Computing power allocation method based on cloud-edge-end intelligent collaboration
By monitoring the load of edge computing nodes in real time and adjusting task allocation dynamically, combined with the intelligent matching AI algorithm model, the problem of excessive load of edge nodes in cloud edge collaboration is solved, the system performance and real-time performance are improved, and the efficient utilization of resources is achieved and Load balancing.
Patent Information
- Application Number
- CN202411067730.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-08-06
AI Technical Summary
In the process of intelligent cloud-edge end intelligent collaboration, edge computing nodes have limited resources. Once the load is too high, it may cause task processing delays to increase or fail. How to efficiently allocate computing resources to ensure that tasks are processed in a timely manner and avoid node overloading becomes a challenge.
By building a cloud-edge collaborative system, the load situation of edge computing nodes is monitored in real time, task allocation is dynamically adjusted, and the optimal AI algorithm model is intelligently matched according to task requirements and node load conditions, and tasks are allocated to nodes with lower load for processing.
Effectively avoid node overload, ensure timely processing of tasks, improve system performance and stability, maximize the use of computing resources, reduce task processing delays, enhance the real-time and flexibility of the system, and adapt to application scenarios of different scales and complexities.
Smart Images

Figure CN118585405B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of cloud-edge-end collaboration technology, and specifically to a computing power allocation method based on cloud-edge-end intelligent collaboration. Background Art
[0002] With the rapid development of information technology, the collaborative work of cloud computing, edge computing and terminal devices (i.e., cloud-edge-end) has become a hot topic in current technical research. Cloud computing centers, with their powerful computing power and abundant resource reserves, are able to handle large-scale data and tasks. Edge computing nodes are located near users and can provide low-latency, high-bandwidth services, especially suitable for processing tasks with high real-time requirements. As the data generator and receiver, the collaborative work of terminal devices with edge computing nodes and cloud computing centers can further improve the overall performance of the system.
[0003] However, in practical applications, due to the diversity of task types and computing requirements, how to efficiently and reasonably allocate computing resources to ensure that tasks are processed in a timely manner while avoiding overloading of some nodes and idleness of other nodes has become an urgent problem to be solved. Especially in edge computing nodes, due to their relatively limited resources, once the load is too high, it may lead to increased task processing delays or even task failures.
[0004] In view of this, the art needs a computing power allocation method based on intelligent collaboration between cloud, edge and end to solve the above problems. Summary of the invention
[0005] In order to solve the above technical problems, that is, to solve the problem that in the existing cloud-edge-end intelligent collaboration process, edge computing nodes have relatively limited resources and once the load is too high, it may cause increased task processing delays or even task failures. The present invention provides a computing power allocation method based on cloud-edge-end intelligent collaboration, which aims to improve the overall performance of the system, optimize resource utilization, reduce latency and improve real-time performance, and enhance the scalability and flexibility of the system.
[0006] The present invention provides a computing power allocation method based on cloud-edge-end intelligent collaboration, and the computing power allocation method includes:
[0007] S1: Build a cloud-edge-end collaborative system including a cloud computing center, edge computing nodes, and terminal devices, wherein the cloud computing center includes an AI algorithm model library;
[0008] S2: When the cloud-edge-device collaborative system receives a task, it performs demand analysis and processes the task;
[0009] S3: monitor the load of all edge computing nodes in real time;
[0010] Step S3 specifically includes:
[0011] S31: deploying a monitoring agent on each edge computing node and deploying a monitoring center in the cloud computing center;
[0012] S32: using the monitoring agent to collect target data from the corresponding edge computing node at preset intervals;
[0013] S33: Encapsulating the collected target data into a heartbeat packet and sending the heartbeat packet to the monitoring center, wherein the heartbeat packet has a unique identifier and a timestamp of the edge computing node;
[0014] S34: After receiving the heartbeat packet, the monitoring center records the data and updates the status information of the corresponding edge computing node;
[0015] S35: The monitoring center analyzes the received data and calculates the load of each edge computing node;
[0016] S4: when the load of the edge computing node exceeds a set threshold, allocating part of the tasks on the edge computing node to a target edge computing node;
[0017] S5: Matching the optimal AI algorithm model from the AI algorithm model library according to the assigned part of the tasks and sending it to the target edge computing node;
[0018] S6: Processing the assigned part of the task using the sent optimal AI algorithm model at the target edge computing node.
[0019] In some preferred embodiments, step S2 specifically includes:
[0020] S21: When the cloud-edge-device collaborative system receives a task, performing a demand analysis on the task and determining a priority of the task;
[0021] S221: If it is determined that the priority of the task is high, the cloud computing center is used for processing;
[0022] S222: If it is determined that the priority of the task is medium priority, the edge computing node is used for processing;
[0023] S223: If it is determined that the priority of the task is low, the terminal device is used for processing.
[0024] In some preferred embodiments, the priority is determined by the following formula:
[0025]
[0026] in, , , , , and are weights, is the total priority score, is the normalized computational complexity score, is the score of the normalized data volume, is the normalized latency score, , is the delay sensitivity, is the privacy score of the normalized data. is the normalized network condition score, is the normalized task urgency score, + + + + + =1.
[0027] In certain preferred implementations, the target data includes CPU usage, memory usage, disk I / O usage, network bandwidth usage, task queue length, and task execution delay.
[0028] In some preferred embodiments, in step S4, the target edge computing node is selected as follows:
[0029] ;
[0030] in, For the selected target edge computing node, , , and are weights, A set of candidate target edge computing nodes The nodes, The original edge computing node whose current load exceeds the set threshold, Candidate target edge computing nodes The current load rating of Original edge computing node and candidate target edge computing nodes The network delay between Candidate target edge computing nodes The computing power of Candidate target edge computing nodes Resource utilization, + + + =1.
[0031] In certain preferred embodiments, the load score The calculation formula is:
[0032] ;
[0033] in, Candidate target edge computing nodes The current load, Candidate target edge computing nodes Maximum load capacity;
[0034] Network latency The calculation formula is:
[0035] ;
[0036] in, Original edge computing node and candidate target edge computing nodes The network bandwidth between is the amount of data to be migrated, Original edge computing node and candidate target edge computing nodes The actual bandwidth usage between
[0037] The resource utilization rate is calculated as follows:
[0038] ;
[0039] in, Candidate target edge computing nodes The computing resources currently available, Candidate target edge computing nodes total computing resources.
[0040] In some preferred embodiments, step S5 specifically includes:
[0041] S51: Calculate the matching score between the task and the AI algorithm model using the following formula;
[0042] ;
[0043] in, is the matching score between the task and the AI algorithm model, is the number of features in the task, For the The weight of the feature, For the task Features, For the Among the AI algorithm models Features, is the similarity function between task features and model features;
[0044] S52: Selecting a model with the highest matching score as the optimal AI algorithm model;
[0045] S53: Send the optimal AI algorithm model to the target edge computing node.
[0046] In some preferred embodiments, the similarity function is:
[0047] ;
[0048] Among them, in the above formula, and is expressed as a vector, and They are and The Euclidean norm of .
[0049] In some preferred embodiments, the computing power allocation method further includes:
[0050] S7: Encapsulate the processed results and transmit them to the task initiator for verification.
[0051] It can be seen from the above that the computing power allocation method based on cloud-edge-end intelligent collaboration provided by the present invention has the following beneficial technical effects:
[0052] The present invention can effectively avoid node overload and ensure that tasks are processed in a timely manner by monitoring the load of edge computing nodes in real time and dynamically adjusting task allocation, thereby improving the overall performance and stability of the system. At the same time, according to task requirements and node load conditions, the optimal AI algorithm model is intelligently matched and allocated, which can maximize the use of computing resources and avoid resource waste. In addition, by allocating some tasks from edge computing nodes with higher loads to nodes with lower loads, and processing them using the optimal AI algorithm model, task processing delays can be reduced and the real-time performance of the system can be further improved. In summary, the method can dynamically adjust the computing power allocation according to actual conditions, making the system more flexible and scalable, and able to adapt to application scenarios of different scales and complexities. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:
[0054] Figure 1 This is a flow chart of the computing power allocation method based on cloud-edge-end intelligent collaboration of the present invention. DETAILED DESCRIPTION
[0055] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0056] Based on the background technology, the edge computing nodes in the existing cloud-edge-end intelligent collaboration process have relatively limited resources. Once the load is too high, it may cause increased task processing delays or even task failures. The present invention provides a computing power allocation method based on cloud-edge-end intelligent collaboration, which aims to improve the overall performance of the system, optimize resource utilization, reduce latency and improve real-time performance, and enhance the scalability and flexibility of the system.
[0057] like Figure 1 As shown, the computing power allocation method based on cloud-edge-end intelligent collaboration of the present invention includes:
[0058] S1: Build a cloud-edge-end collaborative system consisting of a cloud computing center, edge computing nodes, and terminal devices, where the cloud computing center includes an AI algorithm model library.
[0059] A large number of AI algorithm models are stored in the AI algorithm model library, which is convenient for subsequent distribution to the designated target edge computing nodes. Of course, all AI algorithm models in the AI algorithm model library can also be called up and used when the cloud computing center processes tasks, providing sufficient algorithm support for the system. The AI algorithm model library can be updated at regular intervals, and new AI algorithm models can be added and infrequently used AI algorithm models can be deleted.
[0060] S2: When the cloud-edge collaborative system receives a task, it performs demand analysis and processes the task.
[0061] In a preferred embodiment, the above step S2 specifically includes:
[0062] S21: When the cloud-edge-device collaborative system receives a task, it performs a demand analysis on the task and determines the priority of the task;
[0063] S221: If it is determined that the priority of the task is high, the cloud computing center is used for processing;
[0064] S222: If it is determined that the priority of the task is medium priority, an edge computing node is used for processing;
[0065] S223: If it is determined that the priority of the task is low, the terminal device is used for processing.
[0066] Further preferably, the priority is determined by the following formula:
[0067] ;
[0068] in, , , , , and are weights, is the total priority score, is the normalized computational complexity score, is the score of the normalized data volume, is the normalized latency score, , is the delay sensitivity, is the privacy score of the normalized data. is the normalized network condition score, is the normalized task urgency score, + + + + + =1.
[0069] In the above example, the computing requirement score range is 1-100. When the computing requirement of a task is 75, , that is, the normalized score is 0.747; for delay sensitivity, the lower the delay tolerance, the higher the real-time requirement of the task, and the higher the delay score should be. At this time, the reciprocal method is directly used for normalization operation, such as delay sensitivity The threshold range is 10-1000 ms, which is the delay sensitivity of a task. =50 ms, then , that is, the latency score is 0.18; for example, data privacy can be a categorical variable. For a task, the data privacy is 3, and the maximum data privacy is 5. , that is, the normalized score of the data privacy score is 0.5; and Can be used with The same normalized scoring method will not be repeated here.
[0070] It should be noted that when the cloud-edge collaborative system receives multiple tasks, in addition to determining whether the subsequent processing will be carried out by the cloud computing center, edge computing node or terminal device based on the priority of each task, the priority of each task can also be used as the basis for queue sorting. For example, when all tasks cannot be processed concurrently, tasks with high priority scores can be processed first.
[0071] S3: Monitor the load of all edge computing nodes in real time.
[0072] Preferably, the above step S3 specifically includes:
[0073] S31: deploy a monitoring agent on each edge computing node and deploy a monitoring center in the cloud computing center, wherein the monitoring agent is responsible for collecting performance data of the corresponding edge computing node, and the monitoring center is used to receive and analyze monitoring data from each edge computing node;
[0074] S32: Use a monitoring agent to collect target data from the corresponding edge computing node at preset intervals, wherein the target data includes CPU usage, memory usage, disk I / O usage, network bandwidth usage, task queue length, and task execution delay. In the above, the preset time can be 5 seconds or other time. The collected target data can be obtained through system commands or API interfaces;
[0075] S33: Encapsulate the collected target data into a heartbeat packet and send the heartbeat packet to the monitoring center, wherein the heartbeat packet has a unique identifier of the edge computing node, a timestamp, and the collected target data. The monitoring center maintains a status list of the edge computing nodes for storing the latest performance data of each edge computing node;
[0076] S34: After receiving the heartbeat packet, the monitoring center records the data and updates the status information of the corresponding edge computing node;
[0077] S35: The monitoring center analyzes the received data and calculates the load of each edge computing node.
[0078] The above step S3 may further include:
[0079] S36: If the monitoring center does not receive a heartbeat packet from an edge computing node within a predetermined time, it is considered that the node may fail or the network may be interrupted. The monitoring center can take corresponding measures, such as sending an alarm, trying to reconnect, or starting a backup node.
[0080] S4: When the load of the edge computing node exceeds a set threshold, some tasks on the edge computing node are allocated to a target edge computing node, where the target edge computing node can be a standby node or other node with low load.
[0081] In a preferred scenario, in the above step S4, the target edge computing node is selected as follows:
[0082] ;
[0083] in, For the selected target edge computing node, , , and are weights, A set of candidate target edge computing nodes The nodes, The original edge computing node whose current load exceeds the set threshold. Candidate target edge computing nodes The current load rating of Original edge computing node and candidate target edge computing nodes The network delay between Candidate target edge computing nodes The computing power of Candidate target edge computing nodes Resource utilization, + + + =1.
[0084] In the above, the candidate target edge computing nodes can be calculated first Current load rating , and then calculate the original edge computing node With each candidate target edge computing node Network delay between , and then evaluate the computing power of each candidate target edge computing node , and then calculate the resource utilization of each candidate target edge computing node , then calculate each candidate target edge computing node according to the above formula The comprehensive score of the edge computing node is finally selected as the target edge computing node. .
[0085] Of course, in addition to using the above greedy algorithm to determine the target edge computing node, a heuristic algorithm can also be used to determine the target edge computing node.
[0086] In the above, the load rating The calculation formula is:
[0087] ;
[0088] in, Candidate target edge computing nodes The current load, Candidate target edge computing nodes Maximum load capacity;
[0089] Network latency The calculation formula is:
[0090] ;
[0091] in, Original edge computing node and candidate target edge computing nodes The network bandwidth between is the amount of data to be migrated, Original edge computing node and candidate target edge computing nodes The actual bandwidth usage between
[0092] The resource utilization rate is calculated as follows:
[0093] ;
[0094] in, Candidate target edge computing nodes The computing resources currently available, Candidate target edge computing nodes total computing resources.
[0095] S5: Match the optimal AI algorithm model from the AI algorithm model library according to the assigned partial tasks and send it to the target edge computing node.
[0096] Preferably, the above step S5 specifically includes:
[0097] S51: Calculate the matching score between the task and the AI algorithm model using the following formula;
[0098] ;
[0099] in, is the matching score between the task and the AI algorithm model, is the number of features in the task, and the number of features in the corresponding AI algorithm model is also , For the The weight of the feature, For the task Features, For the Among the AI algorithm models Features, is the similarity function between the task features and the model features, where the similarity function is:
[0100] ;
[0101] Among them, in the above formula, and is expressed as a vector, and They are and The Euclidean norm of ;
[0102] S52: Selecting a model with the highest matching score as the optimal AI algorithm model;
[0103] S53: Send the optimal AI algorithm model to the target edge computing node.
[0104] S6: Use the optimal AI algorithm model sent down to process part of the assigned tasks at the target edge computing node.
[0105] S7: Encapsulate the processed results and transmit them to the task initiator for verification.
[0106] Specifically, the above step S7 includes:
[0107] S71: The target edge computing node needs to encapsulate the processing results to ensure the integrity and consistency of the data. The encapsulation process may include adding timestamps, node identifiers, task identifiers and other information for subsequent tracking and verification.
[0108] S72: The encapsulated processing result is transmitted back to the task initiator through the network. During the transmission process, encryption, compression and other measures can be taken to protect data security and reduce transmission time.
[0109] S73: After receiving the processing result, the task initiator verifies the result to ensure the integrity and accuracy of the data. The verification process may include checking information such as timestamp, node ID, task ID, and performing data integrity verification.
[0110] S74: Process the results and provide feedback. Once the processing results pass the verification, the task initiator further processes or analyzes the results. At the same time, feedback needs to be provided to the task initiator to inform him of the results and status of the task processing so that he can perform subsequent operations or decisions.
[0111] S75: Record and monitor logs: During the entire process of processing result feedback, detailed log information should be recorded, including the steps of sending, receiving, verifying and processing the processing results. Log records can facilitate subsequent problem tracking and system optimization.
[0112] S76: Process the exception and execute the retry mechanism. If an exception occurs during the feedback of the processing result, such as network failure, data loss, etc., adopt the corresponding exception handling mechanism, such as trying to resend the processing result or requesting to reprocess the task, to ensure the smooth completion of the task and accurate feedback of the result.
[0113] The computing power allocation method based on cloud-edge-end intelligent collaboration of the present invention can be applied to a variety of scenarios, such as smart cities, industrial Internet of Things, autonomous driving and intelligent transportation systems, telemedicine and health management systems, multimedia processing and entertainment systems, disaster warning and emergency response systems, and converter stations.
[0114] Taking the converter station as an example, when the computing power allocation method based on cloud-edge-end intelligent collaboration of the present invention is applied to the converter station scenario, the cloud computing center is used to store and manage the AI algorithm model library of the converter station. These models may include algorithm models such as power load prediction, equipment fault diagnosis, and power quality optimization. The cloud computing center is also responsible for global data analysis and decision support. Edge computing nodes are deployed in the converter station. These nodes can be integrated in devices such as converter valve control units and converter transformer monitoring systems, or deployed as independent computing units. The edge computing nodes are responsible for processing local real-time data in the converter station and making quick decisions. The terminal devices include various sensors, monitoring devices, and actuators in the converter station, such as current and voltage sensors, temperature sensors, and circuit breakers. These devices collect raw data and upload it to the edge computing nodes or the cloud computing center. When the converter station receives a new task (such as adjusting the power transmission power, optimizing the power quality, etc.), the system first performs a task requirement analysis to determine the specific requirements and required resources of the task. According to the task requirements, the system needs to call the AI algorithm model library of the cloud computing center for preliminary processing, or directly perform the task. Assigned to the edge computing node for real-time processing. The system monitors the load of all edge computing nodes in the converter station in real time through the built-in monitoring mechanism, including CPU usage, memory occupancy, network bandwidth and other indicators. The monitoring data is regularly summarized and reported to the cloud computing center for global resource scheduling and optimization. When the load of an edge computing node exceeds the set threshold, the system will dynamically assign part of the tasks to other idle or low-load target edge computing nodes based on factors such as the current network status, the load of other edge computing nodes and task priority. The system matches the optimal algorithm model from the AI algorithm model library of the cloud computing center according to the assigned task type and requirements. After the matching is completed, the system sends the algorithm model to the target edge computing node so that it can perform the corresponding processing task. After receiving the algorithm model, the target edge computing node immediately uses the model to process the assigned task. The processing results may include adjustment instructions for power transmission power, optimization parameters for power quality, etc. These results will be fed back to the control system of the converter station in real time or uploaded to the cloud computing center for further analysis and decision-making. Applying the computing power allocation method based on cloud-edge-end intelligent collaboration of the present invention to the converter station can significantly improve the real-time and low-latency processing capabilities of the system, achieve resource optimization and load balancing through dynamic task allocation, and use the powerful AI algorithm model library of the cloud computing center to provide accurate data processing and decision support. This method can also enhance the reliability and security of the system, reduce data transmission risks, and have a high degree of flexibility and scalability to adapt to the expansion of the converter station scale or the increase in business demand. In addition, the automated monitoring and management mechanism helps to reduce operation and maintenance costs and improve overall operating efficiency.
[0115] It should be noted that the method of the present invention can be performed by a single device, such as a computer or a server. The method of the present invention can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the present invention, and the multiple devices will interact with each other to complete the described method.
[0116] It should be noted that the above describes specific embodiments of the present invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0117] The present invention has the following beneficial effects: by real-time monitoring of the load of edge computing nodes and dynamically adjusting task allocation, node overload can be effectively avoided, ensuring that tasks are processed in a timely manner, thereby improving the overall performance and stability of the system. At the same time, according to task requirements and node load conditions, the optimal AI algorithm model is intelligently matched and allocated, which can maximize the use of computing resources and avoid resource waste. In addition, by allocating some tasks from edge computing nodes with higher loads to nodes with lower loads, and processing them using the optimal AI algorithm model, task processing delays can be reduced and the real-time performance of the system can be further improved. In summary, the method can dynamically adjust the computing power allocation according to actual conditions, making the system more flexible and scalable, and able to adapt to application scenarios of different scales and complexities.
[0118] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, devices, etc. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.
[0119] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0120] Each embodiment of the present invention is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0121] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure is limited to these examples. Based on the concept of the present disclosure, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present invention as described above, which are not provided in detail for the sake of simplicity.
[0122] While the disclosure has been described in conjunction with specific embodiments thereof, many alternatives, modifications and variations of these embodiments will be apparent to those skilled in the art in light of the foregoing description.
[0123] One or more embodiments of the present invention are intended to cover all such replacements, modifications and variations that fall within the scope of protection of the present invention. Therefore, any omissions, modifications, equivalent replacements, improvements, etc. made within the spirit and principles of one or more embodiments of the present invention should be included in the scope of protection of this disclosure.
Claims
1. A computing power allocation method based on cloud-edge-end intelligent collaboration, It is characterized in that The computing power allocation method comprises: S1: Build a cloud-edge-end collaborative system including a cloud computing center, edge computing nodes, and terminal devices, wherein the cloud computing center includes an AI algorithm model library; S2: When the cloud-edge-device collaborative system receives a task, it performs demand analysis and processes the task; Step S2 specifically includes: S21: When the cloud-edge-device collaborative system receives a task, performing a demand analysis on the task and determining a priority of the task; S221: If it is determined that the priority of the task is high, the cloud computing center is used for processing; S222: If it is determined that the priority of the task is medium priority, the edge computing node is used for processing; S223: If it is determined that the priority of the task is low, the terminal device is used for processing; S3: monitor the load of all edge computing nodes in real time; Step S3 specifically includes: S31: deploying a monitoring agent on each edge computing node and deploying a monitoring center in the cloud computing center; S32: using the monitoring agent to collect target data from the corresponding edge computing node at preset intervals; S33: Encapsulating the collected target data into a heartbeat packet and sending the heartbeat packet to the monitoring center, wherein the heartbeat packet has a unique identifier and a timestamp of the edge computing node; S34: After receiving the heartbeat packet, the monitoring center records the data and updates the status information of the corresponding edge computing node; S35: The monitoring center analyzes the received data and calculates the load of each edge computing node; S4: when the load of the edge computing node exceeds a set threshold, allocating part of the tasks on the edge computing node to a target edge computing node; In step S4, the target edge computing node is selected as follows: in, For the selected target edge computing node, , , and are weights, A set of candidate target edge computing nodes The nodes, The original edge computing node whose current load exceeds the set threshold, Candidate target edge computing nodes The current load rating of Original edge computing node and candidate target edge computing nodes The network delay between Candidate target edge computing nodes The computing power of Candidate target edge computing nodes Resource utilization, + + + =1; Load rating The calculation formula is: in, Candidate target edge computing nodes The current load, Candidate target edge computing nodes Maximum load capacity; Network latency The calculation formula is: in, Original edge computing node and candidate target edge computing nodes The network bandwidth between is the amount of data to be migrated, Original edge computing node and candidate target edge computing nodes The actual bandwidth usage between The resource utilization rate is calculated as follows: in, Candidate target edge computing nodes The computing resources currently available, Candidate target edge computing nodes Total computing resources; S5: Matching the optimal AI algorithm model from the AI algorithm model library according to the assigned part of the tasks and sending it to the target edge computing node; S6: Processing the assigned part of the task using the sent optimal AI algorithm model at the target edge computing node.
2. According to the computing power allocation method based on cloud-edge-end intelligent collaboration according to claim 1, It is characterized in that The priority is determined by the following formula: in, , , , , and are weights, is the total priority score, is the normalized computational complexity score, is the score of the normalized data volume, is the normalized latency score, , is the delay sensitivity, is the privacy score of the normalized data. is the normalized network condition score, is the normalized task urgency score, + + + + + =1.
3. According to the computing power allocation method based on cloud-edge-end intelligent collaboration according to claim 1, It is characterized in that The target data includes CPU usage, memory usage, disk I / O usage, network bandwidth usage, task queue length and task execution delay.
4. According to claim 1, the computing power allocation method based on cloud-edge-end intelligent collaboration, It is characterized in that Step S5 specifically includes: S51: Calculate the matching score between the task and the AI algorithm model using the following formula; in, is the matching score between the task and the AI algorithm model, is the number of features in the task, For the The weight of the feature, For the task Features, For the Among the AI algorithm models Features, is the similarity function between task features and model features; S52: Selecting a model with the highest matching score as the optimal AI algorithm model; S53: Send the optimal AI algorithm model to the target edge computing node.
5. According to the computing power allocation method based on cloud-edge-end intelligent collaboration according to claim 4, It is characterized in that The similarity function is: Among them, in the above formula, and is expressed as a vector, and They are and The Euclidean norm of .
6. The computing power allocation method based on cloud-edge-end intelligent collaboration according to any one of claims 1 to 5, It is characterized in that The computing power allocation method further includes: S7: Encapsulate the processed results and transmit them to the task initiator for verification.
Citation Information
Patent Citations
Cloud edge-end collaborative system optimization method and device based on improved binary multivariate universe algorithm
CN116887344A
Cloud side-end collaboration method and system for comprehensive management of intelligent monitoring system
CN117640634A
Intelligent scheduling algorithm based on cloud edge collaboration
CN117762597A
Cloud side-end collaborative computing power resource intelligent monitoring and scheduling method and cloud side-end collaborative computing power resource intelligent monitoring and scheduling system
CN117827441A
Task load balancing scheduling method based on cloud side-end collaboration
CN118394520A
