric for measuring computer performance, especially in scientific computing, deep learning, and high-performance computing. It is defined as the number of floating-point operations a computer can perform per second; the higher the number of operations, the stronger the computing power of the computer. MIPS is a metric used to measure the execution speed of a CPU, reflecting the number of millions of instructions a CPU can process per second. It is suitable for evaluating the performance of a CPU when executing a basic instruction set and is often used in embedded systems and other fields. TOPS refers to the number of trillions of operations that can be performed per second. It is particularly suitable for deep learning tasks in the field of artificial intelligence and focuses on processing "integer" operations, such as 8-bit and 16-bit integer operations.

[0059] In one example, memory and storage performance metrics include at least one of the following: memory bandwidth and storage performance; in one example, memory bandwidth refers to the amount of data that can be transferred in memory per unit time, usually measured in data transfer per second (e.g., GB / s). High memory bandwidth can reduce data transfer latency and improve system efficiency; storage performance refers to the read and write speed of storage devices, including sequential read and write speed and random read and write speed. High-performance storage systems can quickly read and write large amounts of data and reduce data access latency.

[0060] In one example, parallel processing capability metrics may include the number of cores and the number of threads. In this example, the number of cores refers to the number of independent processing units in the processor; the number of threads refers to the number of tasks the processor can handle simultaneously. Multi-core processors and multi-threading technologies can improve a system's parallel processing capability, enabling multiple tasks to run concurrently.

[0061] In one example, the energy efficiency ratio metric could include: Energy Efficiency Ratio (EER). In one example, EER refers to the amount of computation that can be performed per unit of energy consumption, typically measured in the number of computations per watt (e.g., FLOPS / W, TOPS / W). A high EER allows for more computational tasks to be performed with the same amount of energy.

[0062] In one example, the comprehensive computing power indicator may include: computing power index. In one example, the computing power index is a comprehensive indicator that comprehensively assesses the scale, environment, application status, and development trend of computing power in a region or country. It covers multiple factors such as hardware, software, network, and data, and can use numbers or letters to represent levels, or use a score to characterize it.

[0063] In one embodiment, the node selection strategy includes at least one of the following: a minimum time strategy; an optimal performance strategy; and a minimum energy consumption strategy. In one example, the minimum time strategy may include at least one of the following: a minimum inference latency strategy, a minimum training time strategy, and a minimum computation latency strategy. In one example, different minimum time strategies can be applied to different task operation requests. For example, assuming the task operation request is a federated learning training request or a distributed learning training request, the corresponding node selection strategy may include at least the minimum training time strategy. As another example, assuming the task operation request is an inference request, the corresponding node selection strategy may include at least the minimum inference latency strategy. Yet another example, assuming the task operation request is a computation request, the corresponding node selection strategy may include at least the minimum computation latency strategy. In one example, the minimum energy consumption strategy may include using as many green nodes powered by renewable energy sources as possible. In one example, the optimal performance strategy may be set using optimal performance training indication information or by specific model performance metric numerical constraints. The minimum inference latency strategy can be set using minimum inference latency indication information or specific latency value constraints, or it can indicate the selection of the node with the best inference speed, computational efficiency, and performance; the minimum computation latency strategy can be set using minimum computation latency indication information or specific computational value constraints, or it can indicate the selection of the node with the best computational speed, computational efficiency, and performance; the shortest training time strategy can be set using shortest training time indication information or specific training time value constraints, or it can indicate the selection of the node with the best training speed, training efficiency, and performance.

[0064] In one embodiment, the task operation request includes a federated learning training request, and the node selection criterion is the server selection criterion; receiving the task operation request sent by the second network element node includes:

[0065] Receive the federated learning training request sent by the second network element node, which carries the server selection criteria;

[0066] Accordingly, node selection is performed based on node selection criteria, including: determining the federated learning server based on server selection criteria;

[0067] Sending a task operation report to the second network element node includes sending a federated learning training report to the second network element node; wherein the federated learning training report carries the identifier of the federated learning server. In one example, the federated learning training report may contain the identifier of one federated learning server, or it may contain the identifiers of multiple federated learning servers.

[0068] In one embodiment, determining the federated learning server based on server selection criteria includes:

[0069] If the first network element node meets the server selection criteria, then the first network element node selects itself as the federated learning server and notifies the second network element node.

[0070] If the first network element node does not meet the server selection criteria, the first network element node queries the registration and discovery service producer to obtain the set of federated learning producers, queries the federated learning capabilities supported by the available federated learning producers in the federated learning producer set, selects a federated learning server according to the server selection criteria, and notifies the second network element node.

[0071] In one embodiment, the task operation request includes a federated learning training request; the node selection method applied to the first network element node further includes:

[0072] Receive a query request from the second network element node for information on the federated learning capabilities supported by available federated learning producers in the federated learning producer set; wherein, the federated learning producer set is obtained by the second network element node from the registration and discovery service producer;

[0073] The first network element node sends information about the federated learning capabilities supported by available federated learning producers in the federated learning producer set to the second network element node. This allows the second network element node to select one of the available federated learning producers as the federated learning server. In one example, the federated learning producer set can also be called the federated learning producer list, which is a list of machine learning training functions that support federated learning. In one example, the second network element node can determine whether a federated learning producer in the federated learning producer set is available based on the status information (e.g., status information identifier) ​​of each federated learning producer in the set. For example, assuming the status information identifier is "available," it indicates that the federated learning producer is available and can be added to the subsequent federated learning process. Then, the second network element node sends a request to the first network element node to query the information about the federated learning capabilities supported by the available federated learning producers in the federated learning producer set. This allows the second network element node to obtain the information about the federated learning capabilities supported by each available federated learning producer in the set and select one of them as the federated learning server.

[0074] In one embodiment, the federated learning capability information includes at least one of the following: federated learning support capability; federated learning client support capability; federated learning server support capability; supported samples and features; and data information of the second network element node. In one example, the data information of the second network element node may include at least one of the following: the data volume of the second network element node, the sample volume of the second network element node, and the feature volume of the second network element node, etc.

[0075] In one embodiment, the task operation request is a federated learning training request, and the node selection criterion is a client selection criterion; receiving the task operation request sent by the second network element node includes:

[0076] Receive the federated learning training request sent by the second network element node, which carries the client selection criteria;

[0077] Accordingly, node selection is performed according to node selection criteria, including: selecting federated learning clients based on client selection criteria; or, selecting available federated learning clients based on client selection criteria and informing the second network element node so that the second network element node can confirm the federated learning client.

[0078] Send task operation reports to the second network element node, including: sending federated learning training reports to the second network element node.

[0079] In one embodiment, the federated learning training report includes at least one of the following: candidate federated learning clients; green energy usage in federated learning; and changes in federated learning clients. In one example, green energy usage in federated learning may include: green energy information for each federated learning client, and green energy usage throughout the entire federated learning process. In one example, green energy information for each federated learning client may include: whether green energy is used, and the type of green energy used; green energy usage throughout the entire federated learning process may include: the proportion of nodes using green energy. In one example, changes in federated learning clients are used to characterize the addition or removal of federated learning clients during the federated learning training process, and the reasons for these changes, such as exiting the federated learning training task due to the inability to use green energy.

[0080] In one embodiment, the task operation request includes: a distributed learning training request; correspondingly, the node selection criterion is a distributed training node selection criterion, and the task operation report is a distributed training report.

[0081] In one embodiment, the distributed training report includes at least one of the following: candidate distributed training nodes, distributed training green energy usage, and participation information for each distributed training node. In one example, distributed training green energy usage may include the green energy information of the entire distributed training task and / or each distributed training node, such as, but not limited to, whether each distributed training node uses green energy, the type of green energy used, carbon emissions, etc.; it may also include: the green energy usage of the entire learning process, such as the proportion of nodes using green energy and the number of nodes using green energy. In one example, the participation information for each distributed training node may be the contribution of each distributed training node to the distributed training process, the amount of data for each participating node, training time, etc.; in one example, the contribution of each distributed training node to the distributed training process is used to characterize the magnitude of each distributed training node's contribution to the entire distributed training process, and may be expressed as a percentage of contribution or a score. In one example, candidate distributed training nodes may be one or more sets of distributed training nodes, and each set of distributed training nodes contains the identifiers of one or more distributed training nodes.

[0082] In one embodiment, the task operation request includes: an inference request; correspondingly, the node selection criterion is an inference node selection criterion, and the task operation report is an inference report.

[0083] In one embodiment, the inference report includes candidate inference latency information and candidate inference node information. In one example, the candidate inference node information may contain one or more sets of inference nodes, and each set of inference nodes contains identifiers of one or more inference nodes. In one example, candidate inference latency information refers to one or more inference latency information. In one example, a first network element node may send candidate computation latency information to a second network element node, so that the second network element node can select one of the candidate inference latency information for use in the inference process.

[0084] In one embodiment, the task operation request includes: a computation request; correspondingly, the node selection criterion is a computation node selection criterion, and the task operation report is a computation report.

[0085] In one embodiment, the computation report includes candidate computation latency information and candidate computation node information. In one example, the candidate computation node information may contain one or more sets of computation nodes, and each set of computation nodes contains identifiers of one or more computation nodes. In one example, candidate computation latency information refers to one or more computation latency information sets. In one example, a first network element node may send candidate computation latency information to a second network element node, so that the second network element node can select one of the computation latency information sets for use in the computation process.

[0086] In one embodiment, Figure 3 This is a flowchart of another node selection method provided in this application embodiment. This embodiment is applied to the situation of selecting various operating nodes in a network. This embodiment can be executed by a second network element node. Figure 3 As shown, this embodiment includes: S310-S320.

[0087] S310, Send a task operation request to the first network element node.

[0088] S320, Receive the task operation report sent by the first network element node; wherein, the task operation report includes node selection information; the node selection information is obtained by the first network element node in selecting nodes according to the node selection criteria.

[0089] In one embodiment, the node selection criteria are obtained in one of the following ways: sent by the second network element node; pre-configured; or privately owned.

[0090] In one embodiment, the node selection information includes at least one of the following: candidate node information; confirmed node information; node change information.

[0091] In one embodiment, the node selection information is candidate node information; the node selection method applied to the second network element node further includes: feeding back confirmation node information to the first network element node.

[0092] In one embodiment, the task operation report also includes the task execution result; wherein the task execution result is obtained by the first network element node performing the task operation based on the node selected by itself or the node confirmed by the second network element node.

[0093] In one embodiment, the task operation request includes at least one of the following: federated learning training request; distributed learning training request; inference request; computation request.

[0094] In one embodiment, at least one of the federated learning training request, distributed learning training request, inference request, and computation request includes node selection criteria.

[0095] In one embodiment, the node selection criteria include at least one of the following: node green energy information; node infrastructure information; node selection strategy.

[0096] In one embodiment, the node green energy information includes at least one of the following: renewable energy information; carbon emission information; green energy ratio information.

[0097] In one embodiment, the node infrastructure information includes at least one of the following: hardware information; computing performance indicators; memory and storage performance indicators; parallel processing capability indicators; energy efficiency ratio indicators; and comprehensive computing power indicators.

[0098] In one embodiment, the node selection strategy includes at least one of the following: minimum time strategy; optimal performance strategy; minimum energy consumption strategy.

[0099] In one embodiment, the task operation request includes a federated learning training request, and the node selection criterion is the server selection criterion; sending the task operation request to the first network element node includes:

[0100] A federated learning request carrying server selection criteria is sent to the first network element node;

[0101] Correspondingly, receiving the task operation report sent by the first network element node includes: receiving the federated learning training report sent by the first network element node; wherein, the federated learning training report carries the identifier of the federated learning server; the federated learning server is selected by the first network element node based on the server selection criteria.

[0102] In one embodiment, the federated learning server is selected by the first network element node based on the server selection criteria, including:

[0103] If the first network element node meets the server selection criteria, then the first network element node selects itself as the federated learning server and notifies the second network element node.

[0104] If the first network element node does not meet the server selection criteria, the first network element node queries the registration and discovery service producer to obtain the set of federated learning producers, queries the federated learning capabilities supported by the available federated learning producers in the set of federated learning producers, selects a federated learning server according to the server selection criteria, and notifies the second network element node.

[0105] In one embodiment, the task operation request includes a federated learning training request; the node selection method applied to the second network element node includes:

[0106] Query the federated learning producer set from the registration and discovery service producer;

[0107] Send a query request to the first network element node for information on the federated learning capabilities supported by the available federated learning producers in the federated learning producer set;

[0108] Receive information from the second network element node regarding the federated learning capabilities supported by the available federated learning producers in the federated learning producer set;

[0109] Based on the federated learning capability information, select one of the available federated learning producers as the federated learning server.

[0110] In one embodiment, the federated learning capability information includes at least one of the following: federated learning support capability; federated learning client support capability; federated learning server support capability; supported samples and features; and data information of the second network element node.

[0111] In one embodiment, the task operation request is a federated learning training request, and the node selection criterion is a client selection criterion; sending the task operation request to the first network element node includes:

[0112] Send a federated learning training request carrying client selection criteria to the first network element node;

[0113] Accordingly, receiving the task operation report sent by the first network element node includes: receiving the federated learning training report sent by the first network element node; wherein the federated learning training report is generated by the federated learning client through federated learning training, and the federated learning client is selected by the first network element node based on the client selection criteria; or, selecting an available federated learning client based on the client selection criteria and informing the second network element node so that the second network element node can confirm the federated learning client.

[0114] In one embodiment, the federated learning training report includes at least one of the following: candidate federated learning clients; federated learning green energy usage; and changes in federated learning clients.

[0115] In one embodiment, the task operation request includes: a distributed learning training request; correspondingly, the node selection criterion is a distributed training node selection criterion, and the task operation report is a distributed training report.

[0116] In one embodiment, the distributed training report includes at least one of the following: candidate distributed training nodes, distributed training green energy usage, and participation information for each distributed training node.

[0117] In one embodiment, the task operation request includes: an inference request; correspondingly, the node selection criterion is an inference node selection criterion, and the task operation report is an inference report.

[0118] In one embodiment, the inference report includes candidate inference latency information and candidate inference node information.

[0119] In one embodiment, the task operation request includes: a computation request; correspondingly, the node selection criterion is a computation node selection criterion, and the task operation report is a computation report.

[0120] In one embodiment, the computation report includes: candidate computation latency information and candidate computation node information.

[0121] It should be noted that the explanation and implementation of parameters such as task operation request, node selection criteria, task operation report, candidate node information, confirmed node information, and node change information involved in the node selection method applied to the second network element node can be found in the description of the corresponding parameters in the node selection method applied to the first network element node, and will not be repeated here.

[0122] In the following embodiments, taking the first network element node as the Producer and the second network element node as the Consumer as an example, the general node selection process includes two scenarios:

[0123] First, the Consumer determines the node by issuing a task operation request, which includes the node selection criteria. The Producer selects candidate node information according to the node selection criteria and sends the candidate node information to the Consumer. The Consumer then determines and confirms the node information from the candidate node information.

[0124] Secondly, the Producer determines the nodes. The Producer selects nodes based on pre-configured or private node selection criteria sent by the Consumer, obtains confirmation node information, and optionally informs the Consumer of the confirmed nodes.

[0125] Both Consumer and Producer mentioned above can be UE or Figure 1 Any network entity (OSS / BSS, Domain-OAM, Cross-domain OAM, NF, gNB).

[0126] Example 1

[0127] In this embodiment, the first network element node is taken as the Producer and the second network element node is taken as the Consumer. The node selection process is explained by the Consumer confirming the node information, that is, the scenario in which the Consumer determines the node is explained. Figure 4 This is a schematic diagram illustrating an interactive implementation of node selection provided in an embodiment of this application. For example... Figure 4As shown, the node selection process in this embodiment includes the following steps:

[0128] S410, Send task operation request.

[0129] The Consumer sends a task operation request to the Producer, which may include node selection criteria.

[0130] S420. Select available nodes according to the node selection criteria to obtain candidate node information.

[0131] The Producer selects available nodes based on private or pre-configured node selection criteria received from the Consumer, forming candidate node information.

[0132] S430, Send task operation report.

[0133] The Producer sends a task operation report to the Consumer, which includes candidate node information.

[0134] S440, Send confirmation node information.

[0135] After selecting the desired node from the candidate node information, the Consumer obtains the confirmed node information and sends the confirmed node information to the Producer.

[0136] S450. Perform task operations using one or more nodes from the confirmed node information.

[0137] The Producer executes associated task operations based on one or more nodes in the confirmed node information.

[0138] S460, Send task operation report.

[0139] The Producer sends a task operation report to the Consumer, which includes information on node changes during task execution.

[0140] Example 2

[0141] In this embodiment, the first network element node is taken as the Producer and the second network element node is taken as the Consumer. The node selection process is explained by the Producer confirming the node information, that is, the above-mentioned scenario in which the Producer determines the node is explained. Figure 5 This is a schematic diagram illustrating another interactive implementation of node selection provided in an embodiment of this application. For example... Figure 5 As shown, the node selection process in this embodiment includes the following steps:

[0142] S510, Send task operation request.

[0143] The Consumer sends a task operation request to the Producer, which may include node selection criteria.

[0144] S520. Select available nodes according to the node selection criteria and obtain confirmed node information.

[0145] The Producer selects nodes based on private or pre-configured node selection criteria received from the Consumer and generates confirmation node information.

[0146] S530, Send task operation report.

[0147] The Producer can send a task operation report to the Consumer, which may include information about the selected confirmation node; and the Producer receives the response message associated with the task operation report.

[0148] S540. Confirm the confirmation node information selected by the Producer.

[0149] In actual operation, S530 and S540 may not be executed.

[0150] S550, Perform task operations using one or more nodes from the confirmed node information.

[0151] The Producer executes associated task operations based on one or more nodes in the confirmed node information.

[0152] S560, Send task operation report.

[0153] The Producer sends a task operation report to the Consumer, which includes information on node changes during task execution.

[0154] On Figure 4 and 5 In this context, node selection criteria may include: node green energy information, node infrastructure information, and node selection strategy.

[0155] In one example, node green energy information may include the following parameters:

[0156] Renewable energy information, used at the single-node level, indicates the requirements that a single node must meet. For example, a renewable energy indicator shows whether the node uses renewable energy; the type of renewable energy, such as wind, hydro, or solar power, is also included.

[0157] Carbon emission information, used at the single-node granularity, indicates the requirements that a single node must meet. For example, the maximum carbon emission value, in terms of kilograms of carbon dioxide, is equivalent to energy consumption (kWh).

[0158] Green energy ratio information is used at the multi-node granularity level for the entire task. This includes information such as the ratio of green energy to renewable energy nodes, and the minimum number of nodes required.

[0159] Node infrastructure information, such as hardware information, computing performance indicators; memory and storage performance indicators; parallel processing capability indicators; energy efficiency ratio indicators; and comprehensive computing power indicators. Hardware information includes GPU, CPU, DPU, etc.

[0160] In one example, computational performance metrics may include: FLOPS, MIPS, and TOPS.

[0161] Among them, FLOPS (Floating Point Operations Per Second) is an important indicator for measuring computer performance, especially in the fields of scientific computing, deep learning and high-performance computing. It is defined as the number of floating-point operations that a computer can perform per second. The higher the number of floating-point operations, the stronger the computing power of the computer.

[0162] MIPS (Million Instructions Per Second) is a metric that measures the execution speed of a CPU, reflecting the number of millions of instructions a CPU can process per second. It is suitable for evaluating the performance of a CPU when executing a basic instruction set and is often used in embedded systems and other fields.

[0163] TOPS (Tera Operations Per Second) refers to the number of trillions of operations that can be performed per second. It is particularly suitable for deep learning tasks in the field of artificial intelligence and focuses on handling "integer" operations, such as 8-bit and 16-bit integer operations.

[0164] In one example, memory and storage performance metrics may include memory bandwidth and storage performance.

[0165] Memory bandwidth refers to the amount of data that can be transferred in memory per unit of time, usually measured in data transfer per second (e.g., GB / s). High memory bandwidth can reduce data transfer waiting time and improve system efficiency.

[0166] Storage performance refers to the read and write speed of storage devices, including sequential read and write speed and random read and write speed. High-performance storage systems can quickly read and write large amounts of data and reduce data access latency.

[0167] In one example, parallel processing capability metrics may include the number of cores and the number of threads.

[0168] Number of cores and number of threads: The number of cores refers to the number of independent processing units in a processor, while the number of threads refers to the number of tasks that a processor can process simultaneously. Multi-core processors and multi-threading technology can improve the parallel processing capability of a system, enabling multiple tasks to run at the same time.

[0169] In one example, the energy efficiency ratio metric could include: energy efficiency ratio.

[0170] Energy efficiency ratio refers to the amount of computation that can be completed per unit of energy consumption. It is usually measured by the number of operations per watt (such as FLOPS / W, TOPS / W). A high energy efficiency ratio means that more computational tasks can be completed with the same energy consumption.

[0171] In one example, the overall computing power metric may include: computing power index.

[0172] The computing power index is a comprehensive indicator that comprehensively assesses the computing power scale, environment, application status, and development trends of a region or country, covering multiple factors such as hardware, software, network, and data.

[0173] Node selection strategies include minimum inference latency strategy, which uses minimum inference latency indication information or specific latency value constraints; minimum training time strategy, which uses minimum training time indication information or specific training time value constraints; and optimal performance strategy, which uses optimal performance training indication information or specific model performance value constraints.

[0174] Example 3

[0175] Figure 6 This is a schematic diagram illustrating another interactive implementation of node selection provided in this application. This embodiment uses the first network element node as the federated learning producer (also referred to as ML Training Producer (or FL Producer; FLserver)) and the second network element node as the federated learning consumer (also referred to as ML Training Consumer (FLConsumer)) as an example to illustrate the process of the federated learning server (FL server) selecting the federated learning client (FL Client) during the federated learning (FL) process, using node selection criteria as the standard. In this embodiment, the task operation request is the federated learning training request, and the task operation report is the federated learning training report.

[0176] It should be noted that in this embodiment, the ML Training Producer can be an MLTrainingFunction, which can be OAM, NWDAF, or gNB.

[0177] like Figure 6 As shown, the node selection process in this embodiment includes the following steps:

[0178] S610, Perform an ML Training Producer query.

[0179] The ML Training Consumer sends an ML Training Producer query request to the Registration & Discovery Producer, requesting a report on ML Training Producers to obtain a federated learning producer set. The Registration & Discovery Producer sends the identifier of the ML Training Producer, i.e., the Distinguished Name (DN).

[0180] S621. Query the FL capabilities of available ML Training Producers.

[0181] The ML Training Consumer, based on the retrieved ML Training Producer DN, queries the status information of all ML Training Producers in the federated learning producer set to determine the available ML Training Producers, and then queries the FL capabilities information of the available ML Training Producers (i.e., the available federated learning producers mentioned above). FL capability information includes whether it supports FL (i.e., federated learning support capability), whether it supports acting as an FL Server or FL Client (i.e., federated learning client support capability or federated learning server support capability mentioned above), supported samples (User Equipment ID, Network Slice Selection Assistance Information (NSSAI)) and features (Performance Management (PM), Key Performance Indicators (KPIs)), and Producer data information (e.g., data volume, number of samples, number of features, etc.).

[0182] S622. Determine the FL server.

[0183] The ML Training Consumer selects an ML Training Producer as the FL Server from the available ML Training Producers based on the FL capability information of the available ML Training Producers retrieved.

[0184] It should be noted that S621 and S622 are executed when the Consumer determines the FL Server.

[0185] S631, FL training request, requesting confirmation of the FL server.

[0186] The ML Training Consumer sends a federated learning training request to the ML Training Producer based on the retrieved ML Training Producer DN, requesting the ML Training Producer to determine the FL Server. This federated learning training request includes server selection criteria.

[0187] S632. Select and confirm the FL server.

[0188] The ML Training Producer selects and determines the server. This includes:

[0189] Option 1. Assess whether it meets the conditions for being an FL Server (i.e., assess whether it meets the server selection criteria). If it does, it becomes an FL Server and enters S633.

[0190] Option 2. If the requirement to be a FL Server is not met (i.e., the server selection criteria are not met), the ML TrainingProducer queries the Registration & Discovery Producer to obtain the FL capability information indicated by other available ML Training Producers in the federated learning producer set, and selects a federated learning server as the FL Server according to the server selection criteria.

[0191] S633, Send FL training report.

[0192] The ML Training Producer sends an FL training report to the ML Training Consumer, which includes the identifier of the FL Server.

[0193] It should be noted that when the Producer determines the FL Server, S631 and S633 are executed.

[0194] S640, Receive Federal Learning and Training Requests.

[0195] The ML Training Consumer sends a federated learning training request to the ML Training Producer (FL Server), which includes node selection criteria, i.e., the Client selection criteria. In this example, the Client selection criteria include node green energy information, including the following parameters:

[0196] Renewable energy information, used at the single-node level, indicates the requirements that a single node must meet. For example, a renewable energy indicator shows whether the node uses renewable energy; the type of renewable energy, such as wind, hydro, or solar power, is also included.

[0197] Carbon emission information, used at the single-node granularity, indicates the requirements that a single node must meet. For example, the maximum carbon emission value, in terms of kilograms of carbon dioxide, is equivalent to energy consumption (kWh).

[0198] Green energy ratio information is used at the multi-node granularity level for the entire task. This includes information such as the ratio of green energy / renewable energy nodes used, and the minimum number of nodes required.

[0199] S650, perform an ML Training Function query.

[0200] The ML Training Producer (Server) sends an MLTraining Function query request to the Registration & Discovery Producer, requesting a report of the ML Training Producer that can act as a client.

[0201] S660, query the energy supply status of the ML Training Function.

[0202] Based on the ML Training Function identifier found, the ML Training Producer (Server) further queries its FL status, FL capacity information, and green energy supply status.

[0203] S670, select FL Client.

[0204] The ML Training Producer (Server) selects FL Clients based on the consumer's Client selection criteria.

[0205] S680, conduct federal learning training.

[0206] ML Training Producer performs federated learning training.

[0207] S690, Send the Federal Learning and Training Report.

[0208] The ML Training Producer sends federated learning training reports to the ML Training Consumer. Each client's training report includes its green energy usage, such as renewable energy information for each node, carbon emission information, and the renewable energy ratio for all nodes. Additionally, the federated learning training report includes client changes, such as the addition or removal of FL clients based on client selection criteria and the reasons for these changes.

[0209] Example 4

[0210] Figure 7 This is a schematic diagram illustrating another interactive implementation of node selection provided in this application embodiment. This embodiment takes the first network element node as a distributed training producer (also referred to as an ML Training Producer or DL ​​Producer) and the second network element node as a distributed training consumer (or DL ​​Consumer) as an example to explain the process of selecting distributed training nodes in the distributed learning and training process, using the node selection criteria as the basis. In this embodiment, the task operation request is a distributed learning and training request, and the task operation report is a distributed training report.

[0211] It should be noted that in this example, the ML Training Producer can be an MLTrainingFunction, which can be OAM, NWDAF, or gNB. In this example, the node is the MLTrainingFunction.

[0212] S710 sends a distributed learning and training request.

[0213] The ML Training Consumer sends a distributed learning training request to the ML Training Producer, which includes node selection criteria, i.e., distributed training node selection criteria. In this example, the distributed training node selection criteria include node green energy information, infrastructure information, and a minimum training time strategy.

[0214] In one example, the node's green energy information includes the following parameters:

[0215] Renewable energy information, used at the single-node level, indicates the requirements that a single node must meet. For example, a renewable energy indicator shows whether the node uses renewable energy; the type of renewable energy, such as wind, hydro, or solar power, is also included.

[0216] Carbon emission information, used at the single-node granularity, indicates the requirements that a single node must meet. For example, the maximum carbon emission value, in terms of kilograms of carbon dioxide, is equivalent to energy consumption (kWh).

[0217] Green energy ratio information is used at the multi-node granularity level for the entire task. This includes information such as the ratio of green energy / renewable energy nodes used, and the minimum number of nodes required.

[0218] Node infrastructure information, such as CPU, GPU, DPU, memory bandwidth, and storage performance.

[0219] Node selection strategy, minimum training time strategy, minimum training time indication information or constraints based on specific training time values; optimal performance strategy, optimal performance training indication information or constraints based on specific model performance metrics, etc.

[0220] S720, perform an ML Training Function query.

[0221] The ML Training Producer sends an ML TrainingFunction query request to the Registration & Discovery Producer, requesting reports from other ML Training Producers.

[0222] S730, query the energy supply status of the ML Training Function.

[0223] Based on the ML Training Function identifier found, the ML Training Producer further queries the energy supply and infrastructure of the MLTraining Function.

[0224] S741. Perform DL node selection and confirmation.

[0225] When the Producer selects nodes, step S741 is executed. The ML Training Producer selects and confirms DL nodes. Nodes participating in DL training must meet the distributed training node selection criteria. In this example, nodes using renewable energy are prioritized to meet the green energy ratio requirement. Simultaneously, to meet the minimum training time strategy, computationally efficient training nodes are prioritized.

[0226] SS751, Send candidate node information.

[0227] The ML Training Producer sends candidate node information to the Consumer, including candidate distributed training nodes (that meet the distributed training node selection criteria) and information related to the distributed training node selection criteria.

[0228] S752, Confirmation of participation in distributed training nodes.

[0229] The Consumer confirms with the Producer which nodes will participate in the distributed training.

[0230] S753, Select distributed training nodes.

[0231] The ML Training Producer confirms DL participation with the confirmed DL nodes.

[0232] It should be noted that when the Consumer selects a node, S751-S753 are executed.

[0233] S760 and ML Training Producer perform distributed training.

[0234] S770, send distributed training reports.

[0235] Each ML Training Function sends a distributed training report to the ML Training Producer, which includes the status of each node in the distributed training process, such as the use of green energy, and the contribution of each node to the entire training process, in the form of contribution percentage or score.

[0236] S780, send distributed training reports.

[0237] The ML Training Producer sends a distributed training report to the ML Training Consumer, which includes information on changes to distributed training nodes, such as nodes added or removed based on distributed training node selection criteria and the reasons thereof. It also includes information on the renewable energy ratio and carbon emissions of all nodes.

[0238] Example 5

[0239] Figure 8This is a schematic diagram illustrating another interactive implementation of node selection provided in this application embodiment. Taking the first network element node as an AIML Inference Producer and the second network element node as an AIML Inference Consumer as an example, this embodiment explains the process of using node selection criteria to select nodes for Consumer inference request execution. In this embodiment, the task operation request is an inference request, and the task operation report is an inference report.

[0240] It should be noted that in this embodiment, the AIML Inference Producer can be an AIMLInferenceFunction, which can be OAM, NWDAF, or gNB. The node in this embodiment is the AIMLInferenceFunction.

[0241] like Figure 8 As shown, the node selection process in this embodiment includes the following steps:

[0242] S810, Send inference request.

[0243] The AIML Inference Consumer sends an inference request to the AIML Inference Producer, which includes the node selection strategy used in the inference request. In this embodiment, it is assumed that the node selection strategy is the minimum latency strategy, and the minimum inference latency indication information is constrained by a specific latency value.

[0244] S820, perform an AIML Training Function query.

[0245] The AIML Inference Producer queries the Registration & Discovery Producer for the AIMLInferenceFunction and obtains the Data Name (DN) of the AIML Inference Function.

[0246] S831, Select the inference node.

[0247] When the Producer selects a node, S831 is executed. The AIML Inference Producer selects an inference node that meets the inference latency requirements of the AIML Inference Consumer. Multiple nodes can be used for inference.

[0248] S841, Send candidate node information.

[0249] The AIML Inference Producer sends candidate inference node information to the Consumer, including candidate inference nodes (that meet the node selection criteria) and the information involved in the node selection criteria. In one example, there can be multiple nodes used for inference.

[0250] S842, Confirmation of participating inference nodes.

[0251] AIML Inference Consumer confirms with Producer which nodes are involved in inference.

[0252] S843, Select the inference node.

[0253] AIML Inference Producer confirms inference nodes.

[0254] When the Consumer selects a node, S841-S843 are executed.

[0255] S850, Execute the inference process. The AIML Inference Producer executes the inference process.

[0256] S860, Send reasoning report.

[0257] The AIML Inference Producer sends an inference report to the AIML Inference Consumer. This report may include candidate inference latency information and candidate inference node information, allowing the AIML Inference Consumer to select the appropriate inference latency and node. In one example, the computation report may also include inference latency information and inference node information confirmed by the AIML Inference Producer.

[0258] Example 6

[0259] Figure 9 This is a schematic diagram illustrating another interactive implementation of node selection provided in this application embodiment. Taking the first network element node as a Computing Producer and the second network element node as a Computing Consumer as an example, this embodiment explains the process of using node selection criteria to select nodes for Consumer computation requests. In this embodiment, the task operation request is a computation request, and the task operation report is a computation report.

[0260] It should be noted that in this embodiment, the Computing Producer can be a ComputingFunction, which can be OAM, NWDAF, or gNB. The node in this embodiment is the ComputingFunction.

[0261] like Figure 9 As shown, the node selection process in this embodiment includes the following steps:

[0262] S910, Send calculation request.

[0263] The Computing Consumer sends a computation request to the Computing Producer, which includes the node selection strategy used in the computation request. In this embodiment, it is assumed that the node selection strategy is the minimum computation latency strategy, and the minimum computation latency indication information is constrained by a specific latency value.

[0264] S920, Perform a Computing Function query.

[0265] The Computing Producer queries the Registration & Discovery Producer for the Computing Function and obtains the DN of the Computing Function.

[0266] S931. Select computing nodes.

[0267] When the Producer selects a node, S931 is executed. The Computing Producer selects a computing node that meets the computation latency requirements of the Computing Consumer. Multiple nodes can be used for computation.

[0268] S941, Send candidate node information.

[0269] The Computing Producer sends candidate compute node information to the Computing Consumer, including candidate compute nodes (that meet the node selection criteria) and the information involved in the node selection criteria. Multiple nodes can be used for computation.

[0270] S942, Confirmation of participating computing nodes.

[0271] The Computing Consumer confirms the nodes participating in the computation with the Computing Producer.

[0272] S943. Select the computing nodes.

[0273] The Computing Producer confirms the compute nodes.

[0274] When the Computing Consumer selects a node, S941-S943 are executed.

[0275] S950, Execute the calculation process.

[0276] The Computing Producer performs the computation process.

[0277] S960, Send calculation report.

[0278] The Computing Producer sends a computation report to the Computing Consumer. In one example, the computation report includes candidate computation latency information and candidate computation node information, allowing the Computing Consumer to select the appropriate computation latency and computation node. In another example, the computation report may also include computation latency information and computation node information confirmed by the Computing Producer.

[0279] In one embodiment, Figure 10 This is a structural block diagram of a node selection device provided in an embodiment of this application. This embodiment is applied to a first network element node. Figure 10 As shown, the node selection device in this embodiment includes: a receiving module 1010, a selection module 1020, and a sending module 1030.

[0280] The receiving module 1010 is configured to receive task operation requests sent by the second network element node;

[0281] Select module 1020 and configure it to select nodes based on node selection criteria;

[0282] The sending module 1030 is configured to send a task operation report to the second network element node; wherein, the task operation report includes node selection information.

[0283] In one embodiment, the node selection criteria are obtained in one of the following ways: sent by the second network element node; pre-configured; or privately owned.

[0284] In one embodiment, the node selection information includes at least one of the following: candidate node information; confirmed node information; node change information.

[0285] In one embodiment, the node selection information is candidate node information; the node selection method applied to the first network element node further includes: receiving confirmation node information fed back by the second network element node.

[0286] In one embodiment, the node selection device applied to the first network element node further includes:

[0287] The execution module is configured to perform task operations based on the node selected by the first network element node or the node confirmed by the second network element node.

[0288] Correspondingly, the task operation report also includes the task execution results.

[0289] In one embodiment, the task operation request includes at least one of the following: federated learning training request; distributed learning training request; inference request; computation request.

[0290] In one embodiment, at least one of the federated learning training request, distributed learning training request, inference request, and computation request includes node selection criteria.

[0291] In one embodiment, the node selection criteria include at least one of the following: node green energy information; node infrastructure information; node selection strategy.

[0292] In one embodiment, the node green energy information includes at least one of the following: renewable energy information; carbon emission information; green energy ratio information.

[0293] In one embodiment, the node infrastructure information includes at least one of the following: hardware information; computing performance indicators; memory and storage performance indicators; parallel processing capability indicators; energy efficiency ratio indicators; and comprehensive computing power indicators.

[0294] In one embodiment, the node selection strategy includes at least one of the following: minimum time strategy; optimal performance strategy; minimum energy consumption strategy.

[0295] In one embodiment, the task operation request includes a federated learning training request, and the node selection criterion is the server selection criterion; receiving the task operation request sent by the second network element node includes:

[0296] Receive the federated learning training request sent by the second network element node, which carries the server selection criteria;

[0297] Accordingly, node selection is performed based on node selection criteria, including: determining the federated learning server based on server selection criteria;

[0298] Sending a task operation report to the second network element node includes: sending a federated learning training report to the second network element node; wherein the federated learning training report carries the identifier of the federated learning server.

[0299] In one embodiment, determining the federated learning server based on server selection criteria includes:

[0300] If the first network element node meets the server selection criteria, then the first network element node selects itself as the federated learning server and notifies the second network element node.

[0301] If the first network element node does not meet the server selection criteria, the first network element node queries the registration and discovery service producer to obtain the set of federated learning producers, queries the federated learning capabilities supported by the available federated learning producers in the federated learning producer set, selects a federated learning server according to the server selection criteria, and notifies the second network element node.

[0302] In one embodiment, the task operation request includes a federated learning training request; the node selection device applied to the first network element node further includes:

[0303] The receiving module is also configured to receive a query request from the second network element node for information on federated learning capabilities supported by available federated learning producers in the federated learning producer set; wherein, the federated learning producer set is obtained by the second network element node from the registration and discovery service producer.

[0304] The sending module is also configured to send information about the federated learning capabilities supported by the available federated learning producers in the federated learning producer set to the second network element node, so that the second network element node can select one of the available federated learning producers as the federated learning server.

[0305] In one embodiment, the federated learning capability information includes at least one of the following: federated learning support capability; federated learning client support capability; federated learning server support capability; supported samples and features; and data information of the second network element node.

[0306] In one embodiment, the task operation request is a federated learning training request, and the node selection criterion is a client selection criterion; receiving the task operation request sent by the second network element node includes:

[0307] Receive the federated learning training request sent by the second network element node, which carries the client selection criteria;

[0308] Accordingly, node selection is performed according to node selection criteria, including: selecting federated learning clients based on client selection criteria; or, selecting available federated learning clients based on client selection criteria and informing the second network element node so that the second network element node can confirm the federated learning client.

[0309] Send task operation reports to the second network element node, including: sending federated learning training reports to the second network element node.

[0310] In one embodiment, the federated learning training report includes at least one of the following: candidate federated learning clients; federated learning green energy usage; and changes in federated learning clients.

[0311] In one embodiment, the task operation request includes: a distributed learning training request; correspondingly, the node selection criterion is a distributed training node selection criterion, and the task operation report is a distributed training report.

[0312] In one embodiment, the distributed training report includes at least one of the following: candidate distributed training nodes, distributed training green energy usage, and participation information for each distributed training node.

[0313] In one embodiment, the task operation request includes: an inference request; correspondingly, the node selection criterion is an inference node selection criterion, and the task operation report is an inference report.

[0314] In one embodiment, the inference report includes candidate inference latency information and candidate inference node information.

[0315] In one embodiment, the task operation request includes: a computation request; correspondingly, the node selection criterion is a computation node selection criterion, and the task operation report is a computation report.

[0316] In one embodiment, the computation report includes: candidate computation latency information and candidate computation node information.

[0317] The node selection device provided in this embodiment is configured to achieve... Figure 2 The node selection method applied to the first network element node in the illustrated embodiment is similar in principle and technical effect to the node selection device provided in this embodiment, and will not be described again here.

[0318] In one embodiment, Figure 11 This is a structural block diagram of another node selection device provided in an embodiment of this application. This embodiment is applied to a second network element node. Figure 11 As shown, the node selection device in this embodiment includes a sending module 1110 and a receiving module 1120.

[0319] The sending module 1110 is configured to send a task operation request to the first network element node.

[0320] The receiving module 1120 is configured to receive a task operation report sent by the first network element node; wherein, the task operation report includes node selection information; the node selection information is obtained by the first network element node in selecting nodes according to the node selection criteria.

[0321] In one embodiment, the node selection criteria are obtained in one of the following ways: sent by the second network element node; pre-configured; or privately owned.

[0322] In one embodiment, the node selection information includes at least one of the following: candidate node information; confirmed node information; node change information.

[0323] In one embodiment, the node selection information is candidate node information; the node selection method applied to the second network element node further includes: feeding back confirmation node information to the first network element node.

[0324] In one embodiment, the task operation report also includes the task execution result; wherein the task execution result is obtained by the first network element node performing the task operation based on the node selected by itself or the node confirmed by the second network element node.

[0325] In one embodiment, the task operation request includes at least one of the following: federated learning training request; distributed learning training request; inference request; computation request.

[0326] In one embodiment, at least one of the federated learning training request, distributed learning training request, inference request, and computation request includes node selection criteria.

[0327] In one embodiment, the node selection criteria include at least one of the following: node green energy information; node infrastructure information; node selection strategy.

[0328] In one embodiment, the node green energy information includes at least one of the following: renewable energy information; carbon emission information; green energy ratio information.

[0329] In one embodiment, the node infrastructure information includes at least one of the following: hardware information; computing performance indicators; memory and storage performance indicators; parallel processing capability indicators; energy efficiency ratio indicators; and comprehensive computing power indicators.

[0330] In one embodiment, the node selection strategy includes at least one of the following: minimum time strategy; optimal performance strategy; minimum energy consumption strategy.

[0331] In one embodiment, the task operation request includes a federated learning training request, and the node selection criterion is the server selection criterion; sending the task operation request to the first network element node includes:

[0332] A federated learning request carrying server selection criteria is sent to the first network element node;

[0333] Correspondingly, receiving the task operation report sent by the first network element node includes: receiving the federated learning training report sent by the first network element node; wherein, the federated learning training report carries the identifier of the federated learning server; the federated learning server is selected by the first network element node based on the server selection criteria.

[0334] In one embodiment, the federated learning server is selected by the first network element node based on the server selection criteria, including:

[0335] If the first network element node meets the server selection criteria, then the first network element node selects itself as the federated learning server and notifies the second network element node.

[0336] If the first network element node does not meet the server selection criteria, the first network element node queries the registration and discovery service producer to obtain the set of federated learning producers, queries the federated learning capabilities supported by the available federated learning producers in the set of federated learning producers, selects a federated learning server according to the server selection criteria, and notifies the second network element node.

[0337] In one embodiment, the task operation request includes a federated learning training request; the node selection device applied to the second network element node includes:

[0338] The query module is configured to query the federated learning producer set from the registration and discovery service producers;

[0339] The sending module is also configured to send a query request to the first network element node for information on the federated learning capabilities supported by the federated learning producers available in the federated learning producer set.

[0340] The receiving module is also configured to receive information on federated learning capabilities supported by available federated learning producers in the federated learning producer set sent by the second network element node.

[0341] The module is configured to select one of the available federated learning producers as the federated learning server based on federated learning capability information.

[0342] In one embodiment, the federated learning capability information includes at least one of the following: federated learning support capability; federated learning client support capability; federated learning server support capability; supported samples and features; and data information of the second network element node.

[0343] In one embodiment, the task operation request is a federated learning training request, and the node selection criterion is a client selection criterion; sending the task operation request to the first network element node includes:

[0344] Send a federated learning training request carrying client selection criteria to the first network element node;

[0345] Accordingly, receiving the task operation report sent by the first network element node includes: receiving the federated learning training report sent by the first network element node; wherein the federated learning training report is generated by the federated learning client through federated learning training, and the federated learning client is selected by the first network element node based on the client selection criteria; or, selecting an available federated learning client based on the client selection criteria and informing the second network element node so that the second network element node can confirm the federated learning client.

[0346] In one embodiment, the federated learning training report includes at least one of the following: candidate federated learning clients; federated learning green energy usage; and changes in federated learning clients.

[0347] In one embodiment, the task operation request includes: a distributed learning training request; correspondingly, the node selection criterion is a distributed training node selection criterion, and the task operation report is a distributed training report.

[0348] In one embodiment, the distributed training report includes at least one of the following: candidate distributed training nodes, distributed training green energy usage, and participation information for each distributed training node.

[0349] In one embodiment, the task operation request includes: an inference request; correspondingly, the node selection criterion is an inference node selection criterion, and the task operation report is an inference report.

[0350] In one embodiment, the inference report includes candidate inference latency information and candidate inference node information.

[0351] In one embodiment, the task operation request includes: a computation request; correspondingly, the node selection criterion is a computation node selection criterion, and the task operation report is a computation report.

[0352] In one embodiment, the computation report includes: candidate computation latency information and candidate computation node information.

[0353] The node selection device provided in this embodiment is configured to achieve... Figure 3 The node selection method applied to the second network element node in the illustrated embodiment is similar in principle and technical effect to the node selection device provided in this embodiment, and will not be described again here.

[0354] In one embodiment, Figure 12 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application. Figure 12 As shown, the device provided in this application includes: a processor 1210, a memory 1220, and a communication module 1230. The device may contain one or more processors 1210. Figure 12Taking a processor 1210 as an example. The number of memory units 1220 in this device can be one or more. Figure 12 Taking a memory 1220 as an example, the processor 1210, memory 1220, and communication module 1230 of this device can be connected via a bus or other means. Figure 12 Taking a bus connection as an example, in this embodiment, the device can be either a first network element node or a second network element node.

[0355] The memory 1220, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, and modules, such as program instructions / modules corresponding to the device in any embodiment of this application (e.g., receiving module 1010, selection module 1020, and transmitting module 1030 applied to the node selection device of the first network element node). The memory 1220 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created according to the use of the device, etc. Furthermore, the memory 1220 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 1220 may further include memory remotely located relative to the processor 1210, and these remote memories can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0356] When the communication device is the first network element node, the device provided above can be configured to execute the node selection method applied to the first network element node provided in any of the above embodiments, and has the corresponding functions and effects.

[0357] When the communication device is a second network element node, the device provided above can be configured to execute the node selection method for the second network element node provided in any of the above embodiments, and has the corresponding functions and effects.

[0358] This application embodiment also provides a storage medium containing computer-executable instructions. When executed by a computer processor, the computer-executable instructions are used to perform a node selection method applied to a first network element node. The method includes: receiving a task operation request sent by a second network element node; selecting a node according to node selection criteria; and sending a task operation report to the second network element node; wherein the task operation report contains node selection information.

[0359] This application embodiment also provides a storage medium containing computer-executable instructions. When executed by a computer processor, the computer-executable instructions are used to execute a node selection method applied to a second network element node. The method includes: sending a task operation request to a first network element node; receiving a task operation report sent by the first network element node; wherein the task operation report contains node selection information; the node selection information is obtained by the first network element node in selecting a node according to node selection criteria.

[0360] Those skilled in the art will understand that the term user equipment covers any suitable type of wireless user equipment, such as mobile phones, portable data processing devices, portable web browsers, or vehicle-mounted mobile stations.

[0361] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. For example, some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, although this application is not limited thereto.

[0362] Embodiments of this application can be implemented by executing computer program instructions through the data processor of a mobile device, for example, in a processor entity, or through hardware, or through a combination of software and hardware. The computer program instructions can be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages.

[0363] Any block diagram of logical flow in the accompanying drawings of this application may represent program steps, or may represent interconnected logic circuits, modules, and functions, or may represent a combination of program steps and logic circuits, modules, and functions. The computer program may be stored on memory. Memory may be of any type suitable to the local technical environment and may be implemented using any suitable data storage technology, such as, but not limited to, read-only memory (ROM), random access memory (RAM), optical storage devices and systems (Digital Video Disc (DVD) or Compact Disk (CD)), etc. Computer-readable media may include non-transitory storage media. The data processor may be of any type suitable to the local technical environment, such as, but not limited to, general-purpose computers, special-purpose computers, microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and processors based on multi-core processor architectures.

[0364] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A node selection method, characterized in that, Applied to the first network element node, including: Receive task operation requests sent by the second network element node; Nodes are selected according to node selection criteria; Send a task operation report to the second network element node; wherein the task operation report includes node selection information.

2. The method according to claim 1, characterized in that, The node selection criteria can be obtained in one of the following ways: sent by the second network element node; pre-configured; or privately owned.

3. The method according to claim 1, characterized in that, The node selection information includes at least one of the following: candidate node information; confirmed node information; node change information.

4. The method according to claim 3, characterized in that, The node selection information is candidate node information; the method further includes: Receive confirmation node information from the second network element node.

5. The method according to claim 1 or 4, characterized in that, The method further includes: The task operation is executed according to the node selected by the first network element node or the node confirmed by the second network element node; Accordingly, the task operation report also includes the task execution results.

6. The method according to claim 1, characterized in that, The task operation request includes at least one of the following: federated learning training request; distributed learning training request; inference request; computation request.

7. The method according to claim 6, characterized in that, At least one of the federated learning training request, the distributed learning training request, the inference request, and the computation request includes a node selection criterion.

8. The method according to any one of claims 1-4, characterized in that, The node selection criteria include at least one of the following: node green energy information; node infrastructure information; node selection strategy.

9. The method according to claim 8, characterized in that, The green energy information of the node includes at least one of the following: renewable energy information; carbon emission information; green energy ratio information.

10. The method according to claim 8, characterized in that, The node infrastructure information includes at least one of the following: hardware information; computing performance indicators; memory and storage performance indicators; parallel processing capability indicators; energy efficiency ratio indicators; and comprehensive computing power indicators.

11. The method according to claim 8, characterized in that, The node selection strategy includes at least one of the following: minimum time strategy; best performance strategy; minimum energy consumption strategy.

12. The method according to claim 7, characterized in that, The task operation request is a federated learning training request, and the node selection criterion is a server selection criterion. The process of receiving the task operation request sent by the second network element node includes: Receive a federated learning training request sent by the second network element node, carrying the server selection criteria; Accordingly, the node selection based on the node selection criteria includes: determining the federated learning server based on the server selection criteria; Sending a task operation report to the second network element node includes: sending a federated learning training report to the second network element node; wherein the federated learning training report carries the identifier of the federated learning server.

13. The method according to claim 12, characterized in that, The process of determining the federated learning server based on the server selection criteria includes: If the first network element node meets the server selection criteria, then the first network element node selects itself as the federated learning server and notifies the second network element node. If the first network element node does not meet the server selection criteria, the first network element node queries the registration and discovery service producer to obtain the set of federated learning producers, queries the federated learning capabilities supported by the available federated learning producers in the set of federated learning producers, selects a federated learning server according to the server selection criteria, and notifies the second network element node.

14. The method according to claim 7, characterized in that, The task operation request is a federated learning training request; The method further includes: The system receives a query request from the second network element node regarding the federated learning capabilities supported by available federated learning producers in the federated learning producer set; wherein the federated learning producer set is obtained by the second network element node from the registration and discovery service producer. Send information about the federated learning capabilities supported by available federated learning producers in the federated learning producer set to the second network element node, so that the second network element node can select one of the available federated learning producers as the federated learning server.

15. The method according to claim 13 or 14, characterized in that, The federated learning capability information includes at least one of the following: federated learning support capability; federated learning client support capability; federated learning server support capability; supported samples and features; Data information of the second network element node.

16. The method according to claim 6, characterized in that, The task operation request is a federated learning training request, and the node selection criterion is a client selection criterion. The process of receiving the task operation request sent by the second network element node includes: Receive the federated learning training request sent by the second network element node, which carries the client selection criteria; Accordingly, the node selection based on node selection criteria includes: Federated learning client selection is performed based on the aforementioned client selection criteria; or... Based on the client selection criteria, an available federated learning client is selected and the second network element node is informed so that the second network element node can confirm the federated learning client. Sending a task operation report to the second network element node includes sending a federated learning training report to the second network element node.

17. The method according to claim 16, characterized in that, The federated learning training report shall include at least one of the following: candidate federated learning clients; federated learning green energy usage; and changes in federated learning clients.

18. The method according to claim 1, characterized in that, The task operation request includes: a distributed learning and training request; correspondingly, the node selection criterion is a distributed training node selection criterion, and the task operation report is a distributed training report.

19. The method according to claim 18, characterized in that, The distributed training report includes at least one of the following: candidate distributed training nodes, distributed training green energy usage, and participation information for each distributed training node.

20. The method according to claim 1, characterized in that, The task operation request includes: a reasoning request; correspondingly, the node selection criterion is a reasoning node selection criterion, and the task operation report is a reasoning report.

21. The method according to claim 20, characterized in that, The inference report includes candidate inference latency information and candidate inference node information.

22. The method according to claim 1, characterized in that, The task operation request includes a calculation request; correspondingly, the node selection criterion is a calculation node selection criterion, and the task operation report is a calculation report.

23. The method according to claim 22, characterized in that, The calculation report includes: candidate calculation latency information and candidate calculation node information.

24. A node selection method, characterized in that, Applied to the second network element node, including: Send a task operation request to the first network element node; The system receives a task operation report sent by the first network element node; wherein the task operation report includes node selection information; the node selection information is obtained by the first network element node through node selection according to node selection criteria.

25. The method according to claim 24, characterized in that, The node selection information is candidate node information; the method further includes: Feedback confirmation node information to the first network element node.

26. The method according to claim 24 or 25, characterized in that, The task operation report also includes the task execution result; wherein, the task execution result is obtained by the first network element node performing task operations based on the node it selects or the node confirmed by the second network element node.

27. The method according to claim 24, characterized in that, The task operation request includes a federated learning training request, and the node selection criterion is a server selection criterion; sending the task operation request to the first network element node includes: A federated learning training request carrying the server selection criteria is sent to the first network element node; Accordingly, receiving the task operation report sent by the first network element node includes: receiving the federated learning training report sent by the first network element node; wherein the federated learning training report carries the identifier of the federated learning server; the federated learning server is selected by the first network element node based on the server selection criteria.

28. The method according to claim 27, characterized in that, The federated learning server is selected by the first network element node based on the server selection criteria, including: If the first network element node meets the server selection criteria, then the first network element node selects itself as the federated learning server and notifies the second network element node. If the first network element node does not meet the server selection criteria, the first network element node queries the registration and discovery service producer to obtain the set of federated learning producers, queries the federated learning capabilities supported by the available federated learning producers in the set of federated learning producers, selects a federated learning server according to the server selection criteria, and notifies the second network element node.

29. The method according to claim 24, characterized in that, The task operation request includes a federated learning training request; the method includes: Query the federated learning producer set from the registration and discovery service producer; Send a query request to the first network element node for information on the federated learning capabilities supported by the available federated learning producers in the federated learning producer set; Receive information from the second network element node regarding the federated learning capabilities supported by the available federated learning producers in the federated learning producer set; Based on the federated learning capability information, one of the available federated learning producers is selected as the federated learning server.

30. The method according to claim 24, characterized in that, The task operation request includes: a federated learning training request, wherein the node selection criterion is a client selection criterion; the step of sending the task operation request to the first network element node includes: Send a federated learning training request carrying client selection criteria to the first network element node; Accordingly, receiving the task operation report sent by the first network element node includes: receiving the federated learning training report sent by the first network element node; wherein the federated learning training report is generated by the federated learning client through federated learning training, and the federated learning client is selected by the first network element node based on the client selection criteria.

31. A communication device, characterized in that, include: Memory, and one or more processors; The memory is configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the method as described in any one of claims 1-23 or 24-30.

32. A storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1-23 or 24-30.

Images