Edge computing power sharing method and device, equipment and storage medium

Abstract

The invention discloses an edge computing power sharing method, an edge computing power sharing device, edge computing power sharing equipment and a medium, and the edge computing power sharing method comprises the following steps: connecting edge nodes through a decentralized P2P network, constructing a network connection module, and realizing computing power resource sharing among the nodes; collecting and monitoring computing power resource state data of each edge node through a computing power resource management module, judging nodes in an idle state through an evaluation algorithm, and adding computing power resources of the nodes into a shared pool; the node matching module is used for screening out an optimal matching node from the shared pool by utilizing a preset screening algorithm according to task demand information released by a computing power demand side; and the distribution execution module sends the task instruction to the selected optimal node from the cloud, and after the task is executed, the node releases the computing power resource and performs expense settlement through the charging module. According to the method, the utilization efficiency of computing power resources is optimized through decentralized node management and an intelligent matching mechanism, so that edge computing resources are more easily integrated and utilized.

Description

Technical Field

[0001] The present invention relates to the field of computing power sharing, and particularly to a method, device, equipment and storage medium for edge computing power sharing. Background Art

[0002] With the rapid development of technologies such as the Internet of Things, artificial intelligence, and 5G communication, data has shown explosive growth. The traditional cloud computing model faces problems such as high network bandwidth pressure, high data transmission latency, and increased privacy and security risks. As an emerging computing paradigm, edge computing sinks computing, storage, and network resources to the network edge close to the data source or user, which can effectively relieve the pressure on the cloud computing center, reduce network latency, and improve the real-time performance and security of data processing. However, edge devices usually have limited computing resources and are difficult to independently complete complex computing tasks. Moreover, edge computing resources are distributed dispersedly and the utilization rate is uneven, making it difficult to meet the growing computing needs. Therefore, there is an urgent need for an edge computing power sharing method to solve the problem that existing edge computing resources are difficult to be integrated and applied. Summary of the Invention

[0003] Embodiments of the present invention provide a method, device, equipment and storage medium for edge computing power sharing, aiming to solve the problem that existing edge computing resources are difficult to be integrated and applied.

[0004] In a first aspect, embodiments of the present invention provide an edge computing power sharing method, which is applied to an edge computing power sharing system including a network connection module, a computing power resource management module, a node matching module, a distribution execution module, and a charging module. The method includes: constructing a decentralized P2P network through the network connection module to connect edge nodes;

[0005] collecting the computing power resource status data of the edge nodes through the computing power resource management module;

[0006] evaluating the computing power resource status data of the edge nodes according to a preset computing power status evaluation algorithm, and when the computing power resources of the edge nodes are determined to be in an idle state, adding the edge nodes to the sharing pool;

[0007] according to the task requirement information published by the computing power demand side, screening out the optimal matching nodes from the sharing pool by the node matching module according to a preset screening algorithm;

[0008] sending the task instruction of the computing power demand side from the cloud to the optimal matching node through the distribution execution module, and the optimal matching node executes the task;

[0009] When the optimal matching node completes the task, releasing its computing power resources through the computing power resource management module and performing cost settlement through the charging module.

[0010] In a second aspect, an embodiment of the present invention further provides an edge computing power sharing device, including units for executing the edge computing power sharing method described above.

[0011] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to execute the steps of the above-mentioned edge computing power sharing method.

[0012] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program, and the computer program includes program instructions that can implement the steps of the above-mentioned edge computing power sharing method when executed by a processor.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0014] In the technical solution of the present invention, edge nodes are connected through a decentralized P2P network to construct a network connection module to achieve sharing of computing power resources between nodes; the computing power resource management module collects and monitors the status data of computing power resources of each edge node, and the evaluation algorithm determines the nodes in the idle state and adds their computing power resources to the sharing pool; the node matching module uses a preset screening algorithm to screen out the optimal matching nodes from the sharing pool according to the task requirement information published by the computing power demand side; the distribution execution module sends the task instructions from the cloud to the selected optimal node, and after the task is completed, the node releases the computing power resources and settles the fees through the billing module. This method optimizes the utilization efficiency of computing power resources through a decentralized node management and intelligent matching mechanism, making it easier to integrate and utilize edge computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a flowchart of the edge computing power sharing method provided by the present invention;

[0017] Figure 2 It is the first sub-flowchart of the edge computing power sharing method provided by the present invention;

[0018] Figure 3 It is the second sub-flowchart of the edge computing power sharing method provided by the present invention;

[0019] Figure 4 It is the third sub - flowchart of the edge computing power sharing method provided by the present invention;

[0020] Figure 5 It is the fourth sub - flowchart of the edge computing power sharing method provided by the present invention;

[0021] Figure 6 It is the fifth sub - flowchart of the edge computing power sharing method provided by the present invention;

[0022] Figure 7 It is a schematic block diagram of the unit of the edge computing power sharing device provided by the present invention;

[0023] Figure 8 It is a schematic block diagram of the computer device provided by the embodiments of the present invention; Detailed implementation manners

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0025] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprise" indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations.

[0026] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in this specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0027] It should be further understood that the term "and / or" used in this specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the related listed items, and includes these combinations.

[0028] In order to solve the problem that it is difficult to integrate and apply edge computing resources in the prior art, the present invention proposes an edge computing power sharing method, which is applied to an edge computing power sharing system, including a network connection module, a computing power resource management module, a node matching module, a distribution execution module, and a billing module.

[0029] In the embodiments of the present invention, the so-called network connection module is responsible for constructing a decentralized P2P network and connecting all edge nodes. Specifically, the network connection module is responsible for ensuring that the edge nodes in the system can communicate with each other without relying on a centralized server. Through the decentralized P2P network, nodes can directly exchange data and allocate tasks with other nodes, thereby enhancing the robustness and scalability of the system. This enables efficient sharing of computing power resources among edge nodes and real-time response to task requirements. The construction of the P2P network is the basis for distributed resource sharing in the edge computing system.

[0030] In the embodiments of the present invention, the so-called computing power resource management module is responsible for collecting and monitoring the status data of the computing power resources of each edge node. This module continuously collects the resource usage of nodes, such as CPU, memory, GPU, bandwidth, etc., and stores this data in a centralized database or a distributed storage system. Through real-time monitoring and data collection, the computing power resource management module can evaluate the computing power utilization rate of each node, and then determine which nodes' computing power resources are idle or under low load. These idle resources can then be added to the shared pool for task requesters to use. The computing power resource management module is the core part of resource scheduling and optimization, ensuring that the system can perform intelligent resource allocation according to the real-time situation.

[0031] The so-called node matching module is responsible for screening the most suitable edge nodes to execute tasks according to the task requirement information published by the computing power requesters. When the computing power requesters publish tasks, these tasks usually include specific requirements for computing resources, such as computing power, latency, geographical location, etc. The node matching module uses preset screening algorithms, such as matching algorithms based on node performance and geographical location, to select the most suitable nodes from the shared pool to execute tasks according to these task requirements. Through precise matching, the node matching module can ensure the maximization of task execution efficiency and resource utilization rate, while improving the response speed and flexibility of the system.

[0032] The function of the so-called distribution and execution module is to send task instructions from the cloud to the edge node with the best match and ensure the smooth progress of task execution. Once the node matching module selects the edge node with the best match, the distribution and execution module is responsible for delivering the task instructions to this node. This module also needs to handle potential problems in task transmission, such as network latency, transmission errors, etc., to ensure that the task instructions can reach the target node reliably. In addition, the distribution and execution module will track the execution status of the task and provide real-time feedback to the cloud or the computing power requester to ensure that the task is completed on time. If an exception occurs during task execution, the distribution and execution module can respond quickly, make adjustments or reallocate tasks.

[0033] The so-called billing module is responsible for calculating and settling the fees of the computing power demanders. After the task is completed, the billing module settles the fees according to the preset billing rules, such as by resource usage, computing time, task complexity, etc. The billing module usually needs to consider multiple factors, including the computing power resources used, the length of time for task execution, resource consumption, etc. To ensure the transparency and accuracy of fee settlement, the billing module may use technologies such as blockchain or smart contracts to record the transaction process and prevent tampering and fraud. The billing module not only settles accounts with the computing power demanders but also generates detailed bill reports for all parties to view and audit.

[0034] Based on the edge computing power sharing system, with reference to Figures 1 to 6 , the efficient sharing and application of edge computing power are realized through the following edge computing power sharing method.

[0035] S110. Build a decentralized P2P network through the network connection module to connect edge nodes;

[0036] Build a decentralized P2P network through the network connection module. When each edge node joins the network, it broadcasts its presence information to other nodes in the network using its own identifier. Among them, edge nodes such as edge servers or devices, and identifiers such as device IDs or public network IPs. This information is transmitted through the P2P network to ensure that each node can establish a connection with other nodes. Use the STUN protocol and TURN protocol to realize network address discovery and communication, so that even in a NAT environment, effective connection between nodes can be achieved.

[0037] Furthermore, with reference to Figure 2 , step S110 includes:

[0038] S111. Establish the network connection of the edge node through the network connection module according to the distributed node discovery mechanism;

[0039] S112. Use the STUN protocol through the network connection module to achieve NAT penetration and cooperate with the ICE protocol for connectivity detection;

[0040] S113. Dynamically allocate the communication parameters of the edge node through the network connection module relying on the virtual network controller.

[0041] In an embodiment of the present invention, the network connection module initializes the connections between edge nodes according to the distributed node discovery mechanism. When a node joins the network, it announces its existence to the entire P2P network through a broadcast mechanism or other distributed discovery protocols, such as mDNS, Bonjour, etc. These edge nodes search for other nodes according to predetermined policies, such as geographical location, resource load, latency, and other factors, to ensure that each node can discover and establish communication connections with other nodes. Specifically, the distributed node discovery mechanism ensures that each edge node knows the status of other nodes in the network and their resource availability. In this way, edge nodes can achieve effective interconnection and intercommunication without relying on a centralized server. This is the key technology for the successful operation of P2P networks.

[0042] In a P2P network, many edge nodes may be within a local area network (LAN) and communicate with the external Internet through a NAT (Network Address Translation) device. In this case, directly connecting two nodes may be restricted by NAT, resulting in an inability to establish an effective P2P connection. To solve this problem, the network connection module uses the STUN protocol for NAT traversal. Through the STUN protocol, a node can discover its public IP and port information, thus helping nodes behind NAT devices find each other's communication paths. In addition, the ICE protocol is used to further improve connectivity. After NAT traversal is completed, the ICE protocol uses a STUN or TURN server to detect and select the optimal network path. The ICE protocol can effectively handle the complexity in different network environments, ensuring the maximization of connectivity between edge nodes and establishing stable connections even in complex network topologies. Through the synergistic effect of the STUN protocol and the ICE protocol, the network connection module can dynamically select the best communication path during NAT traversal, thus ensuring unobstructed communication between nodes and avoiding connection failures caused by network structure problems.

[0043] In a P2P network, as the number of nodes increases and resource requirements change, the network topology may change dynamically. For this reason, the network connection module relies on a virtual network controller (VNC) to dynamically allocate and adjust the communication parameters of edge nodes. When a node joins or leaves the network, the virtual network controller will reallocate the relevant communication parameters to ensure that the newly joined node can seamlessly integrate into the network and the old node can release resources in a timely manner when it exits, avoiding network bottlenecks or resource waste. Through this dynamic adjustment mechanism, the network connection module can optimize the performance of the entire P2P network and ensure the stability and efficiency of the edge computing power sharing system.

[0044] S120. Collect the computing power resource status data of the edge nodes through the computing power resource management module;

[0045] After establishing a connection, the computing power resource management module starts monitoring the computing power resource status of each edge node. Nodes regularly collect data on computing resources through a lightweight monitoring program, including CPU usage, memory occupancy, network bandwidth, GPU status, etc. The so-called lightweight monitoring program can be a self-developed performance monitoring tool. The collected data is encrypted and uploaded to the cloud or a shared database, and is analyzed in real time. This process ensures that the resource status of all nodes is always monitored, facilitating subsequent task scheduling and resource management.

[0046] Further, referring to Figure 3 , step S120 includes:

[0047] S121. Real-time collect the computing power parameters of the edge node through the monitoring program carried on the computing power resource management module, where the computing power parameters include CPU usage, GPU usage, memory usage, and network bandwidth;

[0048] S122. Encrypt the computing power parameters by adopting the AES-256 algorithm, and report them to the cloud after GZIP compression.

[0049] In an embodiment of the present invention, a dedicated monitoring program is built into the computing power resource management module for real-time collecting the computing power parameters of the edge node. The computing power parameters of the edge node include but are not limited to CPU usage, GPU usage, memory usage, network bandwidth, etc. These parameters can reflect the current computing load and resource usage of the edge node, and are key data for scheduling resources and optimizing task allocation. The real-time monitoring data of these computing power parameters will be continuously collected and uploaded to the cloud for analysis and storage when necessary. In this way, the cloud can obtain the resource status of the edge node at any time, and then optimize the scheduling and load balancing of the computing power resources.

[0050] To ensure that the computing power resource status data is not leaked or tampered with during the transmission process, the computing power resource management module uses the AES-256 encryption algorithm to encrypt the collected computing power parameters. AES-256 is a symmetric encryption algorithm, which is widely used in occasions requiring high security and has strong encryption strength.

[0051] Specifically, before uploading the computing power data to the cloud, the monitoring program will first encrypt the data by using the AES-256 algorithm. The encrypted computing power parameter data will be encapsulated into ciphertext, and only the cloud receiver with the correct decryption key can restore the original computing power parameter data. Through this encryption measure, the transmission of the computing power data is strongly protected, preventing security risks such as network attacks or data eavesdropping.

[0052] In addition to encryption processing, the computing power resource management module also compresses the encrypted data using the GZIP compression algorithm. In practical applications, since there may be certain limitations on the network bandwidth between edge nodes and the cloud, compressing the data can not only save network bandwidth but also improve data transmission efficiency. Especially when the number of nodes is large and the amount of computing power parameter data is huge, GZIP compression can significantly enhance the speed and throughput of data upload. By compressing the encrypted data, the computing power resource management module further optimizes the efficiency and security of the data transmission process. This step can ensure that sensitive information is not leaked during data upload and that network resources are efficiently utilized for data exchange.

[0053] S130. Evaluate the computing power resource status data of the edge node according to a preset computing power status evaluation algorithm. When the computing power resources of the edge node are determined to be in an idle state, add the edge node to the sharing pool.

[0054] Through a preset computing power status evaluation algorithm, the system dynamically evaluates the computing power resources of each edge node. This algorithm is based on multiple resource usage metrics (such as CPU load, memory occupancy, etc.) to determine which nodes' computing power resources are in an idle state. Specifically, when the CPU and memory usage rates of a certain node are lower than a certain threshold, the system marks the computing power resources of this node as idle and adds it to the sharing pool. The resources in the sharing pool can be used by other nodes with demands to ensure the maximization of resource utilization.

[0055] Further, referring to Figure 4 , step S130 includes:

[0056] S131. The cloud decrypts the encrypted and compressed computing power parameters and calculates the computing power parameters through a dynamic threshold algorithm to identify the node load status of the edge node.

[0057] S132. When the shareable idle computing power resources of the edge node meet the preset conditions for computing power sharing, mark the edge node.

[0058] S133. Incorporate all the marked edge nodes into the sharing pool as candidate computing power providers.

[0059] In an embodiment of the present invention, the cloud first receives the encrypted and compressed computing power resource data uploaded from the edge nodes. Since these data have been encrypted by AES-256 and compressed by GZIP, the cloud system needs to perform two operations to restore the original computing power parameters. After decrypting and decompressing to obtain the computing power parameters, the cloud system will apply the dynamic threshold algorithm to evaluate the computing power resources of the nodes. The dynamic threshold algorithm calculates the load status of the nodes based on a comprehensive analysis of the current computing power parameters of the nodes. The cloud will set a dynamic threshold for each computing power parameter according to the actual requirements and resource conditions. These thresholds are not static, but are dynamically adjusted according to the operating conditions of the nodes, load changes, and changes in the resource requirements of the shared pool. For example, the dynamic threshold of the CPU may change with the fluctuations of the system load. When the load is low, the threshold will be correspondingly reduced, thereby increasing the possibility of sharing computing power nodes. Through the dynamic threshold algorithm, the cloud will calculate the actual load of each computing power parameter in real time and compare it with the corresponding dynamic threshold. If the utilization rate of a certain computing power parameter is lower than its preset threshold, that is, the resources are not fully utilized, then the node may be regarded as being in an idle state. The dynamic threshold algorithm not only considers a single computing power parameter, but also comprehensively evaluates the usage of various resources. For example, if the CPU and memory utilization rates of a certain node are low, but the network bandwidth is in a high load state, then the node may still not be suitable to be added to the shared pool unless its bandwidth load is lower than a certain threshold. The result of the comprehensive evaluation will provide the cloud with an accurate node load status.

[0060] Through the calculation of the dynamic threshold algorithm, the cloud system can obtain the load status of each edge node. When the computing resources of a certain node are in an idle state and the utilization rates of its various resources are lower than the preset computing power sharing conditions, the cloud system will determine that the node has idle computing power resources that can be shared and can be included in the shared pool as a computing power provider. When the computing power resources of the edge node are evaluated as being in an idle state, the cloud will mark the node, that is, mark its identity as the "sharable" state. The marking process can be achieved by adding a "sharing status" marking field to the status database of the node for subsequent shared pool management. In addition, the cloud will further screen the nodes according to the preset sharing conditions. These conditions ensure that the nodes added to the shared pool have sufficient idle computing power resources and can provide stable computing power when needed.

[0061] Once an edge node is marked as shareable, the cloud will include this node in the sharing pool. The sharing pool is a centralized computing power resource pool that contains all available edge computing nodes, and these nodes can provide computing resources to other tasks or nodes. The cloud will add the information of the nodes marked as "shareable" to the sharing pool database as candidate nodes for computing power provision. These nodes can respond to the computing power requirements of other nodes or tasks at any time according to the scheduling strategy of the cloud. The cloud ensures that the computing power resources of the nodes in the pool are fully utilized by continuously monitoring the node status in the sharing pool. If the load of a node changes, the cloud will re-evaluate the sharing eligibility of this node and make necessary dynamic adjustments.

[0062] S140. According to the task requirement information published by the computing power demand side, the optimal matching node is selected from the sharing pool by the node matching module according to a preset screening algorithm;

[0063] The node matching module intelligently selects the optimal matching node in the sharing pool according to the task requirement information published by the computing power demand side. For example, the computing power demand side may require specific computing capabilities, low latency, geographical proximity, etc. The node matching module selects the nodes that meet the requirements from the sharing pool according to these conditions through a preset screening algorithm, such as a weighted matching algorithm based on multi-dimensional conditions. The selected optimal matching node will be used to execute the tasks of the computing power demand side.

[0064] Further, referring to Figure 5 , step S140 includes:

[0065] S141. Obtain the task requirement information submitted by the demand side through the node matching module, and the task requirement information includes a description of computing power requirements, network latency tolerance, and cost budget;

[0066] S142. Construct a dynamic matching model according to the task requirement information through the node matching module;

[0067] S143. Perform feature weighted scoring on all the candidate nodes for computing power provision in the sharing pool through the dynamic matching model;

[0068] S144. Perform weighted scoring based on the computing power performance matching degree, network latency, and cost budget compliance through the dynamic matching model, and select the node with the highest comprehensive score as the optimal matching node.

[0069] In an embodiment of the present invention, the computing power requester will publish its task requirement information on a cloud platform or an edge computing platform. The task requirement information usually includes a description of computing power requirements, network latency tolerance, and cost budget. After the requester submits the task requirement information, the node matching module will obtain this information through an interface and pass it to the dynamic matching model as the basis for subsequent screening of the optimal nodes.

[0070] Based on the task requirement information submitted by the requester, the node matching module constructs a dynamic matching model. The purpose of this model is to comprehensively consider multiple factors and perform multi-dimensional node matching according to the requirements of the task. The dynamic matching model first considers the degree of matching between the computing power requirements of the task and the resources of the candidate nodes in the shared pool. For tasks with network latency sensitivity, the matching model will evaluate the network connection of each node, taking into account the maximum latency value required by the task. If the task has strict requirements for latency, the model will preferentially select edge nodes located close to the requester, which can minimize network latency. After completing the matching of computing power and latency, the matching model will also screen the nodes according to the budget information of the requester. Each node in the shared pool has a corresponding rental cost for computing power resources, and these costs may be based on factors such as the usage of computing resources, the geographical location of the node, and the service duration. The dynamic matching model will screen out nodes that meet the budget range according to the budget requirements of the requester to ensure cost control by the requester.

[0071] Once the dynamic matching model is constructed, it will perform feature weighted scoring on all candidate nodes in the shared pool according to the task requirement information. Each matching factor in the dynamic matching model will be assigned different weights according to its impact on task completion. For example, for some tasks, computing power performance may occupy a higher weight, while for real-time video processing tasks, network latency may be more important, so the weights need to be flexibly adjusted according to the nature of the task. Each candidate node will be scored according to the degree of compliance of its various features with the task requirements. Specifically, the features of each candidate node will be compared with the standards of the task requirements to calculate a matching score. This score will take into account the weighted scores of various factors. For example, a node with a higher computing power matching score will get a higher score in this item, and a node with lower latency will get a higher weight in the latency score. Finally, through the weighted scores of various features, a comprehensive score will be generated, indicating the degree of matching between each node and the task requirements. The higher the comprehensive score of a node, the more it meets the task requirements of the requester in multiple matching dimensions.

[0072] Based on the weighted scoring of the dynamic matching model, the node matching module will screen all candidate nodes in the shared pool and select the optimal matching node with the highest comprehensive score. Once the optimal matching node is selected, the node matching module will send a task allocation request to this node in the shared pool and schedule specific computing tasks to this node for execution according to the task requirements. During the node execution process, the cloud system will also monitor the execution progress, resource usage, latency, etc. of the task to ensure that the task can be completed as expected.

[0073] S150. Send the task instruction of the computing power requester from the cloud to the optimal matching node through the distribution and execution module, and the optimal matching node executes the task;

[0074] Once the optimal matching node is selected, the distribution and execution module sends the task instruction from the cloud to this node. The task can be any computationally intensive operation, such as data processing, artificial intelligence model training, etc. This module ensures the reliability of the task instruction transmission and handles possible network latency problems. During the task execution process, the distribution and execution module will monitor the progress of the task in real time to ensure that the node completes the computing task according to the task requirements.

[0075] Further, referring to Figure 6 , step S150 includes:

[0076] S151. Distribute the task instruction of the computing power requester to the optimal matching node through the encrypted instruction channel of the cloud by the distribution and execution module;

[0077] S152. Use the dynamic key negotiation by the security module to generate a temporary session key to encrypt the communication link, and after verifying the requester's permission according to the RBAC model, deploy the task to the containerized isolation environment for execution;

[0078] S153. Establish a P2P direct connection channel between the optimal matching node and the computing power requester via the network connection module, report the task execution progress to the cloud through heartbeat packets, and immediately send a status notification when the task is completed or abnormally terminated;

[0079] S154. Report the task execution progress reported by the optimal matching node to the cloud in real time.

[0080] In an embodiment of the present invention, the cloud platform sends task instructions from the cloud to the optimal matching node through the distribution execution module to initiate the execution of the task. To ensure the security of the task instructions during transmission, the distribution execution module encrypts the task instructions through the encrypted instruction channel of the cloud and then transmits them to the optimal matching node. Task instructions usually include information such as computing resource allocation, data input / output location, and task execution time. After the instruction transmission, the optimal matching node will decrypt and parse the task instructions and prepare to execute the corresponding task.

[0081] After the task instructions are sent to the optimal matching node, the optimal matching node needs to establish a stable and reliable communication channel with the cloud of the requester. To reduce network transit and improve communication efficiency, the optimal matching node establishes a P2P direct connection channel with the requester through the network connection module to ensure that data can be directly transmitted between nodes without passing through an intermediate server. The optimal matching node requests to establish a P2P direct connection with the cloud through the network connection module, and the cloud platform will conduct connection negotiation based on factors such as network topology and node location. Finally, the task instructions will be transmitted in real time on the P2P channel, thus ensuring that the task execution instructions can reach the optimal matching node quickly and reliably.

[0082] To further enhance the security of the communication link, this embodiment introduces a dynamic key negotiation mechanism. A temporary session key is generated by the security module and used to encrypt the communication between the requester and the optimal matching node. The session key is only valid during the current session, ensuring that even if the communication link is eavesdropped, the eavesdropper cannot decrypt the transmitted content. Once the temporary session key is generated, the security module will use this key to encrypt the communication link, ensuring the confidentiality and integrity of subsequent data such as task instructions and execution progress reports during transmission. After the task instructions are transmitted and encrypted, the security module also needs to verify the permissions of the requester according to the Role-Based Access Control (RBAC) model to ensure that it has the right to execute the task. In this way, user permissions can be managed more flexibly and efficiently, ensuring that only authorized users can execute specific tasks. The security module first obtains the role and corresponding permission information of the requester from the cloud database, and then compares this information with the permission requirements in the task instructions. If the requester has the permission to execute the task, the verification passes; otherwise, the task instructions will be rejected and an error message indicating insufficient permissions will be returned to the requester. After the permission verification is completed, the optimal matching node deploys the task to a containerized isolation environment for execution. Containerization technology ensures that each task runs in an independent environment without interference, improving the security and resource utilization of the system. Specifically, Docker, Kubernetes, etc. are used. The optimal matching node first creates a new container instance to execute the current task. The container instance has an independent file system, network interface, and resource quota, ensuring that the task runs in an isolated environment. The task instructions that have been transmitted through the encrypted instruction channel and decrypted are loaded into the container to prepare for execution.

[0083] Once the task starts to execute, the optimal matching node needs to provide real-time feedback on the task execution progress to the cloud platform. To ensure the reliability of task execution and the real-time nature of monitoring, the optimal matching node will regularly report the task execution progress to the cloud through the heartbeat packet mechanism. During the task execution process, the optimal matching node periodically sends heartbeat packets to the cloud to ensure that the cloud can obtain the latest status of the task in real time. In the heartbeat packet, the optimal matching node will transmit the current execution status of the current task to the cloud, including: the current execution progress of the task, the usage of computing resources, the amount of data processed, or the steps of the calculation. The cloud will update the execution status of the task according to the heartbeat packet information and provide corresponding feedback and scheduling instructions based on the task progress.

[0084] During the task execution process, the optimal matching node not only regularly sends heartbeat packets to report the progress, but also immediately sends a status notification to the cloud platform when the task is completed or abnormally terminated, so that corresponding measures can be taken in a timely manner. When the optimal matching node successfully completes the task, the task completion notification will send a task completion status notification to the cloud through the P2P direct connection channel. This notification usually includes the completion time, results of the task, and any warnings or abnormal situations that may occur during the execution process. The task abnormal termination notification will be used by the cloud for subsequent task data processing, result analysis, and recording. If any exceptions occur during the task execution, such as hardware failures, network disconnections, timeouts, etc., the optimal matching node will immediately send an error message to the cloud through the status notification. The cloud platform can adopt corresponding recovery or retry mechanisms according to the error type. If the task fails, the cloud can also notify the computing power requester so that it can decide whether to reschedule or select other nodes to execute according to the importance of the task. Additionally, in the case of abnormal task termination, the optimal matching node can also upload detailed error information to the cloud through the log recording function, helping the cloud platform analyze and quickly locate the cause of the problem for repair or optimization.

[0085] Throughout the entire task execution process, the cloud platform can monitor the task execution situation in real time through the received heartbeat packets and status notifications. If any exceptions occur during the task execution, the cloud can immediately intervene and make adjustments to ensure that the task can be successfully completed. Through this mechanism, the cloud platform can timely adjust the task scheduling strategy to achieve more flexible and efficient edge computing power resource management.

[0086] S160. After the optimal matching node completes the task, it releases its computing power resources through the computing power resource management module and settles the fees through the billing module.

[0087] After the optimal matching node completes the task, the node will release the computing power resources it occupies through the system and return to a state available for other tasks. The billing module settles the computing power requester according to the preset billing rules (such as calculating time by task, resource usage, etc.). By using blockchain technology or smart contracts, the transparency and immutability of the billing process are ensured, guaranteeing the fairness of the transaction between both parties.

[0088] In an embodiment of the present invention, a task allocation logic of the intelligent matching module adopts a multi-dimensional weighted scoring model, and the specific formula system is constructed as follows:

[0089] First, device screening is performed, and the following Boolean logic formula is used for device screening conditions:

[0090]

[0091] Among them, Denote the available resources (number of CPU cores / GB of memory / GB of GPU video memory / disk resources) of device i, which are from the real-time data of the resource monitoring module; R r Denote the task requirement resources, which come from the task specifications submitted by the requester; U i Denote the current load rate (%) of device i, which is collected by the resource monitoring agent; θ U Denote the load threshold, which is a preset security parameter of the system. By default, it can be 80% or can be changed according to requirements when specifically used; C i Is the quotation of device i, which is set by the provider on the client side; B is the budget of the requester, which comes from the task release parameters.

[0092] After the devices are screened, the formula of the comprehensive scoring model of the selected device i is as follows:

[0093] S i = w M ·M i + w L ·L′ i + w C ·C′ i + w G ·G′ i

[0094] Among them, S i Is the total score of device i; M i Is the computing power matching degree of device i; L′ i Is the network quality factor of device i; C′ i Is the economic benefit factor of device i; G′ i Is the geographical affinity of device i. wi is the weight of each corresponding parameter above, where:

[0095] w M + w L + w C + w G = 1

[0096] The initial value is preset as (0.4, 0.3, 0.2, 0.1) and can be dynamically adjusted according to the task type.

[0097] The formula for the computing power matching degree is:

[0098]

[0099] Among them, k r Is the resource type weight, which reflects the importance difference of heterogeneous resources. This weight is usually (CPU: 0.5, MEM: 0.3, GPU: 0.2). Through logarithmic function processing, it can prevent the score from being distorted due to the excess of a single resource.

[0100] The formula for the network quality factor is as follows:

[0101]

[0102] Where D i is the end-to-end delay (ms), which is measured by the probe of the P2P module; P i is the packet loss rate (%), which is obtained from the data monitored by the network quality; D max and P max are respectively the maximum thresholds allowed by the system, and can be defaulted to (500ms, 5%) or set independently according to requirements when specifically used. λ is the attenuation coefficient, which is used to control the delay influence gradient, and is specifically defaulted to 0.005 or set independently according to requirements.

[0103] The formula for the economic benefit factor is as follows:

[0104]

[0105] Where is the average quotation of the current online devices; σ C is the standard deviation of the quotations; α is the price sensitivity coefficient, and the larger the value, the stronger the preference for low prices.

[0106] The formula for the geographical affinity is as follows:

[0107]

[0108] Where loc refers to the longitude and latitude coordinates, which are obtained through device GPS / IP positioning; γ: distance attenuation factor; δ is the reward value for the same region; reg is the administrative region code, which is used to determine whether it is in the same city or the same province.

[0109] After completing the total score of the devices, it is also necessary to select the optimal device. The formula for the optimal device selection is as follows:

[0110]

[0111] After the computing power demand side provides a task, the cloud will select the device i with the highest score * , and if there is a tie, it will preferentially select the one with a higher M i .

[0112] Figure 7 is a schematic block diagram of an edge computing power sharing device 600 provided by an embodiment of the present invention. As Figure 7 shown, corresponding to the above edge computing power sharing method, the present invention also provides an edge computing power sharing device 600. The edge computing power sharing device includes units for executing the above edge computing power sharing device method, and the device can be configured in terminals such as desktop computers, tablet computers, smart phones, etc. Specifically, please refer to Figure 7, the edge computing power sharing device includes:

[0113] A network construction unit 610 for constructing a decentralized P2P network connection with edge nodes through the network connection module;

[0114] A computing power monitoring unit 620 for collecting the computing power resource status data of the edge nodes through the computing power resource management module;

[0115] A shared pool editing unit 630 for evaluating the computing power resource status data of the edge nodes according to a preset computing power status evaluation algorithm, and adding the edge nodes to the shared pool when the computing power resources of the edge nodes are judged to be in an idle state;

[0116] A task matching unit 640 for, according to the task requirement information published by the computing power demand side, screening out the optimal matching nodes from the shared pool through the node matching module according to a preset screening algorithm;

[0117] A task execution unit 650 for sending the task instructions of the computing power demand side from the cloud to the optimal matching nodes through the distribution and execution module, and the optimal matching nodes execute the tasks;

[0118] A task settlement unit 660 for, when the optimal matching nodes complete the tasks, releasing their computing power resources through the computing power resource management module and performing cost settlement through the billing module.

[0119] The network construction unit 610 includes:

[0120] A node network connection construction unit for establishing the network connection of the edge nodes through the network connection module according to the distributed node discovery mechanism;

[0121] A NAT penetration unit for implementing NAT penetration through the network connection module using the STUN protocol and cooperating with the ICE protocol for connectivity detection;

[0122] A communication parameter allocation unit for dynamically allocating the communication parameters of the edge nodes through the network connection module relying on the virtual network controller.

[0123] The computing power monitoring unit 620 includes:

[0124] A computing power parameter acquisition unit for real-time collecting the computing power parameters of the edge nodes through a monitoring program carried on the computing power resource management module, where the computing power parameters include CPU usage rate, GPU usage rate, memory usage rate, and network bandwidth;

[0125] The uplink communication unit is used to encrypt the computing power parameters by using the AES-256 algorithm, and report them to the cloud after GZIP compression.

[0126] The shared pool editing unit 630 includes:

[0127] The load status recognition unit is used for the cloud to decrypt the encrypted and compressed computing power parameters, calculate the computing power parameters by using the dynamic threshold algorithm, and recognize the node load status of the edge node;

[0128] The idle node marking unit is used to mark the edge node when the shareable idle computing power resources of the edge node meet the preset conditions for computing power sharing;

[0129] The shared pool injection is used to include all the marked edge nodes in the shared pool as candidate computing power providers.

[0130] The task matching unit 640 includes:

[0131] The demand acquisition unit is used to obtain the task demand information submitted by the demander through the node matching module, and the task demand information includes computing power demand description, network delay tolerance and cost budget;

[0132] The dynamic matching model construction unit is used to construct a dynamic matching model according to the task demand information through the node matching module;

[0133] The feature weighted scoring unit is used to perform feature weighted scoring on all the candidate computing power providers in the shared pool through the dynamic matching model;

[0134] The optimal node matching unit is used to perform weighted scoring based on the computing power performance matching degree, network delay and cost budget compliance through the dynamic matching model, and select the node with the highest comprehensive score as the optimal matching node.

[0135] The task execution unit 650 includes:

[0136] The task instruction transmission unit is used to distribute the task instructions of the computing power demander to the optimal matching node through the encrypted instruction channel of the cloud by the distribution execution module;

[0137] The security protection unit is used to encrypt the communication link by using the dynamic key negotiation to generate a temporary session key through the security module, and after verifying the demander's permission according to the RBAC model, deploy the task to the containerized isolation environment for execution;

[0138] An execution monitoring unit, configured to establish a P2P direct connection channel with the computing power demand side via the optimal matching node through the network connection module, report the task execution progress to the cloud through heartbeat packets, and immediately send a status notification when the task is completed or abnormally terminated;

[0139] A final reporting unit, configured to report the task execution progress to the cloud in real time by the optimal matching node.

[0140] The above-mentioned edge computing power sharing device 600 can be implemented in the form of a computer program, and this computer program can run on a computer device as shown in Figure 8 shown.

[0141] Please refer to Figure 8 , Figure 8 which is a schematic block diagram of a computer device provided by an embodiment of the present application. The computer device 500 can be a terminal or a server. Among them, the terminal can be an electronic device with communication functions such as a desktop computer, a tablet computer, a smart phone, etc. The server can be an independent server or a server cluster composed of multiple servers.

[0142] Refer to Figure 8 , the computer device 500 includes a processor 502, a memory, and a network interface 505 connected through a system bus 501. Among them, the memory can include a non-volatile storage medium 503 and an internal memory 504.

[0143] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions, and when these program instructions are executed, the processor 502 can be made to execute an edge computing power sharing method.

[0144] The processor 502 is configured to provide computing and control capabilities to support the operation of the entire computer device 500.

[0145] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can be made to execute an edge computing power sharing method.

[0146] The network interface 505 is configured to communicate with other devices through a network. Those skilled in the art can understand that Figure 7 the structure shown in

[0147] Among them, the processor 502 is used to run the computer program 5032 stored in the memory to implement the steps of the above method.

[0148] It should be understood that in the embodiments of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0149] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, and the storage medium is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above method.

[0150] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by the processor, the processor executes the steps of the above method.

[0151] The storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, an optical disk or other various computer-readable storage media that can store program codes.

[0152] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0153] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0154] The steps in the method embodiments of the present invention can be adjusted, combined, and deleted according to actual needs. The units in the device embodiments of the present invention can be combined, divided, and deleted according to actual needs. In addition, the functional units in each embodiment of the present invention can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0155] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention.

[0156] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for sharing edge computing power, characterized in that: Applied to the edge computing power sharing system, including a network connection module, a computing power resource management module, a node matching module, a distribution execution module, and a billing module, the method includes: Constructing a decentralized P2P network to connect edge nodes through the network connection module; Collecting computing power resource status data of the edge node through the computing power resource management module; Evaluate the computing power resource status data of the edge node according to a preset computing power status evaluation algorithm, and when the computing power resources of the edge node are judged to be in an idle state, add the edge node to the shared pool; According to the task demand information released by the computing power demander, the node matching module selects the best matching node from the shared pool according to a preset screening algorithm; The task instruction of the computing power demander is sent from the cloud to the optimal matching node through the distribution execution module, and the optimal matching node performs the task execution; When the optimal matching node completes the task, its computing resources are released through the computing resource management module, and the fee settlement is performed through the billing module.

2. The edge computing power sharing method according to claim 1, characterized in that: The step of building a decentralized P2P network to connect edge nodes through the network connection module includes: Establishing a network connection of the edge node through the network connection module according to a distributed node discovery mechanism; The network connection module uses the STUN protocol to achieve NAT penetration, and cooperates with the ICE protocol to perform connectivity detection; The communication parameters of the edge nodes are dynamically allocated through the network connection module by means of a virtual network controller.

3. The edge computing power sharing method according to claim 1, characterized in that: The step of collecting the computing resource status data of the edge node through the computing resource management module includes: The computing power parameters of the edge node are collected in real time through a monitoring program installed in the computing power resource management module, wherein the computing power parameters include CPU utilization, GPU utilization, memory utilization, and network bandwidth; The computing power parameters are encrypted by using the AES-256 algorithm and reported to the cloud after being compressed by GZIP.

4. The edge computing power sharing method according to claim 3 is characterized in that: The step of evaluating the computing power resource status data of the edge node according to a preset computing power status evaluation algorithm, and when the computing power resources of the edge node are judged to be in an idle state, adding the edge node to the shared pool comprises: The cloud decrypts the encrypted and compressed computing power parameters, calculates the computing power parameters through a dynamic threshold algorithm, and identifies the node load status of the edge node; When the sharable idle computing power resources of the edge node meet the preset conditions for computing power sharing, marking the edge node; All the marked edge nodes are included in the shared pool as candidate nodes for providing computing power.

5. The edge computing power sharing method according to claim 4, characterized in that: The step of selecting the best matching node from the shared pool by the node matching module according to a preset screening algorithm based on the task demand information released by the computing power demander includes: Obtaining task requirement information submitted by the demander through the node matching module, wherein the task requirement information includes a description of computing power requirements, network delay tolerance, and cost budget; Constructing a dynamic matching model according to the task requirement information through the node matching module; Performing feature-weighted scoring on all candidate nodes providing computing power in the shared pool by using the dynamic matching model; The dynamic matching model performs weighted scoring based on computing power performance matching, network latency, and cost budget compliance, and selects the node with the highest comprehensive score as the optimal matching node.

6. The edge computing power sharing method according to claim 5, characterized in that: The step of sending the task instruction of the computing power demander from the cloud to the optimal matching node through the distribution execution module, and the optimal matching node performing the task execution includes: The task instructions of the computing power demander are distributed to the optimal matching node through the distribution execution module via the encrypted instruction channel of the cloud; Establishing a P2P direct connection channel with the computing power demander through the optimal matching node via the network connection module, reporting the task execution progress to the cloud through a heartbeat packet, and sending a status notification immediately when the task is completed or abnormally terminated; The optimal matching node reports the task execution progress to the cloud in real time.

7. The edge computing power sharing method according to claim 6, characterized in that: The edge computing power sharing system further includes a security module, and after the step of distributing the task instructions of the computing power demander to the optimal matching node through the encrypted instruction channel of the cloud by the distribution execution module, the step further includes: The security module uses dynamic key negotiation to generate a temporary session key to encrypt the communication link, and after verifying the authority of the demander according to the RBAC model, the task is deployed to the containerized isolation environment for execution.

8. An edge computing power sharing device, characterized in that: Used to execute the edge computing power sharing method as described in any one of claims 1 to 7.

9. A computer device, characterized in that: The computer device comprises a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to execute the steps of the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 can be implemented.

Images