Calculation power scheduling system and method of calculation power network

Abstract

The invention discloses a computing power scheduling system and method for a computing power network, and relates to the technical field of computing power distribution scheduling, and the method comprises the steps: obtaining all computing power nodes in the computing power network, and constructing a computing power node interaction network model; obtaining a computing power task initiating node and computing power task attributes; determining a target attribute cluster of the computing power task based on the computing power task attributes, and recording all computing power nodes in the target attribute cluster as demand target nodes of the computing power task; based on the computing power task initiating node, determining an execution trust target node of the computing power task; solving an intersection of the demand target node of the computing power task and the execution trust target node of the computing power task, and recording the intersection as a comprehensive target node; and determining an execution computing power node of the computing power task from the comprehensive target node in combination with the balanced load. The method has the advantages that by constructing the interactive network model of the cooperative trust relationship and the attribute group clustering, multi-dimensional node dynamic screening and optimal matching are realized, and the efficiency and the reliability of computing power scheduling are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of computing power allocation and scheduling, and particularly to a computing power scheduling system and method for a computing power network. Background Art

[0002] With the continuous expansion of the scale of the computing power network, the dynamic scheduling of computing power resources has become a key challenge in improving network efficiency. In the prior art, computing power scheduling often relies on simple load balancing or static node selection strategies, lacking comprehensive consideration of the collaborative trust relationship between nodes, dynamic attribute clustering, and the complexity of the network topology. Traditional methods are difficult to effectively handle trust transfer between nodes, multi-hop collaboration path optimization, and dynamic feedback adjustment after task execution, resulting in low resource utilization and insufficient reliability of task allocation. In addition, node selection in the prior art usually ignores the impact of node importance on global load balancing and cannot adapt to the dynamic requirements of large-scale heterogeneous computing power networks. Therefore, there is an urgent need for an intelligent computing power scheduling method that integrates trust assessment, clustering analysis, load balancing, and dynamic adjustment. Summary of the Invention

[0003] To solve the above technical problems, a computing power scheduling system and method for a computing power network are provided. This technical solution solves the problems that the above traditional methods are difficult to effectively handle trust transfer between nodes, multi-hop collaboration path optimization, and dynamic feedback adjustment after task execution, resulting in low resource utilization and insufficient reliability of task allocation.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows: A computing power scheduling method for a computing power network, comprising: Obtain all computing power nodes in the computing power network and construct a computing power node interaction network model. The computing power node interaction network model includes a collaborative trust relationship and a clustering relationship. Based on the collaborative trust relationship, construct an interaction relationship chain between computing power nodes, and based on the clustering relationship, construct an attribute group clustering of computing power nodes; Obtain a computing power task initiation node and computing power task attributes; Determine the target attribute group clustering of the computing power task based on the computing power task attributes, and record all computing power nodes within the target attribute group clustering as the demand target nodes of the computing power task; Based on the computing power task initiation node, combined with the interaction relationship chain between computing power nodes, determine the execution trust target node of the computing power task; Find the intersection of the demand target nodes of the computing power task and the execution trust target nodes of the computing power task, and record it as the comprehensive target node; Combined with load balancing, determine the execution computing power nodes of the computing power task from the comprehensive target nodes; Allocate the computing power task to the execution computing power nodes for task execution, and at the same time, combined with the task execution result feedback, adjust the interaction relationship chain between computing power nodes.

[0005] As a preferred solution, the specific steps of constructing the interaction relationship chain between computing power nodes based on the collaborative trust relationship include: Based on the physical distance between computing power nodes, an initial collaborative trust weight is added between each pair of all computing power nodes; If the same computing power task has been completed collaboratively between computing power nodes, a collaborative relationship chain is formed between the computing power nodes. Based on the completion status of the computing power tasks completed collaboratively between the computing power nodes and the historical collaborative trust weight, the collaborative trust weight between the computing power nodes is determined; If no collaborative relationship chain is formed between computing power nodes, but there are at most three collaborative relationship chains that can connect the computing power nodes, a multi-hop relationship chain is formed between the computing power nodes. Based on the collaborative trust weights between the computing power nodes that make up the multi-hop relationship chain, the collaborative trust weights of the computing power nodes at both ends of the multi-hop relationship chain are calculated; If neither a collaborative relationship chain nor a multi-hop relationship chain is formed between computing power nodes, the collaborative trust weight between the computing power nodes is the initial collaborative trust weight; Summarize all collaborative relationship chains and multi-hop relationship chains to obtain the interaction relationship chain between computing power nodes.

[0006] As a further preferred solution: Combining load balancing, determining the computing power nodes for executing computing power tasks from the comprehensive target nodes specifically includes: Determine the computing power load requirements of the computing power task and the computing power load capabilities of each computing power node in the comprehensive target nodes; Set N = 1, and determine whether there are N computing power nodes in the comprehensive target nodes whose combined computing power load capabilities are greater than the computing power load requirements of the computing power task. If so, combine the N computing power nodes into a node topology structure. If not, increment the value of N and return to continue the determination. If the value of N exceeds the number of computing power nodes included in the comprehensive target nodes, an alarm signal is output to the background terminal; Use the ratio of the computing power load requirements of the computing power task to the sum of the computing power load capabilities of all computing power nodes in the node topology structure as the task load rate of the node topology structure; Select the node topology structure with the minimum task load rate, and use all the computing power nodes in it as the computing power nodes for executing the computing power task.

[0007] Furthermore, in combination with the above computing power scheduling method for the computing power network, a computing power scheduling system for the computing power network is also proposed, including: A node management module, used to obtain all computing power nodes in the computing power network and construct a computing power node interaction network model including collaborative trust relationships and clustering relationships; The node management module specifically includes: A weight initialization unit, which adds an initial collaborative trust weight to the computing power nodes based on the physical distance; A collaboration relationship construction unit generates a collaboration relationship chain and dynamic weights according to the completion situation of historical collaboration tasks; A multi-hop relationship calculation unit calculates the multi-hop trust weight for nodes without direct collaboration by multiplying the weights of up to three relationship chains, and takes the maximum value as the final weight; A task parsing module is used to obtain the computing power task initiating node and computing power task attributes, and determine the target attribute group clustering and its required target nodes according to the attributes; A trust evaluation module calculates the collaboration trust weight between the computing power task initiating node and other nodes based on the interaction relationship chain, and filters the execution trust target nodes; A node matching module calculates the intersection of the required target nodes and the execution trust target nodes to generate comprehensive target nodes, and selects the execution computing power nodes from them in combination with the load balancing algorithm; The node matching module specifically includes: A load evaluation unit monitors the computing power load capacity of the comprehensive target nodes in real time; A topology generation unit searches for the smallest node combination that meets the task requirements through an iterative algorithm, and calculates the task load rate of each combination; An optimal decision-making unit selects the node topology with the lowest task load rate as the execution computing power node; A dynamic adjustment module updates the collaboration trust relationship and interaction relationship chain according to the feedback of the task execution result Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention realizes multi-dimensional dynamic screening and optimal matching of nodes by constructing an interactive network model of collaboration trust relationship and attribute group clustering, significantly improving the efficiency and reliability of computing power scheduling. Through trust collaboration and multi-hop path optimization, based on the calculation of collaboration history and multi-hop trust weights, it enhances the credibility of task allocation, reduces the participation of inefficient nodes, load balancing and efficient utilization of resources. By iteratively searching for the node combination with the lowest load rate and combining the node importance correction coefficient, it avoids local overload and improves the global resource utilization rate. Dynamic feedback and network adaptability: adjust the trust relationship chain in real time according to the task execution result, enabling the network to have self-optimization ability. Robustness in complex scenarios: introduce attribute group clustering to quickly lock the target node set, reduce the search complexity in large-scale networks, and effectively solve the problems of trust transfer, computing power resource collaboration and dynamic load balancing in the computing power network. It is applicable to scenarios such as cloud computing and edge computing, and has wide application value. Description of the Drawings

[0008] Figure 1 It is a flowchart of the computing power scheduling method for the computing power network proposed in Embodiment 1; Figure 2 It is a flowchart of the method for constructing an interactive relationship chain between computing power nodes with a collaboration trust relationship proposed in Embodiment 1; Figure 3 Flowchart of the method for determining the execution computing power node of the computing power task from the comprehensive target node proposed in the first embodiment. Detailed implementation manner

[0009] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variants.

[0010] Refer to Figure 1 As shown, a computing power scheduling method for a computing power network includes: Obtain all computing power nodes in the computing power network and construct a computing power node interaction network model. The computing power node interaction network model includes a cooperation trust relationship and a clustering relationship. Based on the cooperation trust relationship, construct an interaction relationship chain between computing power nodes, and based on the clustering relationship, construct an attribute group clustering of computing power nodes. For example, divide GPU / TPU nodes into a compute-intensive community, suitable for AI training tasks, and divide large-capacity hard disk nodes into a storage-intensive community, suitable for data analysis tasks; Obtain the computing power task initiation node and the computing power task attributes; Based on the computing power task attributes, determine the target attribute group clustering of the computing power task, and record all computing power nodes within the target attribute group clustering as the demand target nodes of the computing power task. For example, matrix operation task → compute-intensive community, data preprocessing task → storage-intensive community; Based on the computing power task initiation node, combined with the interaction relationship chain between computing power nodes, determine the execution trust target node of the computing power task; Find the intersection of the demand target nodes of the computing power task and the execution trust target nodes of the computing power task, and record it as the comprehensive target node; Combined with load balancing, determine the execution computing power node of the computing power task from the comprehensive target node; Allocate the computing power task to the execution computing power node for task execution, and at the same time, adjust the interaction relationship chain between computing power nodes in combination with the feedback of the task execution result.

[0011] Refer to Figure 2 As shown, in this embodiment, constructing the interaction relationship chain between computing power nodes based on the cooperation trust relationship specifically includes: Based on the physical distance between computing power nodes, attach an initial cooperation trust weight to each pair of all computing power nodes; Specifically, the initial cooperation trust weight is determined in the following manner; Determine the physical distance between computing power nodes and set the maximum value of the initial collaborative trust weight. Specifically, the maximum value of the initial collaborative trust weight is set to 0.8. Then, use linear mapping or non-linear mapping to determine the additional initial collaborative trust weights between pairs of computing power nodes. The calculation formula for linear mapping is: , where is the initial collaborative trust weight between computing power node i and computing power node j, is the physical distance between computing power node i and computing power node j, is the maximum value of the physical distance between computing power nodes, is the minimum value of the physical distance between computing power nodes; In the initial stage of system operation, due to the lack of collaboration data between nodes, when determining the initial collaborative trust weight, consider the physical distance, which is most critical for communication delay, to determine the initial collaborative trust weights between pairs of computing power nodes; If the computing power nodes have completed the same computing power task in collaboration, a collaboration relationship chain is formed between the computing power nodes. Based on the completion situation of the computing power tasks completed in collaboration between the computing power nodes and the historical collaborative trust weights, determine the collaborative trust weights between the computing power nodes; The collaborative trust weights between computing power nodes are calculated using 0.8 × historical collaborative trust weight + 0.2 × the completion score of the computing power task completed in this collaboration for dynamic configuration. The completion score of the computing power task completed in collaboration is calculated through , where is the task demand data volume, is the actual output data volume, is the task demand data accuracy, is the actual output data accuracy, is the average communication delay of this collaboration, is the system's maximum allowable delay threshold, usually set to 500 ms. If no collaboration relationship chain is formed between computing power nodes, but there are at most three collaboration relationship chains that can connect the computing power nodes, a multi-hop relationship chain is formed between the computing power nodes. Based on the collaborative trust weights between the computing power nodes that make up the multi-hop relationship chain, calculate the collaborative trust weights of the computing power nodes at both ends of the multi-hop relationship chain; If there is only one multi-hop relationship chain between two computing power nodes, the collaborative trust weight between the two computing power nodes is the product of the collaborative trust weights of the computing power nodes that make up the multi-hop relationship chain; If there are multiple multi-hop relationship chains between two computing power nodes, the collaborative trust weight between the two computing power nodes is the maximum value among the products of the collaborative trust weights of the computing power nodes of all multi-hop relationship chains; For example, if there is only one multi-hop relationship chain A→B, B→D between node A and node D, then the collaborative trust weight between node A and node D is: , is the collaborative trust weight between nodes AB, is the collaborative trust weight between nodes BD; If there are two multi-hop relationship chains A→B, B→D and A→C, C→D between nodes A and D, then the collaborative trust weight between node A and node D is: and the maximum value in; is the collaborative trust weight between nodes AC, is the collaborative trust weight between nodes CD.

[0012] If there is neither a collaborative relationship chain nor a multi-hop relationship chain between computing power nodes, the collaborative trust weight between computing power nodes is the initial collaborative trust weight; Summarize all collaborative relationship chains and multi-hop relationship chains to obtain the interaction relationship chain between computing power nodes.

[0013] By collecting the execution results of computing power tasks in real time (such as task success rate, response latency), a dynamic feedback mechanism is established to adaptively correct the collaborative trust relationship chain. Specifically, if a certain node performs excellently in task execution (such as high computing efficiency or low error rate), the direct collaborative trust weight between it and the initiating node will be enhanced positively according to the preset rules, and at the same time, the overall trust weight of its multi-hop path will be improved synchronously; on the contrary, if the node has problems with poor task completion quality, its trust weight will be automatically reduced, and the trust attenuation of the associated multi-hop path will be triggered. This closed-loop feedback mechanism enables the network to quickly identify inefficient or ineffective nodes and gradually exclude them from the preferred path through weight attenuation, while strengthening the collaborative priority of highly reliable nodes.

[0014] Based on the interaction relationship chain between computing power nodes constructed above, by screening out the computing power nodes whose collaborative trust weight with the computing power task initiating node is greater than the trust threshold as the execution trust target nodes of the computing power task, it can be ensured that the execution nodes of the computing power task are high-quality collaborative nodes of the task initiating node, and it is ensured that the computing power task can be collaboratively executed by highly reliable nodes.

[0015] Furthermore, as Figure 3 shown, in this embodiment: Combining load balancing, determining the execution computing power node of the computing power task from the comprehensive target nodes specifically includes: Determine the computing power load demand of the computing power task and the computing power load capacity of each computing power node in the comprehensive target nodes; Set N = 1, and determine whether the sum of the computing power load capacities of N computing power nodes in the comprehensive target node is greater than the computing power load demand of the computing power task. If so, combine the N computing power nodes into a node topology structure. If not, increment the value of N by 1 and return to continue the judgment. If the value of N exceeds the number of computing power nodes included in the comprehensive target node, output an alarm signal to the background terminal; Use the ratio of the computing power load demand of the computing power task to the sum of the computing power load capacities of all computing power nodes in the node topology structure as the task load rate of the node topology structure; Filter out the node topology structure with the smallest task load rate, and use all the computing power nodes in it as the execution computing power nodes for the computing power task.

[0016] Avoid local overload and improve the global resource utilization rate by iteratively searching for the node combination with the minimum load rate, and improve the global rationality of the computing power scheduling in the computing power network. Embodiment

[0017] Furthermore, on the basis of Embodiment 1, this embodiment introduces a node importance correction coefficient when determining the execution computing power nodes of the computing power task, as follows: Based on the collaboration relationship chain of each node, combine the PageRank algorithm to determine the node importance of each computing power node in the computing power network; Among them, the calculation formula for the node importance of the computing power node is: , where is the node importance of the i-th computing power node, is the set of nodes with a collaboration relationship chain with the i-th computing power node, is the node importance of the v-th computing power node in is the number of nodes with a collaboration relationship chain with the v-th computing power node in, M is the total number of nodes in the network, e is the damping coefficient, usually taking a value of 0.85; v is the v-th computing power node in; The calculation formula for the task load rate of the node topology structure after introducing the node importance correction coefficient is: ; Among them, is the task load rate of the i-th node topology structure after introducing the node importance correction coefficient, X is the computing power load demand of the computing power task, is the computing power load capacity of the k-th computing power node in the i-th node topology structure, is the i-th node topology structure, K is the union of all node topology structures, is the node importance of the k-th computing power node in the i-th node topology structure, is the node importance of the $u$-th computing power node in $K$, where $k$ is the $k$-th computing power node in the $i$-th node topology, and $u$ is the $u$-th computing power node in $K$. and is the weight coefficient, which is determined based on the task initiation frequency in the network. If the task initiation frequency in the network is high, it is necessary to ensure the availability of high-importance nodes. At this time, increase value. Conversely, increase value.

[0018] The node importance correction mechanism realizes the precise allocation of computing power resources and the global load optimization. Specifically, the system first adopts a progressive search strategy in the comprehensive target node set according to the load demand of the computing power task: starting from a single node ($N = 1$), gradually expanding to multi-node combinations, and screening the smallest node topology that meets the task computing power demand. The task load rate of each candidate node combination is calculated by a formula, which not only includes the ratio of the task demand to the total computing power of the nodes, but also introduces a node importance correction coefficient based on the PageRank algorithm. This coefficient quantifies its global influence in the network by analyzing the centrality of the node in the collaboration relationship chain, the historical task contribution degree, and the multi-hop path participation frequency, so that the system preferentially selects high-importance node-light load combinations. Embodiment

[0019] This embodiment combines Embodiment 1 and Embodiment 2 to propose a computing power scheduling system for a computing power network, including: A node management module, used to obtain all computing power nodes in the computing power network and construct a computing power node interaction network model including collaboration trust relationships and clustering relationships; A task parsing module, used to obtain the computing power task initiation node and computing power task attributes, and determine the target attribute group clustering and its required target nodes according to the attributes; A trust evaluation module, which calculates the collaboration trust weight between the computing power task initiation node and other nodes based on the interaction relationship chain and screens the trusted target nodes for execution; A node matching module, which calculates the intersection of the required target nodes and the trusted target nodes for execution to generate comprehensive target nodes, and selects the computing power nodes for execution from them in combination with the load balancing algorithm; A dynamic adjustment module, which updates the collaboration trust relationship and the interaction relationship chain according to the feedback of the task execution result; The node management module includes: A weight initialization unit, which attaches an initial collaboration trust weight to the computing power nodes based on the physical distance; A collaboration relationship construction unit, which generates a collaboration relationship chain and dynamic weights according to the completion situation of historical collaboration tasks; The multi-hop relationship calculation unit, for nodes without direct collaboration, calculates the multi-hop trust weight through the product of the weights of up to three relationship chains, and takes the maximum value as the final weight. The configuration algorithm of the multi-hop relationship calculation unit is that when there is a single multi-hop path, the product of all collaborative trust weights in the path is used to calculate the weight between nodes; when there are multiple multi-hop paths, the parallel calculation unit compares the product results of all paths and retains the maximum value; The node matching module includes: The load evaluation unit monitors the computing power load capacity of the comprehensive target node in real time; The topology generation unit searches for the smallest node combination that meets the task requirements through an iterative algorithm, and calculates the task load rate of each combination; The optimal decision-making unit selects the node topology with the lowest task load rate as the execution computing power node.

[0020] In summary, the advantages of the present invention are as follows: By constructing an interactive network model of collaborative trust relationships and attribute group clustering, multi-dimensional dynamic screening and optimal matching of nodes are realized, significantly improving the efficiency and reliability of computing power scheduling. Through trust collaboration and multi-hop path optimization, based on the calculation of collaboration history and multi-hop trust weights, the credibility of task allocation is enhanced, the participation of inefficient nodes is reduced, load balancing and efficient resource utilization are achieved. By iteratively searching for the node combination with the lowest load rate and combining the node importance correction coefficient, local overload is avoided and the global resource utilization rate is improved. Dynamic feedback and network adaptability: The trust relationship chain is adjusted in real time according to the task execution results, enabling the network to have self-optimization capabilities. Robustness in complex scenarios: The introduction of attribute group clustering quickly locks the target node set, reducing the search complexity in large-scale networks, effectively solving the problems of trust transfer, computing power resource collaboration, and dynamic load balancing in the computing power network, being applicable to scenarios such as cloud computing and edge computing, and having broad application value.

[0021] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A computing power scheduling method for a computing power network, characterized in that, Including: Obtain all computing power nodes in the computing power network and construct a computing power node interaction network model. The computing power node interaction network model includes a collaborative trust relationship and a clustering relationship. Based on the collaborative trust relationship, construct an interaction relationship chain between computing power nodes, and based on the clustering relationship, construct an attribute group clustering of computing power nodes; Obtain the computing power task initiation node and the computing power task attributes; Determine the target attribute group clustering of the computing power task based on the computing power task attributes, and record all the computing power nodes within the target attribute group clustering as the demand target nodes of the computing power task; Based on the computing power task initiation node, combined with the interaction relationship chain between computing power nodes, determine the execution trust target node of the computing power task; Find the intersection of the demand target nodes of the computing power task and the execution trust target nodes of the computing power task, and record it as the comprehensive target node; Combined with the balanced load, determine the execution computing power nodes of the computing power task from the comprehensive target nodes; Allocate the computing power task to the execution computing power nodes for task execution, and at the same time, combined with the feedback of the task execution results, adjust the interaction relationship chain between computing power nodes.

2. The computing power scheduling method of a computing power network according to claim 1, characterized in that, The specific method of constructing the interaction relationship chain between computing power nodes based on the collaborative trust relationship includes: Based on the physical distance between computing power nodes, attach an initial collaborative trust weight to each pair of all computing power nodes; If the computing power nodes have collaborated to complete the same computing power task, a collaborative relationship chain is formed between the computing power nodes. Based on the completion situation of the computing power tasks completed by the computing power nodes in collaboration and the historical collaborative trust weight, determine the collaborative trust weight between the computing power nodes; If no collaborative relationship chain is formed between the computing power nodes, but there are at most three collaborative relationship chains that can connect the computing power nodes, a multi-hop relationship chain is formed between the computing power nodes. Based on the collaborative trust weights between the computing power nodes that make up the multi-hop relationship chain, calculate the collaborative trust weight between the two computing power nodes at both ends of the multi-hop relationship chain; If neither a collaborative relationship chain nor a multi-hop relationship chain is formed between the computing power nodes, the collaborative trust weight between the computing power nodes is the initial collaborative trust weight; Summarize all the collaborative relationship chains and multi-hop relationship chains to obtain the interaction relationship chain between computing power nodes.

3. The computing power scheduling method of a computing power network according to claim 2, characterized in that, The determination method of the collaborative trust weight between the two computing power nodes at both ends of the multi-hop relationship chain is: If there is only one multi-hop relationship chain between two computing power nodes, the collaborative trust weight between the two computing power nodes is the product of the collaborative trust weights of the computing power nodes that make up the multi-hop relationship chain; If there are multiple multi-hop relationship chains between two computing power nodes, the collaborative trust weight between the two computing power nodes is the maximum value among the products of the collaborative trust weights of the computing power nodes of all multi-hop relationship chains.

4. The computing power scheduling method of a computing power network according to claim 1, characterized in that, The specific method of determining the execution trust target node of the computing power task based on the computing power task initiation node, combined with the interaction relationship chain between computing power nodes, includes: Screen out the computing power nodes with a collaborative trust weight greater than the trust threshold between the computing power task initiation node as the execution trust target nodes of the computing power task.

5. A computing power scheduling method for a computing power network according to claim 1, characterized in that The specific method of determining the execution computing power nodes of the computing power task from the comprehensive target nodes combined with the balanced load includes: Determine the computing power load demand of the computing power task and the computing power load capacity of each computing power node in the comprehensive target nodes; Set N = 1, and determine whether the sum of the computing power load capabilities of N computing power nodes in the comprehensive target node is greater than the computing power load demand of the computing power task. If so, combine the N computing power nodes into a node topology structure. If not, increment the value of N by 1 and return to continue the determination. If the value of N exceeds the number of computing power nodes included in the comprehensive target node, output an alarm signal to the background terminal; Use the ratio of the computing power load demand of the computing power task to the sum of the computing power load capabilities of all computing power nodes in the node topology structure as the task load rate of the node topology structure; Filter out the node topology structure with the minimum task load rate, and use all the computing power nodes therein as the execution computing power nodes for the computing power task.

6. The computing power scheduling method of a computing power network according to claim 5, characterized in that, When determining the execution computing power nodes for the computing power task, a node importance correction coefficient can also be introduced, specifically including: Based on the collaboration relationship chain of each node, combine the PageRank algorithm to determine the node importance of each computing power node in the computing power network; Then the formula for the task load rate of the node topology structure introducing the node importance correction coefficient is: ； Among them, is the task load rate of the i-th node topology introducing the node importance correction coefficient, X is the computing power load demand of the computing power task, is the computing power load capacity of the k-th computing power node in the i-th node topology, is the i-th node topology, K is the union of all node topologies, is the node importance of the k-th computing power node in the i-th node topology, k is the k-th computing power node in the i-th node topology, u is the u-th computing power node in K, is the node importance of the u-th computing power node in K, and are weight coefficients.

7. A computing power scheduling system for a computing power network, characterized in that A method for scheduling the computing power of a computing power network as described in any one of claims 1-6, including: A node management module, configured to obtain all computing power nodes in the computing power network and construct a computing power node interaction network model including collaboration trust relationships and clustering relationships; A task parsing module, configured to obtain the computing power task initiating node and the computing power task attributes, and determine the target attribute group clustering and its required target nodes according to the attributes; A trust evaluation module, calculating the collaboration trust weight between the computing power task initiating node and other nodes based on the interaction relationship chain, and screening the execution trust target nodes; A node matching module, calculating the intersection of the required target nodes and the execution trust target nodes to generate a comprehensive target node, and selecting the execution computing power nodes therefrom in combination with the load balancing algorithm; A dynamic adjustment module, updating the collaboration trust relationship and the interaction relationship chain according to the feedback of the task execution result.

8. The computing power scheduling system of a computing power network according to claim 7, wherein The node management module includes: A weight initialization unit, attaching an initial collaboration trust weight to the computing power node based on the physical distance; A collaboration relationship construction unit, generating a collaboration relationship chain and dynamic weights according to the completion situation of historical collaboration tasks; A multi-hop relationship calculation unit, for nodes without direct collaboration, calculating the multi-hop trust weight through the product of the weights of up to three relationship chains, and taking the maximum value as the final weight.

9. The computing power scheduling system of a computing power network according to claim 8, characterized in that, The multi-hop relationship calculation unit is configured: When there is a single multi-hop path, calculate the weight between nodes by multiplying all the collaboration trust weights in the path; When there are multiple multi-hop paths, compare the multiplication results of all paths through a parallel calculation unit and retain the maximum value.

10. The computing power scheduling system of a computing power network according to claim 7, characterized in that, The node matching module includes: A load evaluation unit, monitoring the computing power load capacity of the comprehensive target node in real time; A topology generation unit, searching for the minimum node combination that meets the task requirements through an iterative algorithm, and calculating the task load rate of each combination; An optimal decision-making unit, selecting the node topology with the lowest task load rate as the execution computing power node.

Images