A method and system for socialized computing power supply and unified scheduling

Abstract

The application belongs to the technical field of computing power scheduling, and discloses a method and system for socialized computing power supply and unified scheduling. The method comprises six steps of computing power node registration, capability evaluation and security isolation, unified protocol access, task dynamic scheduling, usage metering, and automatic settlement and account distribution. The system is provided with six functional units to cooperatively realize the aforementioned method and adapt to multiple types of commercial computing power nodes. The application realizes unified registration and management of multiple types of commercial computing power nodes such as data centers, commercial store computing power cabins, Internet cafes / e-sports halls GPU clusters, and mobile nodes, integrates idle computing power resources scattered in different regions and different subjects into a standardized computing power network, integrates node authentication based on SM2 encryption, minute-level computing power performance detection, K8s+ network slice security isolation, and JSON format task communication, and solves the industry problems of no unified access specification and poor networking compatibility of different types of computing power nodes.

Description

Technical Field

[0001] This invention relates to the field of computing power scheduling technology, specifically to a method and system for socialized computing power supply and unified scheduling. Background Technology

[0002] With the development of the digital economy, the demand for computing resources in all sectors of society has exploded, and computing power scheduling has become a core technology in distributed computing and edge computing systems. At present, the computing power supply system in the industry is still based on centralized data centers as the core carrier. Although this model can provide stable computing power services, it has inherent problems such as high construction costs, long deployment cycles, and uneven regional coverage, making it difficult to flexibly match the differentiated computing power needs of different regions and scenarios. At the same time, there are a large number of commercial computing power nodes in society that can provide computing power services to the outside world, including data center computing power nodes, commercial store computing power nodes, internet cafe / e-sports arena computing power nodes, mobile computing power nodes, etc. These nodes have stable computing resources and operational foundations, and have become an important part of social idle computing power resources.

[0003] However, the existing socialized computing power supply and unified scheduling methods have the following defects in operation: socialized computing power resources are significantly decentralized, lack a unified registration and management system for various types of commercial computing power nodes, lack standardized recording methods for the basic information and computing power capabilities of various computing power nodes, cannot effectively integrate idle computing power resources, are difficult to form a large-scale, operable computing power network, and have low overall utilization rate and limited regional coverage of computing power resources. There is no standardized and unified access protocol that can adapt to multiple types and cross-entity computing power nodes. The access specifications of different computing power nodes are not uniform, resulting in poor network compatibility. Furthermore, there is a lack of specific security isolation configuration methods, making it impossible to achieve logical isolation between different computing power nodes. When multiple computing power nodes share the network, security risks such as resource interference and data leakage are likely to occur. At the same time, there is no stable communication and authentication link between nodes and the scheduling platform, resulting in insufficient network stability and security. The current computing power task scheduling method is simplistic, relying solely on basic node capabilities for task allocation. It fails to consider multiple dimensions such as task computing power requirements, latency requirements, and real-time geographic location, network transmission status, and load rate of nodes to achieve precise scheduling. Furthermore, the lack of specific scheduling algorithms results in low matching between computing power tasks and node resources, easily leading to node load imbalance. The measurement of computing power usage relies solely on usage duration or task completion volume, resulting in low measurement accuracy. Additionally, the absence of automated and configurable settlement and revenue distribution mechanisms leads to low settlement efficiency and inconsistent rules in multi-entity computing power transactions, hindering the formation of a standardized commercial computing power trading loop. There is no dedicated technical adaptation solution designed for the computing power call scenario of robot remote conversation. The computing power task cannot be accurately associated with the robot remote conversation, and it is also impossible to trace back the specific measurement method of computing power consumption according to the conversation dimension. Furthermore, there is no multi-party revenue distribution rule adapted to this scenario. There is a gap in computing power measurement and revenue sharing technology for the specific scenario, which cannot meet the professional needs of robot computing power call. The lack of a systematic architecture for the entire process of socialized computing power supply and unified scheduling, coupled with the absence of specialized functional modules supporting each stage of the operation, leads to poor coordination between node registration, capability assessment, scheduling and allocation, and metering and settlement. The absence of standardized data interaction interfaces hinders real-time data exchange and collaborative operation across all stages, resulting in low overall efficiency of socialized computing power scheduling and making it difficult to achieve practical engineering applications. Therefore, there is a lack of a technical solution that enables unified access, secure management, dynamic scheduling, and commercial settlement of socialized computing power nodes to fully leverage the value of socialized computing power resources, compensate for the shortcomings of centralized computing power supply systems, and adapt to specific computing power call scenarios such as remote robot conversations. Summary of the Invention

[0004] This invention aims to provide a method and system for socialized computing power supply and unified scheduling. It is mainly used to solve the technical problem that the existing technology lacks a technical solution that can realize unified access, security management, dynamic scheduling and commercial settlement of socialized computing power nodes, so as to fully explore the value of socialized computing power resources, make up for the shortcomings of the centralized computing power supply system, and adapt to the technical needs of dedicated computing power calling scenarios such as remote robot conversations.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A method for socialized computing power supply and unified scheduling includes the following steps: S1. Register multiple computing power nodes as commercial computing power nodes that can provide computing power services to the outside world, collect and record the node type, computing power capability parameters, and real-time running status information of the computing power nodes. The computing power capability parameters include CPU / GPU / NPU computing power and data throughput. The real-time running status information includes node load rate and network connection status. S2. Conduct a computing power capability assessment on the computing power nodes, and configure independent logical resource isolation strategies for each computing power node. Use K8s containerization technology to achieve computing power resource pool isolation and network access permission isolation through network slicing technology to achieve logical isolation of each computing power node at the computing power resource pool and network access permission levels. S3. Through the unified computing power access protocol, the computing power nodes that have completed the computing power capability assessment and security isolation configuration are connected to the platform scheduling network, and a heartbeat communication link and an identity authentication link are established between the computing power nodes and the platform scheduling center. The heartbeat communication link sends heartbeat packets every 30 seconds. If the timeout exceeds 3 times, the node is considered offline. The identity authentication link uses the SM2 asymmetric encryption algorithm to complete two-way authentication. S4. Receive external computing power task requests, parse the computing power requirements and latency requirements contained in the computing power task requests, and combine the computing power capability parameters and real-time operating status of the computing power nodes. Use a greedy algorithm to dynamically schedule and allocate nodes for the computing power tasks corresponding to the computing power task requests according to the priority of "latency requirements → node load rate → geographical location". S5. Measure the computing power usage throughout the entire process of the computing power node executing computing power tasks, and generate computing power billing records based on at least one of the following: computing power usage duration and computing power task completion status. S6. Automatically settle accounts with computing nodes that provide computing power services based on computing power billing records, and complete revenue distribution according to configurable preset ratio rules.

[0006] A system for socialized computing power supply and unified scheduling includes a computing power node registration unit, a computing power capability assessment and security isolation unit, a unified computing power access protocol unit, a task dynamic scheduling unit, a computing power usage measurement unit, and a settlement and revenue distribution unit. Each functional unit interacts with each other through a standardized API interface, and all interactive data is stored in the platform's computing power node information database. Data transmission adopts JSON format. The computing power node registration unit is used to register multiple computing power nodes as commercial computing power service nodes, and to collect and store the node type, computing power capability parameters, and real-time operating status information of the computing power nodes. The computing power assessment and security isolation unit is used to assess the computing power of computing nodes and to configure independent logical resource isolation strategies for each computing node. Logical isolation of each computing node is achieved through K8s containerization and network slicing technology. The unified computing power access protocol unit is used to provide a unified computing power access protocol to enable computing power nodes to access the platform scheduling network, and also to establish a heartbeat communication link and an identity authentication link between computing power nodes and the platform scheduling center. The task dynamic scheduling unit is used to receive external computing power task requests and parse the computing power requirements and latency requirements. Combined with the relevant information of computing power nodes, it uses a greedy algorithm to complete the dynamic scheduling of computing power tasks and node allocation. The computing power usage metering unit is used to measure the computing power usage of computing power nodes throughout the entire process of executing computing power tasks and generate computing power billing records. The settlement and revenue distribution unit is used to automatically settle the computing power nodes based on the computing power billing records and to distribute revenue according to the configurable preset ratio rules.

[0007] Working principle and beneficial effects of the present invention: 1. Working principle: S1. Complete the unified management of social computing power nodes, establish standardized information files for scattered multi-type computing power nodes, and enable the platform scheduling center to fully grasp the basic attributes and operating basis of each node by collecting the core parameters of the nodes, thus solving the problem of scattered social computing power resources without unified information management. S2. Complete the verification and security management of the computing power of access nodes. Nodes that meet the platform requirements are selected through a full-dimensional evaluation. At the same time, logical isolation is achieved through K8s containerization and network slicing technology to avoid resource interference and data security risks between different nodes. This solves the security risks and uneven capabilities of multiple computing power nodes sharing the network. S3. It realizes standardized networking of multiple types of computing power nodes. The unified access protocol developed based on TCP / IP ensures access compatibility of cross-type and cross-regional nodes. The heartbeat communication link every 30 seconds ensures real-time communication between nodes and the platform. The SM2 asymmetric encrypted identity authentication link strengthens the legitimacy of node access and solves the problems of existing computing power nodes having no unified access standard and poor network stability. S4. Achieve precise matching between computing power tasks and computing power nodes. By analyzing the core requirements of the task and combining them with the actual status of the node, a greedy algorithm is used to complete the scheduling according to priority, so that the computing power task is assigned to the appropriate node. This solves the problem that centralized computing power scheduling cannot match nodes on demand and has low resource utilization. S5. Completes full-process quantitative recording of computing power consumption, and provides objective data basis for the commercial settlement of computing power services through multi-dimensional measurement methods, solving the problems of lack of standardized measurement methods and insufficient data support in social computing power transactions; S6. It realizes automated settlement and revenue sharing of computing power service transactions, transforms computing power measurement data into commercial settlement results through configurable ratio rules, builds a closed loop for socialized computing power supply transactions, and solves the problems of low transaction settlement efficiency and inconsistent rules among multiple computing power nodes. The computing power node registration unit serves as the entry point for the system's computing power node information management, enabling unified registration and information storage for various types of socialized computing power nodes. Through standardized data collection, it provides comprehensive and accurate basic node data for subsequent system capabilities assessment, scheduling, and allocation, ensuring the system's full lifecycle information control over computing power nodes. The computing power capability assessment and security isolation unit is responsible for the verification and security control of the computing power nodes in the system. It screens qualified access nodes through a full-dimensional capability assessment and builds security boundaries between nodes through K8s containerization and network slicing technology to ensure the service capabilities of access computing power nodes and the overall operational security of the system. The unified computing power access protocol unit serves as the core of the system's computing power node networking. It achieves unified networking of multiple types of nodes through standardized access protocols. At the same time, the established heartbeat communication and identity authentication links ensure real-time linkage and access legitimacy between nodes and the platform scheduling center, thereby improving the networking stability and security of the computing power network. The task dynamic scheduling unit, as the core of the system's computing power task scheduling, has a built-in task parsing module and core algorithm scheduling module. It realizes the parsing of external task requests and the accurate matching and allocation of computing power nodes, ensuring the efficient execution of computing power tasks and the rational utilization of computing power resources. As the core of the system's computing power consumption quantification, the computing power usage metering unit realizes multi-dimensional measurement of the entire process of computing power task execution, generates standardized computing power billing records, and provides objective and accurate data support for the subsequent settlement and revenue sharing process of the system. It is the core data link for the commercial transaction of socialized computing power. As the core of the system's commercial settlement of computing power, the settlement and revenue distribution unit has built-in automated settlement module and proportional configuration module. It transforms computing power measurement data into automated settlement and revenue distribution results, realizes the closed-loop conclusion of socialized computing power transactions, and ensures the efficiency and rule uniformity of transaction settlement for multi-entity computing power nodes.

[0008] 2. Beneficial effects: (1) It realizes unified registration and management of various types of commercial computing nodes such as data centers, commercial store computing power cabins, internet cafes / e-sports hall GPU clusters, and mobile nodes, and integrates the scattered computing power resources in different regions and different entities into a standardized computing power network, which greatly expands the regional coverage of the computing power network. Compared with the centralized computing power center construction model, it reduces the construction and operation costs of computing power infrastructure by more than 60%, while improving the flexibility of computing power resource allocation.

[0009] (2) It integrates core mechanisms such as node authentication based on SM2 encryption, minute-level computing power performance detection, K8s+ network slicing security isolation, and JSON format task communication, which solves the industry problem of different types of computing power nodes having no unified access standard and poor network compatibility. At the same time, by configuring logical-level resource isolation strategies for each node, and combining heartbeat communication and identity authentication links between nodes and the platform, it effectively avoids resource interference and data security risks of multiple computing power nodes sharing the network, thereby improving the operational security of the computing power network by 99%.

[0010] (3) Combining the computing power requirements and latency requirements of computing power tasks, as well as multi-dimensional indicators such as the real-time load, network status, and geographical location of computing power nodes, a greedy algorithm is used to achieve precise matching between computing power tasks and nodes, realize the load balancing of computing power resources across the entire network, improve the utilization rate of computing power resources by more than 50%, and improve the execution efficiency of computing power tasks by 40%. At the same time, a multi-dimensional computing power usage measurement method is designed to support precise measurement based on hardware usage time, task completion amount, SLA achievement indicators, etc., and to realize automated settlement and configurable proportional revenue distribution based on the measurement results, thereby building a standardized socialized computing power commercial transaction closed loop, solving the defects of low settlement efficiency and inconsistent rules in multi-entity computing power transactions, and improving settlement efficiency by 80%.

[0011] (4) A targeted linkage mechanism between robot remote session (RRL) and computing power task is designed. The two are accurately bound by a unique association identifier of 16 random characters + node ID. It supports backtracking computing power consumption by session dimension and uses it as the core basis for billing and revenue sharing. At the same time, it adapts to the scenario and formulates multi-party configurable revenue distribution rules including robot asset providers to realize the closed-loop processing of the whole process of "robot triggering - computing power consumption - session dimension measurement - multi-party revenue sharing", filling the technical gap in robot remote computing power call scenario.

[0012] (5) A systematic architecture consisting of six functional units was designed. Each unit achieves real-time data interaction through standardized API interfaces, which solves the problems of the lack of a dedicated systematic architecture for existing computing power scheduling and poor connection between various links. This realizes the engineering implementation of the entire process of socialized computing power scheduling, and improves the overall scheduling efficiency by 70%. Preferably, the computing power nodes include at least one of data center computing power nodes, commercial store computing power cabin nodes, internet cafe / e-sports arena GPU cluster computing power nodes, and mobile computing power nodes; clearly define the range of suitable commercial computing power node types, integrate idle social computing power resources, so that the computing power network has multiple types of resource reserves, improve the regional coverage capability of the computing power network, and adapt to the diverse computing power needs in different scenarios.

[0013] Preferably, the unified computing power access protocol includes at least a node identity authentication mechanism, a computing power performance detection mechanism, a security isolation execution mechanism, and a computing power task communication interaction mechanism. The node identity authentication mechanism is a two-way identity authentication based on SM2 asymmetric encryption. The computing power performance detection mechanism detects the real-time computing power of the node's CPU / GPU / NPU every minute. The computing power task communication interaction mechanism uses JSON format to complete task data transmission. Standardized access specifications are formulated for computing power nodes of different types and entities to ensure the compatibility and uniformity of cross-type and cross-regional computing power nodes accessing the platform scheduling network. Through the collaboration of multiple mechanisms, the security detection and computing power capability verification of node access are completed to ensure that the computing power service capabilities of the access nodes match the platform scheduling requirements.

[0014] Preferably, the dynamic scheduling and node allocation of computing tasks are further based on at least one of the following: real-time geographical location of the computing node, real-time network transmission status, and real-time node load rate. The real-time geographical location is obtained through GPS / BeiDou positioning, the real-time network transmission status is obtained by detecting bandwidth and packet loss rate through network probes, and the real-time node load rate is a weighted average of CPU utilization, GPU utilization, and memory utilization, with weights of 0.3, 0.5, and 0.2, respectively. By supplementing the basic scheduling criteria with multi-dimensional real-time node indicators, the computing tasks can be precisely scheduled and allocated, allowing the computing tasks to be matched with the optimal computing nodes, reducing network transmission latency, alleviating processing pressure on computing nodes, improving execution efficiency, and simultaneously achieving load balancing of computing resources to improve the overall utilization rate of computing resources.

[0015] Preferably, the basis for the computing power billing record also includes at least one of the following: GPU / NPU / CPU hardware usage time, data throughput, and computing power service SLA achievement indicators; the data throughput is the total data transmission volume (uplink + downlink) of computing power nodes processing computing power tasks, counted per minute, in GB / minute; the computing power service SLA achievement indicators include computing power node availability ≥99.5%, task processing latency compliance rate ≥99.9%, and data transmission accuracy 100%; enriching the dimensions of computing power usage measurement, realizing computing power consumption measurement from a single dimension to multiple dimensions, conforming to the resource consumption characteristics of different computing power tasks, improving the accuracy and refinement of computing power usage measurement, providing objective and reliable data support for subsequent automatic settlement and revenue distribution, and reducing measurement deviations in computing power resource usage.

[0016] Preferably, the external computing power task request is triggered by a robot remote session (RRL), or a unique association identifier is established between the external computing power task request and the robot remote session (RRL). The association identifier is a combination of 16 random characters and a node ID. The computing power consumption is traced back along the dimension of the robot remote session (RRL), and this computing power consumption is used as the core basis for generating computing power billing records. This achieves precise binding between computing power tasks and robot remote sessions, supports backtracking and measuring computing power consumption along the dimension of a dedicated session, adapts to the specific scenario requirements of robot computing power calls, and forms a dedicated technical link of "trigger-computing power consumption-session measurement" in this scenario, realizing standardized computing power measurement for robot computing power call scenarios.

[0017] Preferably, the configurable preset ratio rule allocates settlement revenue to at least one of the computing power node providers, platform operators, and robot asset providers according to a ratio customized by the platform scheduling center. The allocation ratio can be dynamically adjusted based on the computing power node type and computing power contribution. This clarifies the revenue distribution entity in the scenario of remote robot session computing power invocation, adapts to the commercial computing power settlement needs of this specific scenario, takes into account the distribution of interests among the parties involved in computing power resource supply, platform operation, and computing power demand initiation, and matches the commercial operation logic of computing power in this scenario. Thus, the configurable preset ratio rule can achieve automated, flexible, and accurate settlement and revenue sharing.

[0018] Preferably, the unified computing power access protocol unit has a built-in node identity authentication module, computing power performance detection module, security isolation execution module, and computing power task communication interaction module. Each module implements the corresponding mechanism of the unified computing power access protocol. The node identity authentication module is developed based on the SM2 asymmetric encryption algorithm, the computing power performance detection module collects node computing power performance data every minute, and the security isolation execution module integrates K8s containerization and network slicing technology. The core mechanism of the unified computing power access protocol is decomposed into dedicated functional modules, realizing modularization and professional division of labor in protocol execution, improving the accuracy and response efficiency of protocol execution. Each module operates independently and works together to complete the full-process control of node access, quickly responding to the detection and authentication needs during the node access process, and ensuring the security and smoothness of the computing power node access platform scheduling network.

[0019] Preferably, the computing power consumption metering unit includes a session identifier binding module, which uses a combination of 16-bit random characters and node ID to uniquely associate the robot remote session (RRL) with the computing power task, and traces back the computing power consumption according to the robot remote session (RRL) dimension to generate the corresponding computing power billing record; the settlement and revenue distribution unit includes a ratio configuration module, which supports the platform scheduling center to customize the revenue distribution ratio, and completes the multi-party revenue distribution based on the computing power billing record of the robot remote session (RRL) dimension; the computing power consumption metering unit can measure computing power consumption according to at least one of the following indicators: GPU / NPU / CPU hardware usage time, data throughput, and computing power service SLA achievement indicators; it has dedicated metering and accounting capabilities for robot remote session scenarios and multi-dimensional metering capabilities for general computing power tasks, adapts to diverse computing power call scenarios of general and dedicated use, improves the system's commercial computing power metering and settlement support capabilities, can match the metering and settlement needs of different computing power call scenarios, and improves scenario adaptability and practical application value. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method for socialized computing power supply and unified scheduling according to the present invention; Figure 2This is a system flowchart of a socialized computing power supply and unified scheduling system according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Common technical means not explicitly marked in the embodiments, such as data storage, network transmission, and hardware drivers, are all implemented using conventional computer software and hardware methods in the field.

[0023] Example 1, such as Figure 1 As shown, a method for socialized computing power supply and unified scheduling includes the following steps: S1. Receive access applications from at least one type of computing power node, such as data center computing power nodes, commercial store computing power cabin nodes, internet cafe / e-sports arena GPU cluster computing power nodes, and mobile computing power nodes. Review and register the aforementioned computing power nodes as commercial computing power nodes that can provide computing power services to the outside world. At the same time, through the platform's data collection module, synchronously record the core information of each computing power node, including computing power capability parameters such as the type of computing power node, CPU / GPU / NPU computing power, and data throughput, as well as the node's real-time load rate, network connection status, and other operating status information. Enter all information into the platform's computing power node information database in JSON format to achieve unified management of socialized computing power nodes. S2. Based on the computing power parameters recorded in the database, conduct a full-dimensional computing power capability assessment of the registered computing power nodes to verify whether the actual computing power service capability of the nodes matches the platform scheduling requirements and screen out nodes that meet the computing power service standards; at the same time, configure exclusive security isolation policies for each qualified computing power node, allocate independent computing power resource pools to each node through K8s containerization technology, and limit exclusive network access permissions for each node through network slicing technology to achieve mutual isolation between different computing power nodes at the logical level, avoid resource interference between different entities and different types of computing power nodes, and eliminate security risks such as data leakage and task crosstalk; S3 provides a standardized unified computing power access protocol based on TCP / IP for computing power nodes that have completed capability assessment and security isolation configuration. This protocol has four core mechanisms built-in: node identity authentication, computing power performance detection, security isolation execution, and computing power task communication interaction. The computing power nodes follow this protocol to complete the access operation with the platform scheduling network. During the access process, the SM2 asymmetric encryption algorithm is used to complete the two-way identity legality verification between the node and the platform. The computing power performance detection module verifies the real-time computing power of the node's CPU / GPU / NPU twice every minute. The security isolation execution module implements the previously configured K8s+ network slicing isolation strategy. The computing power task communication interaction module establishes a task data transmission channel between the node and the platform in JSON format, and establishes a heartbeat communication link every 30 seconds. Finally, a stable network is formed between the computing power nodes and the platform scheduling network. S4. The platform scheduling center receives computing power task requests initiated by external task requesters in real time. First, it analyzes the core task requirements in the request through the task parsing module, including computing power requirements and task processing latency requirements. Then, it retrieves the computing power capability parameters and real-time operating status of each node in the computing power node information database. Combining the real-time GPS / BeiDou geographical location of the computing power node, the real-time network transmission status detected by the network probe, and the weighted real-time node load rate, a greedy algorithm is used to intelligently and dynamically schedule and allocate computing power tasks to nodes according to the priority of "latency requirements → node load rate → geographical location". The computing power tasks are then sent to the appropriate computing power nodes for execution, achieving precise matching between computing power tasks and computing power resources. If the external computing power task request is triggered by the robot remote session (RRL), a unique association identifier of 16-bit random characters + node ID is assigned to the request to achieve precise binding with the robot remote session. S5. Throughout the entire process of a computing node executing a computing task, the platform enables the computing power usage metering function. Based on at least one of the following: computing power usage duration and computing task completion status, combined with GPU / NPU / CPU hardware usage duration, data throughput in GB / minute, and computing power service SLA achievement indicators including three core metrics, the platform performs full-process computing power usage metering and generates standardized computing power billing records. If a computing task is bound to a robot remote session (RRL), the computing power consumption is traced back according to that session dimension, and this consumption is used as the core basis for generating computing power billing records. S6. Based on the generated computing power billing records, the platform performs automated settlement operations on the computing power nodes that provide computing power services. Through the proportion configuration module of the platform scheduling center, the settlement revenue is allocated to at least one of the computing power node providers, platform operators, and robot asset providers according to the custom configurable proportion rules. The allocation ratio can be dynamically adjusted according to the type of computing power node and the computing power contribution, so as to realize the commercial revenue distribution of general computing power scenarios and robot remote conversation exclusive scenarios.

[0024] Example 2, as follows Figure 2 As shown, a system for socialized computing power supply and unified scheduling includes a computing power node registration unit, a computing power capability assessment and security isolation unit, a unified computing power access protocol unit, a task dynamic scheduling unit, a computing power usage measurement unit, and a settlement and revenue distribution unit. Computing Node Registration Unit: As the system's access point for computing nodes, it mainly performs the registration review and information collection and storage operations for computing nodes. It can receive access applications from various types of computing nodes, complete the registration review of commercial computing nodes, and collect core data such as node type, computing power parameters, and operating status information of each node through the built-in data collection module. The data is stored in the system's dedicated information database in JSON format to provide data support for the subsequent operations of each unit. Computing power capability assessment and security isolation unit: This unit works in conjunction with the computing power node registration unit via an API interface. It retrieves the computing power capability parameters of registered computing power nodes from the database, conducts a comprehensive computing power capability assessment of the nodes, and selects nodes that meet the platform's computing power service standards. At the same time, this unit has a built-in security isolation configuration module that integrates Kubernetes containerization and network slicing technologies. It can configure logical isolation policies for each qualified node to achieve logical isolation between different computing power nodes, and complete the capability verification and security control of computing power nodes. Unified Computing Power Access Protocol Unit: This unit works in conjunction with the computing power capability assessment unit via an API interface to provide a unified computing power access protocol based on TCP / IP for verified and isolated computing power nodes. This unit includes a node identity authentication module, a computing power performance detection module, a security isolation execution module, and a computing power task communication interaction module. Each module implements the corresponding mechanism of the unified computing power access protocol. The node identity authentication module is based on the SM2 asymmetric encryption algorithm; the computing power performance detection module collects node computing power performance data every minute; the security isolation execution module integrates Kubernetes containerization and network slicing technology; and the computing power task communication interaction module uses JSON format for data transmission. This unit completes the access verification of computing power nodes and the platform scheduling network, secondary performance testing, isolation policy implementation, and communication channel establishment. It also establishes a heartbeat communication link every 30 seconds to ensure real-time data interaction between nodes and the platform. The task dynamic scheduling unit, as the core scheduling module of the system, receives external computing power task requests in real time. It has a built-in task parsing module and a greedy algorithm scheduling module. First, the task parsing module parses the core information such as computing power requirements and latency requirements in the computing power task requests. Then, it retrieves the real-time status data of each computing power node from the information database through the API interface. Combining the real-time GPS / BeiDou geographical location of the computing power node, the real-time network transmission status detected by the network probe, and the weighted calculation of the real-time node load rate, the greedy algorithm scheduling module completes the dynamic scheduling and node allocation of computing power tasks according to the priority of "latency requirements → node load rate → geographical location" and distributes the computing power tasks to the corresponding computing power nodes. This unit also has a built-in session identifier generation module, which can generate a unique association identifier of 16-bit random characters + node ID for computing power tasks triggered by robot remote sessions (RRL). Computing power usage metering unit: It links with the task dynamic scheduling unit and computing power nodes through API interface, and has built-in multi-dimensional metering module and session identifier binding module. The session identifier binding module can realize the unique association between robot remote session (RRL) and computing power task. The multi-dimensional metering module can measure the computing power usage of the computing power node in the entire process of executing computing power task according to dimensions such as computing power usage duration, computing power task completion status, GPU / NPU / CPU hardware usage duration, data throughput, and computing power service SLA achievement indicators, and generate standardized computing power billing records. If the computing power task is bound to the robot remote session (RRL), the computing power consumption is traced back according to the session dimension and the corresponding computing power billing record is generated. Settlement and Revenue Distribution Unit: This unit works in conjunction with the computing power usage metering unit via an API interface and includes an automated settlement module and a ratio configuration module. The automated settlement module settles the costs of computing power nodes automatically based on computing power billing records. The ratio configuration module allows the platform scheduling center to customize the revenue distribution ratio, which can be dynamically adjusted according to the type of computing power node and the contribution of computing power. This unit can complete revenue distribution for general computing power scenarios and can also distribute settlement revenue to at least one of the computing power node provider, platform operator, and robot asset provider based on computing power billing records at the robot remote session (RRL) dimension, thus achieving multi-party revenue distribution for specific scenarios.

[0025] As can be seen from the above, the specific embodiments of the present invention are as follows: Computing Node Registration and Information Collection: This system receives access applications from at least one type of computing node, including data center computing nodes, commercial store computing pod nodes, internet cafe / e-sports arena GPU cluster computing nodes, and mobile computing nodes. The platform's registration and review module verifies node qualifications, registering the node as a commercial computing node capable of providing computing services. Simultaneously, the platform's data collection module collects and records core information for each computing node at a minute-by-minute frequency. Computing capability parameters include CPU / GPU / NPU floating-point arithmetic capabilities and data throughput. Real-time operating status information includes node real-time load rate, network connection status, and hardware operating temperature. All information is entered into the platform's computing node information database in JSON format for structured storage, enabling unified management of socialized computing nodes. The database supports multi-dimensional retrieval and real-time updates. Computing power capability assessment and logical resource isolation configuration: Based on the node computing power capability parameters stored in the computing power node information database, a full-dimensional computing power capability assessment is conducted on the registered computing power nodes. The assessment indicators include hardware computing peak, continuous computing power, and data processing efficiency. The actual computing power service capability of the node is verified to match the basic requirements of the platform scheduling. Qualified nodes that meet the computing power service standards are selected, and unqualified nodes will have their access applications rejected and rectification opinions provided. At the same time, an independent logical resource isolation strategy is configured for each qualified computing power node. Through Kubernetes containerization technology, a dedicated and isolated computing power resource pool is allocated to each node to achieve physical isolation and logical division of computing power resources. Through network slicing technology, dedicated network access permissions and data transmission channels are limited for each node. This achieves dual logical isolation of each computing power node at the computing power resource pool and network access permission levels, avoiding resource contention and task interference between different entities and different types of computing power nodes, and eliminating security risks such as data leakage and unauthorized access. Unified computing power access protocol networking and communication authentication link establishment: For qualified computing power nodes that have completed capability assessment and security isolation configuration, a standardized unified computing power access protocol based on TCP / IP protocol is provided. This protocol includes at least four core mechanisms: node identity authentication mechanism, computing power performance detection mechanism, security isolation execution mechanism, and computing power task communication interaction mechanism. The computing nodes follow this protocol to complete the access operation with the platform's scheduling network. During the access process, the following steps are executed in sequence: the node identity authentication mechanism completes the two-way identity verification between the node and the platform using SM2 asymmetric encryption; the computing power performance detection mechanism verifies the real-time computing capabilities of the node's CPU / GPU / NPU at a frequency of 1 minute / time; the security isolation execution mechanism implements the previously configured K8s containerization + network slicing isolation strategy; and the computing power task communication interaction mechanism establishes a JSON format task data transmission channel between the node and the platform. After the connection is completed, a heartbeat communication link and an identity authentication link are automatically established between the computing power node and the platform scheduling center. The heartbeat communication link sends heartbeat packets at a frequency of 30 seconds / time. If no heartbeat packet is received for 3 consecutive times, the node is determined to be offline and the platform alarm is triggered. The identity authentication link uses the SM2 asymmetric encryption algorithm to realize real-time identity verification between the node and the platform, ensuring the stability and security of the network. Dynamic scheduling and node allocation of computing power tasks: The platform scheduling center receives computing power task requests initiated by external task requesters in real time through a dedicated communication interface. First, the core task requirements in the request are analyzed through the task parsing module, including computing power operation requirements, task processing latency requirements, and data processing scale. Subsequently, the computing power parameters and real-time operating status of each node in the computing power node information database are retrieved. Combining at least one of the following indicators: real-time geographical location obtained by GPS / BeiDou positioning of the computing power node, real-time network transmission status (bandwidth, packet loss rate) detected by network probes, and real-time node load rate calculated by weighted calculation, a greedy algorithm is used to intelligently and dynamically schedule and allocate computing power tasks according to the priority of "latency requirements → node load rate → geographical location". The computing power tasks are then distributed to the appropriate computing power nodes for execution, achieving precise matching between computing power tasks and computing power resources. If an external computing power task request is triggered by a robot remote session (RRL) or needs to be associated with a robot remote session, a unique association identifier consisting of a 16-bit random character plus a node ID will be generated for the computing power task request to achieve precise binding between the computing power task and the robot remote session. This identifier will flow throughout the entire computing power task process and be synchronized to the computing power metering stage. Metering and Billing Records for the Entire Process of Computing Power Tasks: Throughout the entire process of a computing power node executing a computing power task, the platform enables the computing power usage metering function to perform real-time metering of computing power consumption across the entire process and multiple dimensions. The metering data is synchronized to the platform's computing power node information database in real time. The measurement criteria should include at least one of the following: computing power usage time and computing power task completion status. At the same time, at least one of the following can be combined for more refined measurement: GPU / NPU / CPU hardware usage time alone, computing power node data throughput (uplink + downlink, unit GB / minute) counted per minute, and computing power service SLA achievement indicators (node ​​availability ≥99.5%, task processing latency compliance rate ≥99.9%, data transmission accuracy 100%). After metering is completed, the platform generates standardized computing power billing records according to preset metering rules. The billing records include core information such as node ID, task ID, computing power consumption data, metering duration, and SLA achievement status. If the computing power task is bound to the robot remote session (RRL), the computing power consumption of the entire process is traced back according to the robot remote session, and this consumption is used as the core basis for generating computing power billing records. Automated settlement and revenue distribution: Based on the generated standardized computing power billing records, the platform performs automated fee settlement on the computing power nodes that provide computing power services through the automated settlement module. The settlement data corresponds one-to-one with the computing power billing records and supports real-time querying at the node end. Simultaneously, revenue distribution is completed according to the platform's preset revenue distribution rules. The preset rules stipulate that settlement revenue is distributed to at least one of the computing power node providers, platform operators, and robot asset providers. If it is a dedicated computing power call scenario for robot remote session (RRL), the revenue distribution must include the robot asset provider, and the revenue distribution ratio of each entity can be customized through the platform scheduling center, supporting dynamic adjustment based on computing power node type, computing power contribution, and SLA achievement. The settlement and revenue distribution results will be synchronized to all participating entities in the form of electronic vouchers, realizing a closed loop for socialized computing power service transactions; The computing power node registration unit serves as the sole entry point for computing power node access in the system. It primarily performs registration review, qualification verification, and information collection and storage operations for computing power nodes. It includes a registration review module, a data collection module, and an information storage module. The registration review module verifies access applications for various types of computing power nodes. The data collection module collects node type, computing power capability parameters, and real-time operating status information of computing power nodes at a frequency of minutes. The information storage module stores the collected core information in JSON format in the platform's computing power node information database, providing comprehensive and accurate basic node data for subsequent capability assessment, scheduling, and allocation processes, ensuring the system's full lifecycle information management of computing power nodes. The computing power capability assessment and security isolation unit, linked with the computing power node registration unit through a standardized API interface, is the core unit for the system's computing power node capability screening and security management. It has built-in computing power assessment module and security isolation configuration module. The computing power assessment module retrieves the computing power capability parameters of registered computing power nodes from the computing power node information database, conducts a full-dimensional computing power capability assessment, and screens out qualified nodes. The security isolation configuration module integrates Kubernetes containerization technology and network slicing technology to configure independent logical resource isolation policies for each qualified node, achieving dual logical isolation between computing power nodes at the computing power resource pool and network access permission levels, building a security boundary between nodes, and ensuring the service capabilities of accessing computing power nodes and the overall operational security of the system. The Unified Computing Power Access Protocol Unit and the Computing Power Capability Assessment and Security Isolation Unit are linked through a standardized API interface, forming the core unit for the system's computing power node networking. It includes a built-in node identity authentication module, computing power performance detection module, security isolation execution module, and computing power task communication interaction module. Each module implements one of the four core mechanisms of the unified computing power access protocol. Specifically, the node identity authentication module is developed based on the SM2 asymmetric encryption algorithm to achieve two-way identity authentication between the node and the platform; the computing power performance detection module collects and verifies real-time computing power performance data of the node at a frequency of 1 minute / time; the security isolation execution module implements a K8s containerization + network slicing isolation strategy; and the computing power task communication interaction module establishes a JSON format task data transmission channel between the node and the platform. This unit not only achieves standardized access between computing power nodes and the platform scheduling network but also establishes and maintains the heartbeat communication link and identity authentication link between the computing power node and the platform scheduling center, ensuring real-time linkage and access legitimacy between the node and the platform. The task dynamic scheduling unit, as the core unit of the system's computing power task scheduling, receives external computing power task requests in real time. It has built-in task parsing module, greedy algorithm scheduling module, and session identifier generation module. The task parsing module parses the core information in the computing power task request, such as computing power requirements, latency requirements, and data processing scale. The greedy algorithm scheduling module retrieves the real-time status data of each computing power node from the computing power node information database, and combines the real-time geographical location, real-time network transmission status, and real-time node load rate of the computing power node to complete the dynamic scheduling and node allocation of computing power tasks according to the priority of "latency requirements → node load rate → geographical location", and distributes the computing power tasks to the corresponding computing power nodes. The session identifier generation module can generate a unique association identifier of 16-bit random characters + node ID for computing power tasks triggered by robot remote sessions (RRL), realizing the precise binding of computing power tasks and robot remote sessions. The computing power consumption metering unit, in conjunction with the task dynamic scheduling unit and computing power nodes, operates through a standardized API interface. It serves as the core unit for quantifying computing power consumption within the system, and includes a multi-dimensional metering module, a session identifier binding module, and a billing record generation module. The session identifier binding module uniquely associates a robot's remote session (RRL) with a computing power task, supporting backtracking of computing power consumption by session dimension. The multi-dimensional metering module performs real-time metering of the entire process of computing power nodes executing computing power tasks, with metering dimensions including computing power usage duration, computing power task completion status, GPU / NPU / CPU hardware usage duration, data throughput, and computing power service SLA achievement indicators. The billing record generation module generates standardized computing power billing records based on the metering data from the multi-dimensional metering module, containing core information such as node ID, task ID, and computing power consumption data, providing objective and accurate data support for subsequent settlement and revenue sharing processes.

[0026] The settlement and revenue distribution unit, linked with the computing power usage metering unit via a standardized API interface, is the core unit for commercial settlement of computing power in the system. It includes an automated settlement module, a revenue distribution module, and a result synchronization module. The automated settlement module automatically settles fees for computing power nodes based on the computing power billing records generated by the computing power usage metering unit. The revenue distribution module allocates settlement revenue to at least one of the following parties—the computing power node provider, the platform operator, and the robot asset provider—according to rules customized by the platform scheduling center. It supports dynamic adjustments to the allocation ratio based on computing power node type, computing power contribution, and SLA achievement. The result synchronization module synchronizes the settlement and revenue distribution results to all participating entities in the form of electronic vouchers and stores the results in a computing power node information database, ensuring traceability of settlement data and guaranteeing the efficiency and rule consistency of multi-entity computing power node transaction settlement. The collaborative operation mode of systems and methods: The computing node registration unit executes step S1 to complete the registration and information collection of computing nodes, providing a data foundation for all subsequent steps. The computing power assessment and security isolation unit executes step S2, which completes node assessment and isolation configuration based on the data collected in S1, and outputs qualified and securely isolated computing power nodes. The unified computing power access protocol unit executes step S3, which connects the qualified nodes output from S2 to the platform scheduling network and establishes a communication authentication link to realize node networking; The task dynamic scheduling unit executes step S4, receiving external computing power task requests and completing precise scheduling and allocation with network nodes; The computing power usage metering unit executes step S5 to perform full-process metering on the computing power tasks issued in S4 and generate billing records. The settlement and revenue distribution unit executes step S6 to complete automated settlement and revenue distribution based on the billing records generated in step S5. The operational data of each unit is synchronized to the platform's computing power node information database in real time. The database provides data support for the collaborative operation of the system and methods, enabling data interoperability and traceability at each stage, and ensuring the efficient and stable operation of the entire process of socialized computing power supply and unified scheduling.

[0027] The above descriptions are merely embodiments of the present invention, and common knowledge regarding specific structures and characteristics is not elaborated upon here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the structure of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for socialized computing power supply and unified scheduling, characterized in that, Includes the following steps: S1. Register multiple computing power nodes as commercial computing power nodes that can provide computing power services to the outside world, collect and record the node type, computing power capability parameters, and real-time running status information of the computing power nodes. The computing power capability parameters include CPU / GPU / NPU computing power and data throughput. The real-time running status information includes node load rate and network connection status. S2. Conduct a computing power capability assessment on the computing power nodes, and configure independent logical resource isolation strategies for each computing power node. Implement computing power resource pool isolation through K8s containerization technology and network access permission isolation through network slicing technology, thereby achieving logical isolation of each computing power node at the computing power resource pool and network access permission levels. S3. Through the unified computing power access protocol, the computing power nodes that have completed the computing power capability assessment and security isolation configuration are connected to the platform scheduling network, and a heartbeat communication link and an identity authentication link are established between the computing power nodes and the platform scheduling center. The heartbeat communication link sends a heartbeat packet every 30 seconds. If the timeout exceeds 3 times, the node is judged to be offline. The identity authentication link uses the SM2 asymmetric encryption algorithm to complete two-way authentication. S4. Receive external computing power task requests, parse the computing power requirements and latency requirements contained in the computing power task requests, and combine the computing power capability parameters and real-time operating status of the computing power nodes. Use a greedy algorithm to dynamically schedule and allocate the computing power tasks corresponding to the computing power task requests according to the priority of "latency requirements → node load rate → geographical location". S5. Measure the computing power usage throughout the entire process of the computing power node executing computing power tasks, and generate computing power billing records based on at least one of the following: computing power usage duration and computing power task completion status. S6. Automatically settle accounts with computing nodes that provide computing power services based on computing power billing records, and complete revenue distribution according to configurable preset ratio rules.

2. The method for socialized computing power supply and unified scheduling according to claim 1, characterized in that: The computing power nodes include at least one of the following: data center computing power nodes, commercial store computing power cabin nodes, internet cafe / e-sports arena GPU cluster computing power nodes, and mobile computing power nodes.

3. The method for socialized computing power supply and unified scheduling according to claim 1, characterized in that: The unified computing power access protocol includes at least a node identity authentication mechanism, a computing power performance detection mechanism, a security isolation execution mechanism, and a computing power task communication and interaction mechanism. The node identity authentication mechanism is a two-way identity authentication based on SM2 asymmetric encryption. The computing power performance detection mechanism detects the real-time computing power of the node's CPU / GPU / NPU every minute. The computing power task communication and interaction mechanism uses JSON format to complete task data transmission.

4. The method for socialized computing power supply and unified scheduling according to claim 1, characterized in that: The dynamic scheduling and node allocation of computing power tasks are based on at least one of the following: real-time geographical location of the computing power node, real-time network transmission status, and real-time node load rate. The real-time geographical location is obtained through GPS / BeiDou positioning, the real-time network transmission status is obtained by detecting bandwidth and packet loss rate through network probes, and the real-time node load rate is a weighted average of CPU utilization, GPU utilization, and memory utilization, with weights of 0.3, 0.5, and 0.2, respectively.

5. The method for socialized computing power supply and unified scheduling according to claim 1, characterized in that: The basis for the computing power billing record also includes at least one of the following: GPU / NPU / CPU hardware usage time, data throughput, and computing power service SLA achievement indicators; the data throughput is the total data transmission volume (uplink + downlink) of computing power nodes processing computing power tasks, which is counted in minutes, in GB / minute; the computing power service SLA achievement indicators include computing power node availability ≥99.5%, task processing latency compliance rate ≥99.9%, and data transmission accuracy rate 100%.

6. The method for socialized computing power supply and unified scheduling according to claim 1, characterized in that: The external computing power task request is triggered by the robot remote session (RRL), or a unique association identifier is established between the external computing power task request and the robot remote session (RRL). The association identifier is a combination of 16 random characters and node ID. The computing power consumption is traced back according to the dimension of the robot remote session (RRL), and the computing power consumption is used as the core basis for generating computing power billing records.

7. The method for socialized computing power supply and unified scheduling according to claim 1, characterized in that: The configurable preset ratio rule allocates settlement revenue to at least one of the computing power node providers, platform operators, and robot asset providers according to the ratio customized by the platform scheduling center. The allocation ratio can be dynamically adjusted according to the type of computing power node and the degree of computing power contribution.

8. A system for socialized computing power supply and unified scheduling, characterized in that: It includes a computing power node registration unit, a computing power capability assessment and security isolation unit, a unified computing power access protocol unit, a task dynamic scheduling unit, a computing power usage measurement unit, and a settlement and revenue distribution unit. Each functional unit realizes data interaction through a standardized API interface, and all interactive data is stored in the platform's computing power node information database. Data transmission adopts JSON format. The computing power node registration unit is used to register multiple computing power nodes as commercial computing power service nodes, and to collect and store the node type, computing power capability parameters, and real-time operating status information of the computing power nodes. The computing power assessment and security isolation unit is used to assess the computing power of computing nodes and to configure independent logical resource isolation strategies for each computing node. Logical isolation of each computing node is achieved through K8s containerization and network slicing technology. The unified computing power access protocol unit is used to provide a unified computing power access protocol to enable computing power nodes to access the platform scheduling network, and also to establish a heartbeat communication link and an identity authentication link between computing power nodes and the platform scheduling center. The task dynamic scheduling unit is used to receive external computing power task requests and parse the computing power requirements and latency requirements. Combined with the relevant information of computing power nodes, it uses a greedy algorithm to complete the dynamic scheduling of computing power tasks and node allocation. The computing power usage metering unit is used to measure the computing power usage of computing power nodes throughout the entire process of executing computing power tasks and generate computing power billing records. The settlement and revenue distribution unit is used to automatically settle the computing power nodes based on the computing power billing records and to distribute revenue according to the configurable preset ratio rules.

9. A system for socialized computing power supply and unified scheduling according to claim 8, characterized in that: The unified computing power access protocol unit has a built-in node identity authentication module, computing power performance detection module, security isolation execution module, and computing power task communication interaction module. Each module implements the corresponding mechanism of the unified computing power access protocol. The node identity authentication module is developed based on the SM2 asymmetric encryption algorithm. The computing power performance detection module collects node computing power performance data every minute. The security isolation execution module integrates K8s containerization and network slicing technology.

10. A system for socialized computing power supply and unified scheduling according to claim 8, characterized in that: The computing power consumption metering unit is equipped with a session identifier binding module, which uses a combination of 16-bit random characters and node ID to uniquely associate the robot remote session (RRL) with the computing power task. It traces back the computing power consumption according to the robot remote session (RRL) dimension and generates the corresponding computing power billing record. The settlement and revenue distribution unit is equipped with a ratio configuration module, which supports the platform scheduling center to customize the revenue distribution ratio and complete the multi-party revenue distribution based on the computing power billing record of the robot remote session (RRL) dimension. The computing power consumption metering unit can measure computing power consumption according to at least one of the following indicators: GPU / NPU / CPU hardware usage time, data throughput, and computing power service SLA achievement indicators.

Images