A connector system with trusted execution and resource scheduling capabilities

Abstract

This invention discloses a connector system with trusted execution and resource scheduling capabilities, belonging to the field of data management technology. It solves the problem that existing methods rely solely on information reported by device plugins during initial access to their node resource capability lists, failing to perceive dynamic resource changes in real time. Furthermore, in scenarios with fluctuating resource status, scheduling decisions based on outdated resource information lead to unreasonable task allocation. The system includes a unified access module, a scheduling response module, a trusted execution module, and a log auditing and monitoring module. The trusted execution module initializes the trusted execution environment based on task scheduling requests and distributes dynamic resource packages to computing nodes based on their real-time status. This enables dynamic resource perception and real-time status simulation by combining the real-time status of computing nodes, reducing scheduling failures or performance losses caused by information asymmetry.

Description

Technical Field

[0001] This invention belongs to the field of data management technology, specifically relating to a connector system with trusted execution and resource scheduling capabilities. Background Technology

[0002] As the process of data elementization progresses, the demand for combined applications of data, computing power, and algorithms is growing. Currently, frameworks such as the International Data Space Architecture (IDSA), GAIA-X, and DataMesh have proposed "connector" components to support data flow and integration. In distributed computing, cloud computing, and cross-domain collaboration scenarios, "connectors" serve as key middleware connecting different computing nodes and heterogeneous resources, undertaking the core functions of data interaction, task scheduling, and resource coordination.

[0003] Traditional connector systems primarily focus on basic connectivity and simple routing logic, with "connectivity" as their core design goal. However, as the complexity of digital businesses increases, traditional connectors have gradually revealed the following key shortcomings: Their resource allocation relies on static configuration or simple heuristic rules, making it difficult to dynamically adapt to complex and ever-changing business needs. Furthermore, existing technologies in this field often focus on single data flow stages, lacking unified support for dynamic loading of algorithm resources, elastic scheduling of computing power, and diverse computing modes (such as federated learning and multi-party secure computation).

[0004] Chinese patent CN112559138B discloses a resource scheduling system and method. The method includes a device plugin providing access information to a device manager through a device plugin extension interface; the device manager generating a node resource capability list based on the device plugin's access information and transmitting the node resource capability list to a node capability description file; and the node capability description file deploying the device plugin to a node specified by the server according to the server's service scheduling request. However, the existing method's node resource capability list only relies on the information reported by the device plugin during initial access, and cannot perceive the dynamic changes of resources in real time. In scenarios where resource status fluctuates in real time, scheduling decisions based on outdated resource information lead to unreasonable task allocation. To address the above problems, we propose a connector system with reliable execution and resource scheduling capabilities. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a connector system with reliable execution and resource scheduling capabilities. This solves the problem that existing methods rely solely on information reported by the device plugin during initial access to the node resource capability list, which fails to detect dynamic changes in resources in real time. Furthermore, in scenarios where resource status fluctuates in real time, scheduling decisions based on outdated resource information lead to unreasonable task allocation.

[0006] This invention is implemented as follows: a connector system with trusted execution and resource scheduling capabilities, the system comprising: The unified access module responds to the access command of computing resource elements, registers and identifies computing resource elements based on their types, dynamically configures element matching templates based on the element identification results, and writes the computing resource elements into the element matching templates. The scheduling response module is used to receive and parse task scheduling requests sent by multiple protocols, perform joint matching between task scheduling requests and computing resource elements, bind computing resource elements that match the task scheduling requests, index the element matching templates corresponding to the computing resource elements, and package them into dynamic resource packages. The trusted execution module initializes the trusted execution environment based on the task scheduling request, obtains dynamic resource packages, and distributes the dynamic resource packages to the computing nodes in combination with the real-time status of the computing nodes. The computing nodes generate resource scheduling requests in real time, and respond to the resource scheduling requests by driving the computing nodes to allocate the scheduling data stream to the corresponding data transmission links based on the load balancing strategy. The log auditing and monitoring module is used to record and audit the entire lifecycle of scheduling data streams, data transmission links, and computing power operation behaviors in compliance.

[0007] Preferably, the unified access module includes: The element identification unit responds to the computing resource element access command, identifies the type of computing resource element, and assigns the computing resource element to the corresponding associated unit; The data management unit is used to acquire multi-source heterogeneous data resources, register the data resources, and define the data resource format, schema, access interface, and security policy. The algorithm management unit is used to acquire various algorithm components and to register, load, version control, and configure access permissions for these components. The computing power management unit is used to access computing power nodes from local servers, cloud hosts, and edge nodes, and to register the performance indicators, status, and access paths of the computing power nodes. The computation mode management unit is used to acquire various computation modes and register the available computation modes. The computation modes include local execution, remote call, federated learning, and multi-party secure computation. The configuration center unit dynamically configures the feature matching template based on the feature identification results, writes the computational resource features into the feature matching template, and builds a unified feature resource catalog based on the feature matching template.

[0008] Preferably, the method for dynamically configuring feature matching templates based on feature recognition results includes: The system acquires computing resource elements and element recognition results. Natural language processing technology extracts key features of the elements based on the element recognition results and generates structured tagged element descriptors based on the key features. The tagged element descriptors include element feature type, priority, compliance, sensitivity, and element feature symbol. Based on the feature feature type of the tagged feature descriptor, retrieve the associated candidate templates in the template meta database, export at least one set of candidate templates, generate a candidate template queue composed of candidate templates, and construct vectorized feature priority matrix, feature compliance matrix, and feature sensitivity matrix based on the priority, compliance, and sensitivity of the tagged feature descriptor, respectively. Load vectorized element priority matrices, element compliance matrices, and element sensitivity matrices. Perform weight self-learning training on these matrices and then weight them to obtain weighted element priority matrices, element compliance matrices, and element sensitivity matrices. During weight self-learning training, the matrices are weighted based on the BM25 algorithm. The matrix weight calculation formula is expressed as follows: in, Represents the matrix weights. For matrix The set of elements Represents the matrix weight constraint function. For matrix dimensions, These are the sparsification threshold and the time-related decay coefficient, respectively. The initial weights of the matrix, These are the matrix uncertainty and the second-order norm of the matrix, respectively. The weighted element priority matrix, element compliance matrix, and element sensitivity matrix are superimposed to obtain the weighted adjacency element matrix; Load the candidate template queue consisting of candidate templates, query the template resource combinations of candidate templates in the candidate template queue, and construct the candidate resource combination of candidate templates based on the template resource combination; Calculate the utility score of the weighted adjacency element matrix and the candidate resource combination. Use the candidate template corresponding to the candidate resource combination with the highest utility score as the element matching template. Use the weighted adjacency element matrix as the quantitative input of the element matching template to complete the writing of resource elements into the element matching template. The formula for calculating the utility score of the weighted adjacency element matrix and candidate resource combination is as follows: in, For weighted adjacency element matrix Combined with candidate resources The utility score, These are weighted adjacency element matrices. Combined with candidate resources Interaction weight coefficients, weighted adjacency matrix Loading time for candidate templates Represents the weighted adjacency matrix Medium element connection vector Combined with candidate resources Embedding similarity, Indicates candidate resource combinations Resource Link Vector With weighted adjacency matrix Embedding similarity, These are weighted adjacency element matrices. Combined with candidate resources The second-order norm.

[0009] Preferably, the scheduling response module includes: The task receiving unit is used to receive task scheduling requests sent by multiple protocols, parse the task scheduling requests sent by multiple protocols, generate vectorized task parsing results, and output the task parsing results. The task parsing results include task type, data requirements, algorithm requirements, computing power requirements, and computing mode. The element association unit is used to obtain the task parsing results, calculate the vector matching degree between the task parsing results and the weighted adjacency element matrix using cosine similarity, select the weighted adjacency element matrix with the highest vector matching degree score to associate with the task parsing results, mark the computing resource elements associated with the weighted adjacency element matrix as "associated", and upload the association status to the log audit and monitoring module. The formula for calculating the vector matching degree is as follows: in, Indicates the result of task parsing. With weighted adjacency matrix Vector matching degree; The dynamic packaging unit is used to obtain the correlation of the task parsing results, calculate the element matching template corresponding to the resource element based on the computing resource element index, and package it into a dynamic resource package.

[0010] Preferably, the trusted execution module includes: The trusted environment creation unit is used to obtain task scheduling requests and identify task parsing results, and initialize a trusted execution environment based on the task parsing results; The resource scheduling request unit is used to obtain dynamic resource packages and, in combination with the real-time status of the computing power nodes, distribute the dynamic resource packages to the computing power nodes. The computing power nodes generate resource scheduling requests in real time. The resource scheduling unit, in response to a resource scheduling request, drives the computing nodes to allocate the scheduling data stream to the corresponding data transmission link based on the load balancing strategy.

[0011] Preferably, the method for initializing a trusted execution environment based on task parsing results includes: Load the task parsing results, retrieve the registration identity information of the computing resource elements associated with the task parsing results, verify the registration identity information, and determine whether the registration identity is valid. If the registration is successful, the system receives a Trusted Environment Requirements Report from the element uploader, parses the Trusted Environment Requirements Report, and assesses the trustworthiness of the calculated resource elements. The formula for calculating the credibility of computational resource elements is as follows: in, This indicates the reliability of the calculated resource elements. These are identity credibility, information completeness, environmental reliability, vulnerability risk level, and behavioral compliance. Identity credibility is determined by whether the user is registered and has a valid certificate; information completeness is the ratio of computational resource elements to standard resource elements; and vulnerability risk level is assessed using a reverse evaluation of CVE scores. These are the weighting coefficients for identity credibility, information integrity, environmental reliability, vulnerability risk level, and behavioral compliance, respectively. Determine whether the trustworthiness of computing resource elements meets the standard trusted execution environment threshold; If the credibility of the calculated resource elements meets the standard trusted execution environment threshold, a trusted verification credential is issued and uploaded to the log auditing and monitoring module, written into the trusted environment requirement report, and the trusted execution environment is initialized based on the trusted environment requirement report; If the trustworthiness of computing resource elements does not meet the standard trusted execution environment threshold, the trusted execution hardware and software information, measurement environment information, and authentication policy information are obtained from the trusted environment requirement report based on the computing resource elements. Initiate a request to create a trusted execution environment based on trusted execution hardware and software information, measurement environment information, and authentication policy information; In response to the request to create a trusted execution environment, the computing power resource pool allocates a trusted execution environment initialization memory space, constructs a dynamic isolation domain based on information trust in the trusted execution environment initialization memory space, and generates a random trusted sequence based on a quantum-inspired random adjustment generation algorithm. The random trusted sequence includes trusted public key information, trusted private key information, and a secure exchange key. A random trusted sequence is obtained, and the random trusted sequence is compressed into a single verification point using the quantum signature aggregation method. The single verification point is then merged with the computing resource elements, and the trusted execution environment is initialized based on the merged computing resource elements.

[0012] Preferably, the method for distributing dynamic resource packages to computing nodes based on their real-time status includes: Traverse the computing power nodes, identify the real-time status of the computing power nodes, and use the spatiotemporal causal graph verification framework combined with the element matching template in the dynamic resource package to perform real-time simulation of the real-time status of the computing power nodes, and obtain the computing power occupancy prediction results of the dynamic resource package. The computing resource pool sets an initial priority preemption strategy based on the computing power occupancy prediction results to obtain an initial allocation scheme for computing power nodes. The initial allocation scheme for computing power nodes includes the initial mapping relationship between computing power nodes and dynamic resource packages. The initial allocation scheme for computing nodes is loaded. Based on the real-time status of computing nodes, the availability of node resources, network health status, and storage space are identified. The priority and computing power occupancy of dynamic resource packages are also identified. The availability of node resources, network health status, storage space, priority, and computing power occupancy are normalized and quantified to obtain resource availability, network health, storage space, resource priority, and computing power occupancy. Using resource availability, network health, storage space, resource priority, and computing power occupancy as verification indicators, an indicator reward function is constructed. The indicator reward function is solved to obtain the comprehensive matching value between computing nodes and dynamic resource packages. in, These represent resource availability, network health, storage space availability, resource priority, and computing power utilization, respectively. These are the weighting coefficients for resource availability, network health, storage space availability, resource priority, and computing power utilization, respectively. Indicates storage space constraints; in, For storage space constraints, These are the initial constraint threshold and the time decay coefficient, respectively; Determine whether the comprehensive matching value corresponding to the computing power node meets the preset matching threshold; If the comprehensive matching value corresponding to the computing power node meets the preset matching threshold, establish the mapping relationship between the dynamic resource package and the computing power node in the initial allocation scheme of the computing power node; If the comprehensive matching value corresponding to the computing power node does not meet the preset matching threshold, the initial mapping relationship between the dynamic resource package and the computing power node in the initial allocation scheme is released. The computing power nodes and dynamic resource packages that have not established a mapping relationship are fed back to the computing power resource pool. A weighted round-robin strategy is used to perform secondary matching on the computing power nodes and dynamic resource packages that have not established a mapping relationship, and the preset matching threshold is reduced to obtain the computing power node adjustment allocation scheme. Based on the computing power node adjustment allocation scheme, the adjustment mapping relationship between the dynamic resource package and the computing power node is established. Based on the mapping relationship between dynamic resource packages and computing power nodes, and the adjustment mapping relationship between dynamic resource packages and computing power nodes, the dynamic resource package is divided into multiple data blocks. The computing power node creates a temporary directory based on the initialization trusted execution environment requirements and transmits multiple data blocks to the temporary directory of the computing power node.

[0013] Compared with the prior art, the embodiments of this application have the following main advantages: In this embodiment of the invention, the system is equipped with a trusted execution module. The trusted execution module can initialize the trusted execution environment based on the task scheduling request and distribute dynamic resource packages to the computing power nodes in combination with the real-time status of the computing power nodes. This enables dynamic resource perception and real-time status simulation by combining the real-time status of the computing power nodes. On the one hand, it solves the problem of static information dependence. On the other hand, it can also construct an indicator reward function to calculate a comprehensive matching value, quantify the adaptability of nodes and resource packages from multiple dimensions, avoid the one-sidedness of a single indicator, improve the matching accuracy of resource packages and computing power nodes, and reduce scheduling failures or performance losses caused by information asymmetry.

[0014] In this embodiment of the invention, the unified access module consists of an element identification unit, a data management unit, an algorithm management unit, a computing power management unit, and a configuration center unit. Through the collaborative efforts of multiple units, it supports the registration of heterogeneous computing power nodes such as local servers, cloud hosts, and edge nodes, as well as multi-source heterogeneous data and algorithm components. The unified access module supports the dynamic registration and template configuration of multiple types of computing resource elements (data, algorithms, computing power, computing modes). Combined with multi-protocol task parsing and dynamic packaging, it realizes the efficient integration and flexible scheduling of heterogeneous resources, making the invention adaptable to complex business scenarios. This overcomes the problems of traditional resource management models, such as lack of correlation between resources, serious information silos, and the need for manual cross-system matching during scheduling, which are inefficient and prone to errors.

[0015] In this embodiment of the invention, the scheduling response module consists of a task receiving unit, an element association unit, and a dynamic packaging unit. The task receiving unit parses the task scheduling request, and the task parsing result is vectorized to generate a vector containing key information such as task type, data requirements, algorithm requirements, computing power requirements, and computing mode. The vectorization method not only facilitates subsequent calculation and matching but also improves the efficiency and accuracy of task parsing. Furthermore, the element association unit accurately associates resource elements and packages them into standardized dynamic resource packages, providing an adaptation carrier for subsequent trusted execution and resource scheduling. Multi-dimensional vector matching is used to replace manual rules, avoiding the one-sidedness of single-label matching. At the same time, the dynamic resource package serves as the input to the trusted execution module, and its standardized format and pre-verified resource elements ensure the rapid initialization of the trusted execution environment.

[0016] In this embodiment of the invention, the trusted execution module consists of a trusted environment creation unit, a resource scheduling request unit, and a resource scheduling unit. When the trusted environment creation unit initializes the trusted execution environment based on the task parsing results, it introduces a trusted execution environment initialization process (trustworthiness assessment, dynamic isolation domain, random trusted sequence) and full lifecycle auditing, which ensures the confidentiality, integrity, and code immutability of data during task execution, meets the requirements of high-security scenarios such as privacy computing and sensitive data processing, and improves the matching accuracy of resource packages and computing power nodes by using a spatiotemporal causal graph verification framework to predict computing power usage, multi-dimensional indicator reward function to calculate comprehensive matching value, and weighted round-robin secondary matching mechanism, thereby reducing scheduling failures or performance losses caused by information asymmetry.

[0017] In this embodiment of the invention, when dynamically configuring element matching templates based on element recognition results, key features of computing resource elements are extracted using natural language processing technology. This generates structured, tagged element descriptors containing element feature types, priorities, compliance, sensitivity, and element signatures. These refined descriptors comprehensively and accurately characterize the features of computing resource elements, providing rich semantic information for subsequent resource matching. Furthermore, based on the element feature types of the tagged element descriptors, associated candidate templates are retrieved from the template metadata database to generate a candidate template queue. By dynamically generating the candidate template queue, suitable templates can be quickly selected based on different resource features, improving template matching efficiency. The process from resource recognition and template matching to resource writing is highly automated and efficient, significantly shortening resource configuration time, improving overall system performance, and solving the problems of low efficiency, poor adaptability, and difficult maintenance associated with traditional template configuration.

[0018] In this embodiment of the invention, when initializing the Trusted Execution Environment (TEE) based on the task parsing results, TEE resources are changed from static pre-allocation to dynamic on-demand allocation by tasks. This ensures that TEE initialization is triggered only when a task requires it, avoiding resource hoarding when there are no tasks. Simultaneously, the trust assessment is upgraded from single-index authentication to multi-dimensional comprehensive verification. When the trustworthiness of computing resource elements does not meet the standard trusted execution environment threshold, a request to create a trusted execution environment is initiated. The computing resource pool allocates the initialization memory space for the trusted execution environment and constructs a dynamic isolation domain based on information trustworthiness. This dynamic isolation domain effectively isolates different tasks and resources, preventing the spread of malware or attacks within the system. It compresses random trusted sequences into a single verification point and merges it into the computing resource elements. During TEE initialization, this point is verified to ensure the uniqueness and immutability of TEE startup parameters, thus guaranteeing the security and isolation of task execution and meeting the requirements of high-security scenarios such as privacy computing and sensitive data processing.

[0019] In this embodiment of the invention, when dynamic resource packages are distributed to computing power nodes in combination with the real-time status of the computing power nodes, dynamic indicators such as CPU utilization, memory usage, network latency / bandwidth, and remaining storage space are collected in real time by traversing the computing power nodes. This avoids relying on static initial information and ensures the real-time perception capability of the system. Furthermore, by constructing a spatiotemporal causal graph using the element matching template in the dynamic resource package, the resource usage change trend of nodes under different allocation strategies can be simulated. This allows scheduling decisions to shift from being based on historical static information to being based on future dynamic predictions. The design of the multi-dimensional indicator reward function enables refined matching decisions between computing power nodes and dynamic resource packages. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the architecture of the connector system with trusted execution and resource scheduling capabilities provided by the present invention.

[0021] Figure 2 A schematic diagram illustrating the implementation process of a method for dynamically configuring feature matching templates based on feature recognition results is shown.

[0022] Figure 3 A schematic diagram of the implementation process of the method for initializing a trusted execution environment based on task parsing results is shown.

[0023] Figure 4 The diagram illustrates the process of distributing dynamic resource packages to computing nodes based on their real-time status.

[0024] In the diagram: 100 - Unified Access Module, 110 - Element Identification Unit, 120 - Data Management Unit, 130 - Algorithm Management Unit, 140 - Computing Power Management Unit, 150 - Computing Method Management Unit, 160 - Configuration Center Unit, 200 - Scheduling Response Module, 210 - Task Receiving Unit, 220 - Element Association Unit, 230 - Dynamic Packaging Unit, 300 - Trusted Execution Module, 310 - Trusted Environment Creation Unit, 320 - Resource Scheduling Request Unit, 330 - Resource Scheduling Unit, 400 - Log Auditing and Monitoring Module. Detailed Implementation

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0026] Existing methods rely solely on information reported by device plugins during initial access to their node resource capability lists, failing to detect real-time dynamic changes in resources. In scenarios where resource status fluctuates in real time, scheduling decisions based on outdated resource information lead to unreasonable task allocation. To address these issues, we propose a connector system with trusted execution and resource scheduling capabilities. In short, the system comprises a unified access module 100, a scheduling response module 200, a trusted execution module 300, and a log auditing and monitoring module 400. During operation, the unified access module 100 first responds to computing resource element access commands, registers and identifies computing resource elements based on their types, dynamically configures element matching templates based on the element identification results, and writes the computing resource elements into the element matching templates. Then, the scheduling response module 200 receives and parses task scheduling requests sent via multiple protocols, and determines the appropriate task allocation method based on the task scheduling request and the required computing resources. The system performs element-based matching, binding computing resource elements that match the task scheduling request, and indexing the element matching templates corresponding to the computing resource elements. These elements are then packaged into a dynamic resource package. Finally, the trusted execution module 300 initializes the trusted execution environment based on the task scheduling request, obtains the dynamic resource package, and distributes the dynamic resource package to the computing power nodes based on the real-time status of the computing power nodes. The computing power nodes generate resource scheduling requests in real time and respond to these requests by distributing the scheduling data stream to the corresponding data transmission link based on a load balancing strategy. Simultaneously, the log auditing and monitoring module 400 performs full lifecycle recording and compliance auditing of the scheduling data stream, data transmission link, and computing power operation behavior. In this embodiment of the invention, the system is equipped with a trusted execution module 300. The trusted execution module 300 can initialize the trusted execution environment based on the task scheduling request and distribute dynamic resource packages to the computing power nodes in combination with the real-time status of the computing power nodes. This enables dynamic resource perception and real-time status simulation by combining the real-time status of the computing power nodes. On the one hand, it solves the problem of static information dependence. On the other hand, it can also construct an indicator reward function to calculate a comprehensive matching value, quantify the adaptability of nodes and resource packages from multiple dimensions, avoid the one-sidedness of a single indicator, improve the matching accuracy of resource packages and computing power nodes, and reduce scheduling failures or performance losses caused by information asymmetry.

[0027] This invention provides a connector system with trusted execution and resource scheduling capabilities. Figure 1 A schematic diagram of a connector system architecture with trusted execution and resource scheduling capabilities is shown. Specifically, the connector system with trusted execution and resource scheduling capabilities includes: The unified access module 100 responds to the access command for computing resource elements, registers and identifies computing resource elements based on their types, dynamically configures element matching templates based on the element identification results, and writes the computing resource elements into the element matching templates. In this embodiment of the invention, the unified access module 100 includes: The element identification unit 110 responds to the computing resource element access command, identifies the type of computing resource element, and assigns the computing resource element to the corresponding associated unit; The data management unit 120 is used to acquire multi-source heterogeneous data resources, register the data resources, and define the data resource format, schema, access interface and security policy. The data resources can be structured databases, unstructured files and streaming data. The algorithm management unit 130 is used to acquire various algorithm components and register, load, version control, and configure calling permissions for the algorithm components. The algorithm components can be machine learning models, statistical analysis tools, and cryptographic algorithms. The computing power management unit 140 is used to access computing power nodes of local servers, cloud hosts, and edge nodes, and to register the performance indicators, status, and access paths of computing power nodes. The computation mode management unit 150 is used to acquire multiple computation modes and register the available computation modes. The computation modes include local execution, remote call, federated learning, and multi-party secure computation. The configuration center unit 160 dynamically configures the feature matching template based on the feature identification results, writes the computational resource features into the feature matching template, and constructs a unified feature resource catalog based on the feature matching template.

[0028] In this embodiment of the invention, the configuration center unit 160 is electrically connected to the element identification unit 110, data management unit 120, algorithm management unit 130, and computing power management unit 140. The unified access module 100 is composed of the element identification unit 110, data management unit 120, algorithm management unit 130, computing power management unit 140, and configuration center unit 160. Through the collaborative support of multiple units, it supports the registration of heterogeneous computing power nodes such as local servers, cloud hosts, and edge nodes, as well as multi-source heterogeneous data and algorithm components. The unified access module 100 supports the dynamic registration and template configuration of multiple types of computing resource elements (data, algorithms, computing power, computing modes). Combined with multi-protocol task parsing and dynamic packaging, it realizes the efficient integration and flexible scheduling of heterogeneous resources, making the invention adaptable to complex business scenarios. This overcomes the problems of lack of correlation between resources, serious information silos, and the need for manual cross-system matching during scheduling in traditional resource management models, which are inefficient and prone to errors.

[0029] The scheduling response module 200 is used to receive and parse task scheduling requests sent by multiple protocols, perform joint matching between task scheduling requests and computing resource elements, bind computing resource elements that match the task scheduling requests, index the element matching templates corresponding to the computing resource elements, and package them into dynamic resource packages. In this embodiment of the invention, the scheduling response module 200 includes: The task receiving unit 210 is used to receive task scheduling requests sent by multiple protocols (HTTP, gRPC, MQTT, CoAP protocols), parse the task scheduling requests sent by multiple protocols, generate vectorized task parsing results, and output the task parsing results. The task parsing results include task type, data requirements, algorithm requirements, computing power requirements, and computing mode. The element association unit 220 is used to obtain the task parsing results, calculate the vector matching degree between the task parsing results and the weighted adjacency element matrix using cosine similarity, select the weighted adjacency element matrix with the highest vector matching degree score to associate with the task parsing results, and mark the computing resource elements associated with the weighted adjacency element matrix as "associated" and upload the association status to the log audit and monitoring module 400. The formula for calculating the vector matching degree is as follows: in, Indicates the result of task parsing. With weighted adjacency matrix Vector matching degree; The dynamic packaging unit 230 is used to obtain the association relationship of the task parsing results, calculate the element matching template corresponding to the resource element based on the resource element index, and package it into a dynamic resource package.

[0030] In this embodiment of the invention, the scheduling response module 200 consists of a task receiving unit 210, an element association unit 220, and a dynamic packaging unit 230. The task receiving unit 210 parses the task scheduling request, and the task parsing result is vectorized to generate a vector containing key information such as task type, data requirements, algorithm requirements, computing power requirements, and computing mode. The vectorization method not only facilitates subsequent calculation and matching but also improves the efficiency and accuracy of task parsing. Furthermore, the element association unit 220 accurately associates resource elements and packages them into a standardized dynamic resource package, providing an adaptation carrier for subsequent trusted execution and resource scheduling. Multi-dimensional vector matching is used to replace manual rules, avoiding the one-sidedness of single-label matching. At the same time, the dynamic resource package serves as the input of the trusted execution module 300, and its standardized format and pre-verified resource elements ensure the rapid initialization of the trusted execution environment.

[0031] The trusted execution module 300 initializes the trusted execution environment based on the task scheduling request, obtains the dynamic resource package, and distributes the dynamic resource package to the computing power node in combination with the real-time status of the computing power node. The computing power node generates a resource scheduling request in real time, responds to the resource scheduling request, and drives the computing power node to allocate the scheduling data stream to the corresponding data transmission link based on the load balancing strategy. In this embodiment of the invention, the trusted execution module 300 includes: Trusted environment creation unit 310 is used to obtain task scheduling requests and identify task parsing results, and initialize a trusted execution environment based on the task parsing results; The resource scheduling request unit 320 is used to obtain dynamic resource packages and send the dynamic resource packages to the computing power nodes in combination with the real-time status of the computing power nodes. The computing power nodes generate resource scheduling requests in real time. Resource scheduling unit 330, in response to resource scheduling requests, drives computing nodes to allocate scheduling data streams to corresponding data transmission links based on load balancing strategies.

[0032] In this embodiment of the invention, the trusted execution module 300 consists of a trusted environment creation unit 310, a resource scheduling request unit 320, and a resource scheduling unit 330. When the trusted environment creation unit 310 initializes the trusted execution environment based on the task parsing results, it introduces a trusted execution environment initialization process (trustworthiness assessment, dynamic isolation domain, random trusted sequence) and full lifecycle auditing, which ensures the confidentiality, integrity, and code immutability of data during task execution, meets the requirements of high-security scenarios such as privacy computing and sensitive data processing, and improves the matching accuracy of resource packages and computing power nodes by using a spatiotemporal causal graph verification framework to predict computing power usage, multi-dimensional indicator reward function to calculate comprehensive matching value, and weighted round-robin secondary matching mechanism, thereby reducing scheduling failures or performance losses caused by information asymmetry.

[0033] The Log Audit and Monitoring Module 400 is used to record and audit the entire lifecycle of scheduling data streams, data transmission links, and computing power operation behaviors.

[0034] In this embodiment, the log auditing and monitoring module 400 is connected to the unified access module 100, the scheduling response module 200, and the trusted execution module 300 via Bluetooth or a local area network. The unified access module 100 is equipped with an access layer, which includes a web-based visual management interface, API interfaces (REST, gRPC), an SDK access terminal, and a multi-protocol adapter (HTTP). The access layer is connected to external resource interfaces, which are connected to external data sources, algorithm repositories, computing power nodes, blockchain evidence storage systems, and third-party collaborative systems.

[0035] In this embodiment of the invention, the system is equipped with a trusted execution module 300. The trusted execution module 300 can initialize the trusted execution environment based on the task scheduling request and distribute dynamic resource packages to the computing power nodes in combination with the real-time status of the computing power nodes. This enables dynamic resource perception and real-time status simulation by combining the real-time status of the computing power nodes. On the one hand, it solves the problem of static information dependence. On the other hand, it can also construct an indicator reward function to calculate a comprehensive matching value, quantify the adaptability of nodes and resource packages from multiple dimensions, avoid the one-sidedness of a single indicator, improve the matching accuracy of resource packages and computing power nodes, and reduce scheduling failures or performance losses caused by information asymmetry.

[0036] Traditional template configurations are static and cannot adapt to dynamic changes in feature characteristics or task requirements, leading to template obsolescence and failure. This invention provides a method for dynamically configuring feature matching templates based on feature recognition results. Figure 2 This diagram illustrates the implementation flow of a method for dynamically configuring feature matching templates based on feature recognition results. Specifically, this method includes: S101: Obtain computing resource elements and element recognition results. Natural language processing (NLP) technology extracts key features of the elements based on the element recognition results and generates structured, tagged element descriptors based on these key features. The tagged element descriptors include element feature type, priority, compliance, sensitivity, and element signature. It should be noted that computing resource elements include, but are not limited to, data, algorithms, computing power, and computing modes. By automatically extracting key features using NLP technology and combining candidate template retrieval and weight self-learning training from a template metadata database, the automation of template configuration is achieved. S102: Based on the feature feature type of the tagged feature descriptor, retrieve the associated candidate templates from the template meta-database, export at least one set of candidate templates, generate a candidate template queue composed of candidate templates, and construct vectorized feature priority matrix, feature compliance matrix, and feature sensitivity matrix based on the priority, compliance, and sensitivity of the tagged feature descriptor, respectively; construct feature priority matrix, feature compliance matrix, and feature sensitivity matrix based on the "priority, compliance, and sensitivity" of the tagged descriptor, thereby transforming abstract qualitative descriptions such as "high priority" and "strong compliance" into computable quantitative vectors.

[0037] It should be noted that the candidate template refers to a set of predefined templates retrieved from the template metadata database that have a potential correlation with the feature feature types of the current computing resource elements. It is a candidate set of feature matching templates. The system selects the optimal template as the final feature matching template by analyzing the matching degree between feature features and template characteristics.

[0038] S103: Load the vectorized element priority matrix, element compliance matrix, and element sensitivity matrix. Perform weight self-learning training on the vectorized element priority matrix, element compliance matrix, and element sensitivity matrix, and then perform weighted processing on the matrices to obtain the weighted element priority matrix, element compliance matrix, and element sensitivity matrix. During the weight self-learning training, the matrices are weighted based on the BM25 algorithm. The matrix weight calculation formula is expressed as: in, Represents the matrix weights. For matrix The set of elements Represents the matrix weight constraint function. For matrix dimensions, These are the sparsification threshold and the time-related decay coefficient, respectively. The initial weights of the matrix, These are the matrix uncertainty and the second-order norm of the matrix, respectively. S104. The weighted element priority matrix, element compliance matrix, and element sensitivity matrix are superimposed to obtain the weighted adjacency element matrix. S105, Load the candidate template queue consisting of candidate templates, query the template resource combinations of candidate templates in the candidate template queue, and construct the candidate resource combination of candidate templates based on the template resource combination; S106, calculate the utility score of the weighted adjacency element matrix and the candidate resource combination, take the candidate template corresponding to the candidate resource combination with the highest utility score as the element matching template, use the weighted adjacency element matrix as the quantitative input of the element matching template, and complete the writing of resource elements into the element matching template. In this embodiment of the invention, template selection is changed from type matching to utility optimization, ensuring that the selected template is not only type-matched, but also has the highest collaborative efficiency of resource combination, thus ensuring a safety-efficiency balance between resource elements and task requirements.

[0039] The formula for calculating the utility score of the weighted adjacency element matrix and candidate resource combination is as follows: in, For weighted adjacency element matrix Combined with candidate resources The utility score, These are weighted adjacency element matrices. Combined with candidate resources Interaction weight coefficients, weighted adjacency matrix Loading time for candidate templates Represents the weighted adjacency matrix Medium element connection vector Combined with candidate resources Embedding similarity, Indicates candidate resource combinations Resource Link Vector With weighted adjacency matrix Embedding similarity, These are weighted adjacency element matrices. Combined with candidate resources The second-order norm.

[0040] In this embodiment of the invention, when dynamically configuring element matching templates based on element recognition results, key features of computing resource elements are extracted using natural language processing technology. This generates structured, tagged element descriptors containing element feature types, priorities, compliance, sensitivity, and element signatures. These refined descriptors comprehensively and accurately characterize the features of computing resource elements, providing rich semantic information for subsequent resource matching. Furthermore, based on the element feature types of the tagged element descriptors, associated candidate templates are retrieved from the template metadata database to generate a candidate template queue. By dynamically generating the candidate template queue, suitable templates can be quickly selected based on different resource features, improving template matching efficiency. The process from resource recognition and template matching to resource writing is highly automated and efficient, significantly shortening resource configuration time, improving overall system performance, and solving the problems of low efficiency, poor adaptability, and difficult maintenance associated with traditional template configuration.

[0041] Considering that traditional TEE initialization is mostly static pre-configuration (allocating fixed TEE space to all nodes in advance) or not bound to specific tasks (initializing only based on node type), low-security tasks may consume high-specification TEE resources, and high-security tasks may fail to start due to the lack of pre-configuration of the TEE or the inability of static configuration to meet dynamic requirements, this embodiment of the invention provides a method for initializing a trusted execution environment based on task parsing results. Figure 3 This diagram illustrates the implementation flow of a method for initializing a trusted execution environment based on task parsing results. Specifically, this method includes: S201, Load the task parsing result, retrieve the registration identity information of the computing resource elements associated with the task parsing result, verify the registration identity information, and determine whether the registration identity is approved; S202, If the registration identity is approved, receive the Trusted Environment Requirement Report sent by the element uploader, parse the Trusted Environment Requirement Report, and assess the trustworthiness of the calculated resource elements; The formula for calculating the credibility of computational resource elements is as follows: in, This indicates the reliability of the calculated resource elements. These are identity credibility, information completeness, environmental reliability, vulnerability risk level, and behavioral compliance. Identity credibility is determined by whether the user is registered and has a valid certificate; information completeness is the ratio of computational resource elements to standard resource elements; and vulnerability risk level is assessed using a reverse evaluation of CVE scores. These are the weighting coefficients for identity credibility, information integrity, environmental reliability, vulnerability risk level, and behavioral compliance, respectively. S203, determine whether the trustworthiness of computing resource elements meets the standard trusted execution environment threshold; S204. If the credibility of the computational resource elements meets the standard trusted execution environment threshold, a trusted verification credential is issued and uploaded to the log auditing and monitoring module 400. The trusted environment requirement report is written and the trusted execution environment is initialized based on the trusted environment requirement report. The log recording mechanism facilitates the traceability and auditing of the initialization process of the trusted execution environment, thereby enhancing the reliability and maintainability of the system.

[0042] S205, If the trustworthiness of computing resource elements does not meet the standard trusted execution environment threshold, obtain trusted execution hardware and software information, measurement environment information, and authentication policy information based on the trusted environment requirement report of computing resource elements; S206, Initiate a request to create a trusted execution environment based on trusted execution hardware and software information, measurement environment information, and authentication policy information; S207, in response to the request to create a trusted execution environment, the computing power resource pool allocates a trusted execution environment initialization memory space, constructs a dynamic isolation domain based on information trust in the trusted execution environment initialization memory space, and generates a random trusted sequence based on a quantum-inspired random adjustment generation algorithm. The random trusted sequence includes trusted public key information, trusted private key information, and a secure exchange key, making its randomness inspired by the principles of quantum mechanics and more difficult to predict than traditional pseudo-random numbers. S208: Obtain a random trusted sequence, compress the random trusted sequence into a single verification point using the quantum signature aggregation method, merge the single verification point with the computing resource elements, and initialize the trusted execution environment based on the merged computing resource elements.

[0043] In this embodiment of the invention, when initializing the Trusted Execution Environment (TEE) based on the task parsing results, TEE resources are changed from static pre-allocation to dynamic on-demand allocation by tasks. This ensures that TEE initialization is triggered only when a task requires it, avoiding resource hoarding when there are no tasks. Simultaneously, the trust assessment is upgraded from single-index authentication to multi-dimensional comprehensive verification. When the trustworthiness of computing resource elements does not meet the standard trusted execution environment threshold, a request to create a trusted execution environment is initiated. The computing resource pool allocates the initialization memory space for the trusted execution environment and constructs a dynamic isolation domain based on information trustworthiness. This dynamic isolation domain effectively isolates different tasks and resources, preventing the spread of malware or attacks within the system. It compresses random trusted sequences into a single verification point and merges it into the computing resource elements. During TEE initialization, this point is verified to ensure the uniqueness and immutability of TEE startup parameters, thus guaranteeing the security and isolation of task execution and meeting the requirements of high-security scenarios such as privacy computing and sensitive data processing.

[0044] Traditional resource scheduling relies heavily on statically reported initial resource information from nodes, failing to reflect the dynamic status of nodes in real time. Scheduling based on outdated information often leads to resource overload or idle resources. This invention provides a method for distributing dynamic resource packages to computing nodes based on their real-time status. Figure 4 This diagram illustrates the implementation flow of a method for distributing dynamic resource packages to computing nodes based on their real-time status. The method specifically includes: S301 traverses the computing power nodes, identifies their real-time status, and uses a spatiotemporal causal graph verification framework combined with element matching templates in the dynamic resource package to perform real-time simulation of the computing power nodes' real-time status, obtaining the computing power occupancy prediction results of the dynamic resource package. The spatiotemporal causal graph verification framework uses a graph model to characterize the spatiotemporal dependencies and causal logic between nodes. In the graph model, the spatial dimension is: modeling the network topology relationship between nodes (such as the hierarchical connection between edge nodes and cloud hosts) and resource sharing relationship, while the temporal dimension is: modeling the temporal variation law of resource occupancy. Finally, the rationality of the simulation is verified through causal reasoning to avoid prediction distortion. S302, the computing power resource pool sets an initial priority preemption strategy based on the computing power occupancy prediction results to obtain an initial allocation scheme for computing power nodes, wherein the initial allocation scheme for computing power nodes includes the initial mapping relationship between computing power nodes and dynamic resource packages; S303: Load the initial allocation scheme for computing power nodes. Based on the real-time status of computing power nodes, identify node resource availability, network health status, and storage space, and identify the priority and computing power occupancy of dynamic resource packages. Normalize and quantify the node resource availability, network health status, storage space, priority, and computing power occupancy to obtain resource availability, network health, storage space, resource priority, and computing power occupancy. Among them, resource availability reflects the proportion of CPU, memory, and other resources currently available for allocation to the node, while network health is represented by a combination of latency, bandwidth, and packet loss rate. Using resource availability, network health, storage space, resource priority, and computing power occupancy as verification indicators, construct an indicator reward function, solve the indicator reward function, and obtain the comprehensive matching value between computing power nodes and dynamic resource packages. in, These represent resource availability, network health, storage space availability, resource priority, and computing power utilization, respectively. These are the weighting coefficients for resource availability, network health, storage space availability, resource priority, and computing power utilization, respectively. Indicates storage space constraints; in, For storage space constraints, These are the initial constraint threshold and the time decay coefficient, respectively.

[0045] S304, determine whether the comprehensive matching value corresponding to the computing power node meets the preset matching threshold, wherein the preset matching threshold can be 0.7-0.85; S305, If the comprehensive matching value corresponding to the computing power node meets the preset matching threshold, establish the mapping relationship between the dynamic resource package and the computing power node in the initial allocation scheme of the computing power node; S306, if the comprehensive matching value corresponding to the computing power node does not meet the preset matching threshold, the initial mapping relationship between the dynamic resource package and the computing power node in the initial allocation scheme is released, and the computing power nodes and dynamic resource packages without established mapping relationship are fed back to the computing power resource pool. A weighted round-robin strategy is used to perform secondary matching on the computing power nodes and dynamic resource packages without established mapping relationship, and the preset matching threshold is reduced to obtain the computing power node adjustment allocation scheme. Based on the computing power node adjustment allocation scheme, the adjustment mapping relationship between the dynamic resource package and the computing power node is established to ensure that the unmatched resource packages are allocated as much as possible. S307: Based on the mapping relationship between dynamic resource packages and computing power nodes, and the adjustment mapping relationship between dynamic resource packages and computing power nodes, the dynamic resource package is divided into multiple data blocks. The computing power node creates a temporary directory based on the initialization trusted execution environment requirements and transmits multiple data blocks to the temporary directory of the computing power node.

[0046] In this embodiment of the invention, when dynamic resource packages are distributed to computing power nodes in combination with the real-time status of the computing power nodes, dynamic indicators such as CPU utilization, memory usage, network latency / bandwidth, and remaining storage space are collected in real time by traversing the computing power nodes. This avoids relying on static initial information and ensures the real-time perception capability of the system. Furthermore, by constructing a spatiotemporal causal graph using the element matching template in the dynamic resource package, the resource usage change trend of nodes under different allocation strategies can be simulated. This allows scheduling decisions to shift from being based on historical static information to being based on future dynamic predictions. The design of the multi-dimensional indicator reward function enables refined matching decisions between computing power nodes and dynamic resource packages.

[0047] In this embodiment of the invention, by constructing an integrated architecture that combines unified access, scheduling response, trusted execution, and log auditing, integrated management of data, computing power, algorithms, and computation methods within the connector is achieved, breaking down heterogeneous resource barriers and forming a globally manageable resource pool. Relying on multi-protocol task parsing and dynamic template matching mechanisms, it supports interconnection and interoperability between multiple protocols, cross-platforms, and heterogeneous systems, significantly reducing the technical threshold for enterprises or institutions to deploy data collaboration platforms. Through a configurable, scalable, and visualized system architecture design, combined with flexible scheduling strategies such as dynamic resource package segmentation and weighted round-robin secondary matching, it greatly improves user experience and system compatibility. Simultaneously, based on the dynamic initialization of the trusted execution environment (including multi-dimensional trustworthiness assessment and quantum-enhanced security isolation) and full lifecycle behavior auditing, the system ensures compliant operation and resists complex security threats. Finally, through spatiotemporal causal graph simulation prediction, multi-dimensional indicator reward function matching, and intelligent selection of various computing modes (such as AI inference and joint modeling), the collaborative efficiency and security in data-driven scenarios are significantly improved. This enables the system to not only efficiently support the needs of highly dynamic tasks, but also meet the stringent requirements of high-security fields such as finance, healthcare, and industry, becoming a "trustworthy, efficient, and easy-to-use" core hub for data collaboration that is adaptable to multiple business scenarios.

[0048] In summary, this invention provides a connector system with trusted execution and resource scheduling capabilities. In this embodiment, the system includes a trusted execution module 300. The trusted execution module 300 can initialize a trusted execution environment based on task scheduling requests and distribute dynamic resource packages to computing nodes in conjunction with the real-time status of computing nodes. This enables dynamic resource perception and real-time status simulation by combining the real-time status of computing nodes. On the one hand, it solves the problem of static information dependence; on the other hand, it can construct an indicator reward function to calculate a comprehensive matching value, quantifying the adaptability of nodes and resource packages from multiple dimensions, avoiding the one-sidedness of a single indicator, improving the matching accuracy of resource packages and computing nodes, and reducing scheduling failures or performance losses caused by information asymmetry.

[0049] It should be noted that, for the sake of simplicity, the foregoing embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to the present invention. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0050] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete, or otherwise adjust the features of the various embodiments of the present invention according to the circumstances without conflict or creative effort, thereby obtaining different technical solutions that do not fundamentally depart from the concept of the present invention. These technical solutions also fall within the scope of protection of the present invention.

Claims

1. A connector system with trusted execution and resource scheduling capabilities, characterized in that, The system includes: The unified access module responds to the access command of computing resource elements, registers and identifies computing resource elements based on their types, dynamically configures element matching templates based on the element identification results, and writes the computing resource elements into the element matching templates. The scheduling response module is used to receive and parse task scheduling requests sent by multiple protocols, perform joint matching between task scheduling requests and computing resource elements, bind computing resource elements that match the task scheduling requests, index the element matching templates corresponding to the computing resource elements, and package them into dynamic resource packages. The trusted execution module initializes the trusted execution environment based on the task scheduling request, obtains dynamic resource packages, and distributes the dynamic resource packages to the computing nodes in combination with the real-time status of the computing nodes. The computing nodes generate resource scheduling requests in real time, and respond to the resource scheduling requests by driving the computing nodes to allocate the scheduling data stream to the corresponding data transmission links based on the load balancing strategy. The log auditing and monitoring module is used to record and audit the entire lifecycle of scheduling data streams, data transmission links, and computing power operation behaviors in compliance.

2. The connector system with trusted execution and resource scheduling capabilities as described in claim 1, characterized in that: The unified access module includes: The element identification unit responds to the computing resource element access command, identifies the type of computing resource element, and assigns the computing resource element to the corresponding associated unit; The data management unit is used to acquire multi-source heterogeneous data resources, register the data resources, and define the data resource format, schema, access interface, and security policy. The algorithm management unit is used to acquire various algorithm components and to register, load, version control, and configure access permissions for these components. The computing power management unit is used to access computing power nodes from local servers, cloud hosts, and edge nodes, and to register the performance indicators, status, and access paths of the computing power nodes. The computation mode management unit is used to acquire various computation modes and register the available computation modes. The computation modes include local execution, remote call, federated learning, and multi-party secure computation. The configuration center unit dynamically configures the feature matching template based on the feature identification results, writes the computational resource features into the feature matching template, and builds a unified feature resource catalog based on the feature matching template.

3. The connector system with trusted execution and resource scheduling capabilities as described in claim 2, characterized in that: The method for dynamically configuring feature matching templates based on feature recognition results includes: The system acquires computing resource elements and element recognition results. Natural language processing technology extracts key features of the elements based on the element recognition results and generates structured, tagged element descriptors based on the key features of the elements. Based on the feature feature type of the tagged feature descriptor, retrieve the associated candidate templates in the template meta database, export at least one set of candidate templates, generate a candidate template queue composed of candidate templates, and construct vectorized feature priority matrix, feature compliance matrix, and feature sensitivity matrix based on the priority, compliance, and sensitivity of the tagged feature descriptor, respectively. Load the vectorized element priority matrix, element compliance matrix, and element sensitivity matrix. Perform weight self-learning training on the vectorized element priority matrix, element compliance matrix, and element sensitivity matrix, and then perform weighted processing on the matrices to obtain the weighted element priority matrix, element compliance matrix, and element sensitivity matrix. The weighted element priority matrix, element compliance matrix, and element sensitivity matrix are superimposed to obtain the weighted adjacency element matrix; Load the candidate template queue consisting of candidate templates, query the template resource combinations of candidate templates in the candidate template queue, and construct the candidate resource combination of candidate templates based on the template resource combination; Calculate the utility score of the weighted adjacency feature matrix and candidate resource combinations. Use the candidate template corresponding to the candidate resource combination with the highest utility score as the feature matching template. Use the weighted adjacency feature matrix as the quantitative input of the feature matching template to complete the writing of resource features into the feature matching template.

4. The connector system with trusted execution and resource scheduling capabilities as described in claim 1, characterized in that: The scheduling response module includes: The task receiving unit is used to receive task scheduling requests sent by multiple protocols, parse the task scheduling requests sent by multiple protocols, generate vectorized task parsing results, and output the task parsing results. The task parsing results include task type, data requirements, algorithm requirements, computing power requirements, and computing mode. The element association unit is used to obtain the task parsing results, calculate the vector matching degree between the task parsing results and the weighted adjacency element matrix using cosine similarity, select the weighted adjacency element matrix with the highest vector matching degree score to associate with the task parsing results, mark the computing resource elements associated with the weighted adjacency element matrix as "associated", and upload the association status to the log audit and monitoring module. The dynamic packaging unit is used to obtain the correlation of the task parsing results, calculate the element matching template corresponding to the resource element based on the computing resource element index, and package it into a dynamic resource package.

5. The connector system with trusted execution and resource scheduling capabilities as described in claim 4, characterized in that: The trusted execution module includes: The trusted environment creation unit is used to obtain task scheduling requests and identify task parsing results, and initialize a trusted execution environment based on the task parsing results; The resource scheduling request unit is used to obtain dynamic resource packages and, in combination with the real-time status of the computing power nodes, distribute the dynamic resource packages to the computing power nodes. The computing power nodes generate resource scheduling requests in real time. The resource scheduling unit, in response to a resource scheduling request, drives the computing nodes to allocate the scheduling data stream to the corresponding data transmission link based on the load balancing strategy.

6. The connector system with trusted execution and resource scheduling capabilities as described in claim 5, characterized in that: The method for initializing a trusted execution environment based on task parsing results includes: Load the task parsing results, retrieve the registration identity information of the computing resource elements associated with the task parsing results, verify the registration identity information, and determine whether the registration identity is valid. If the registration is successful, the system will receive a Trusted Environment Requirements Report from the element uploader, parse the Trusted Environment Requirements Report, and assess the trustworthiness of the calculated resource elements. Determine whether the trustworthiness of computing resource elements meets the standard trusted execution environment threshold; If the credibility of the computational resource elements meets the standard trusted execution environment threshold, a trusted verification credential is issued and uploaded to the log auditing and monitoring module, written into the trusted environment requirement report, and the trusted execution environment is initialized based on the trusted environment requirement report.

7. The connector system with trusted execution and resource scheduling capabilities as described in claim 6, characterized in that: The method for initializing a trusted execution environment based on task parsing results further includes: If the trustworthiness of computing resource elements does not meet the standard trusted execution environment threshold, the trusted execution hardware and software information, measurement environment information, and authentication policy information are obtained from the trusted environment requirement report based on the computing resource elements. Initiate a request to create a trusted execution environment based on trusted execution hardware and software information, measurement environment information, and authentication policy information; In response to the request to create a trusted execution environment, the computing power resource pool allocates a trusted execution environment initialization memory space, constructs a dynamic isolation domain based on information trust in the trusted execution environment initialization memory space, and generates a random trusted sequence based on a quantum-inspired random adjustment generation algorithm. The random trusted sequence includes trusted public key information, trusted private key information, and a secure exchange key. A random trusted sequence is obtained, and the random trusted sequence is compressed into a single verification point using the quantum signature aggregation method. The single verification point is then merged with the computing resource elements, and the trusted execution environment is initialized based on the merged computing resource elements.

8. The connector system with trusted execution and resource scheduling capabilities as described in claim 5, characterized in that: The method for distributing dynamic resource packages to computing nodes based on their real-time status includes: Traverse the computing power nodes, identify the real-time status of the computing power nodes, and use the spatiotemporal causal graph verification framework combined with the element matching template in the dynamic resource package to perform real-time simulation of the real-time status of the computing power nodes, and obtain the computing power occupancy prediction results of the dynamic resource package. The computing resource pool sets an initial priority preemption strategy based on the computing power occupancy prediction results to obtain an initial allocation scheme for computing power nodes. The initial allocation scheme for computing power nodes includes the initial mapping relationship between computing power nodes and dynamic resource packages. The initial allocation scheme for computing nodes is loaded. Based on the real-time status of computing nodes, the availability of node resources, network health status, and storage space are identified. The priority and computing space occupied by dynamic resource packages are also identified. The availability of node resources, network health status, storage space, priority, and computing space occupied are normalized and quantified to obtain resource availability, network health, storage space, resource priority, and computing space occupied. Using resource availability, network health, storage space, resource priority, and computing space occupied as verification indicators, an indicator reward function is constructed. The indicator reward function is solved to obtain the comprehensive matching value between computing nodes and dynamic resource packages.

9. The connector system with trusted execution and resource scheduling capabilities as described in claim 8, characterized in that: The method for distributing dynamic resource packages to computing power nodes based on their real-time status also includes: Determine whether the comprehensive matching value corresponding to the computing power node meets the preset matching threshold; If the comprehensive matching value corresponding to the computing power node meets the preset matching threshold, establish the mapping relationship between the dynamic resource package and the computing power node in the initial allocation scheme of the computing power node; If the comprehensive matching value corresponding to the computing power node does not meet the preset matching threshold, the initial mapping relationship between the dynamic resource package and the computing power node in the initial allocation scheme is released. The computing power nodes and dynamic resource packages that have not established a mapping relationship are fed back to the computing power resource pool. A weighted round-robin strategy is used to perform secondary matching on the computing power nodes and dynamic resource packages that have not established a mapping relationship, and the preset matching threshold is reduced to obtain the computing power node adjustment allocation scheme. Based on the computing power node adjustment allocation scheme, the adjustment mapping relationship between the dynamic resource package and the computing power node is established. Based on the mapping relationship between dynamic resource packages and computing power nodes, and the adjustment mapping relationship between dynamic resource packages and computing power nodes, the dynamic resource package is divided into multiple data blocks. The computing power node creates a temporary directory based on the initialization trusted execution environment requirements and transmits multiple data blocks to the temporary directory of the computing power node.

Images