Edge node scheduling method and system based on sinking distributed network

Abstract

The invention provides an edge node scheduling method and system based on a sinking distributed network. The method comprises the following steps: determining a scheduling strategy of an edge node based on a preset load balancing algorithm; receiving a service request from the terminal equipment; identifying a proper service node based on the data type of the service request; distributing the service node to process the service request; and adjusting the scheduling strategy based on the real-time monitored data feedback. Through the scheme of the embodiment of the invention, the problem of data consistency caused by insufficient data synchronization mechanism under the condition of high concurrent requests can be solved. And the problem of communication delay increase caused by lack of an adaptive routing protocol under the condition of dynamically changing network topology is solved. And the problem of computing resource waste caused by an unreasonable task allocation strategy under the condition of resource-limited edge nodes is solved. And under the condition of data intensive application, the problem of data processing bottleneck caused by an imperfect data caching mechanism is solved.

Description

Technical Field

[0001] The present application relates to the field of computer network technology, and in particular to an edge node scheduling method and system based on a distributed network. Background Art

[0002] With the development of the Internet of Things (IoT) and 5G communication technology, edge computing is rapidly emerging as a new computing paradigm. It transfers data processing tasks from traditional centralized cloud platforms to the edge of the network, that is, to devices close to the data source or user. The edge node scheduling method based on the sinking distributed network aims to take advantage of this architecture to efficiently allocate and schedule computing tasks at the edge of the network to improve the system's response speed and service quality. By reducing the need to transmit data to remote data centers, edge computing can significantly reduce latency while alleviating pressure on the core network.

[0003] Although edge computing offers many potential benefits, there are still a number of challenges to achieving effective edge node scheduling:

[0004] 1. Data consistency: In the case of high concurrent requests, how to ensure data synchronization between edge nodes becomes a key issue. If this problem cannot be effectively solved, data inconsistency may occur, which in turn affects the stability of the entire system and user experience.

[0005] 2. Adaptive routing protocol: Dynamically changing network environments require efficient adaptive routing protocols to optimize data paths to minimize communication delays. However, existing solutions often lack sufficient flexibility to cope with complex changes in network conditions, thus limiting system performance.

[0006] 3. Resource utilization: Edge nodes are usually deployed in resource-constrained environments, such as small servers, routers, etc. Therefore, it is crucial to properly dispatch tasks to appropriate edge nodes to avoid resource waste. Improper task allocation will not only cause some nodes to be overloaded, but may also fail to meet user needs due to insufficient resources.

[0007] 4. Cache management: When facing application scenarios that need to process large amounts of data, a good cache strategy can help alleviate storage pressure and speed up data access. However, if the cache mechanism is not designed properly, it is easy to encounter bottlenecks, especially when frequent data updates conflict with query operations.

[0008] 5. Security assurance: If adequate security measures are not taken during the interaction between different fields, sensitive information is vulnerable to threats from attackers. This not only damages personal privacy, but may also cause economic losses to the company.

[0009] In summary, the above problems are the main obstacles to the current implementation of effective distributed network edge scheduling. In order to overcome these difficulties, the present invention proposes a novel method, which aims to provide a more intelligent and secure edge node scheduling solution by comprehensively considering multiple factors such as load balancing algorithm, real-time monitoring feedback adjustment and multi-dimensional evaluation. Summary of the invention

[0010] In view of this, an embodiment of the present disclosure provides an edge node scheduling method based on a sinking distributed network, which at least partially solves the problems existing in the prior art.

[0011] The edge node scheduling method based on the sinking distributed network includes:

[0012] Determine the scheduling strategy of edge nodes based on the preset load balancing algorithm;

[0013] Receiving service requests from terminal devices;

[0014] identifying an appropriate service node based on the data type of the service request;

[0015] Allocating the service node to process the service request;

[0016] Adjust scheduling strategies based on real-time monitoring data feedback.

[0017] Preferably, determining the scheduling strategy of the edge node based on a preset load balancing algorithm includes:

[0018] Collect the historical load of each edge node;

[0019] Determine an importance factor α for measuring historical load;

[0020] Assign historical load weights, where weight W(i) = α*H(i), H(i) is the historical load value of the i-th edge node;

[0021] Determine the edge node scheduling strategy based on the current system status and the historical load of the edge nodes;

[0022] The technical solution claimed in this claim can more accurately identify historical load conditions and adjust scheduling decisions based on historical conditions to optimize load balancing.

[0023] Preferably, collecting the historical load of each edge node includes:

[0024] Update the log records of edge nodes after each task is completed;

[0025] Separate the CPU utilization percentage CPUp(i,t) and memory usage ratio Memoryp(i,t) from the historical data, where i represents the i-th edge node and t represents the time point;

[0026] Use statistical functions to calculate the CPU utilization change ΔCPUp(i,T) and memory usage change ΔMemoryp(i,T) of each edge node in the past t time period. Specifically:

[0027] The CPU utilization change is defined as: ΔCPUp(i,T)=CPUp(i,t+T)-ΔCPUp(i,t);

[0028] The memory usage change is defined as: ΔMemoryp(i,T) = ΔMemoryp(i,t+T) - ΔMemoryp(i,t)

[0029] If ΔCPUp(i,T)+ΔMemoryp(i,T)>θ, it is considered that the resource usage fluctuation in the time period T is significant, and it is necessary to consider adjusting the load balancing strategy, where θ is a preset threshold used to determine whether the fluctuation range of resource usage reaches the level that affects the load balancing strategy.

[0030] Preferably, determining an importance factor α for measuring historical load comprises the following steps:

[0031] The dynamic parameter β is set to represent the intensity of the immediate response demand of the network load;

[0032] The static parameter κ is set to reflect the system stability preference weight;

[0033] Then α is determined by finding the average function α = (β + κ) / 2;

[0034] If β / κ>γ (γ represents the target ratio defined by the user), the algorithm parameters are adjusted to favor fast response rather than robustness;

[0035] Under this condition, β represents the system's emphasis on delay sensitivity, while κ focuses on stability and service quality.

[0036] Preferably, the weight distribution when adjusting the scheduling according to the historical load includes the following details:

[0037] Understand the impact of historical load H(i) on the current demand D;

[0038] Considering the impact of network transmission efficiency, f2 = 1 / NetDelayi, where NetDelayi is the actual delay from the central server to the edge node i;

[0039] Combine the above two items f=f1+f2 to determine the total factor, and sort the processing priority of each edge node according to this comprehensive score;

[0040] If any two nodes i and j satisfy f(i)f(j)>σ, then i should be considered as a high-priority option for scheduling (σ is a small difference threshold determined by the user);

[0041] The value of σ controls the algorithm's ability to distinguish nodes. When σ is small, it will promote more selection results based on fine-grained evaluation.

[0042] Preferably, the steps of calculating the total weight by combining multiple influences include:

[0043] Set a corresponding standardized value ni for each impact item (such as standardized indicator values ​​such as CPU and bandwidth);

[0044] Select a weighted method to calculate the final weighted score Wi;

[0045] For example, Wi=Σni*weighti;

[0046] A filtering criterion F:Wi>λ (where λ represents the minimum effective allocation threshold) is applied to determine that the node is suitable for allocation service;

[0047] The parameter λ is used to exclude low-performance nodes to ensure service quality and performance.

[0048] Preferably, the available service nodes are screened based on multi-dimensional evaluation conditions:

[0049] Check whether the comprehensive availability A of the edge endpoint E is at least equal to the minimum acceptance value L defined by the system, indicating that the node E is in a healthy and operational state (A≥L);

[0050] Additionally check that E's effective service processing capacity SC is higher than the minimum processing level M to ensure actual business operation efficiency (SC>M);

[0051] The above formula ensures that only nodes that meet the minimum specifications and service requirements can be finally selected to assign the service request task.

[0052] Preferably, when the edge node E enters the range to be scheduled, its load status update mechanism is:

[0053] Autonomously sense the local system resource changes CR and regularly report to the central manager CMS;

[0054] If the increase in the latest CR value compared to the information reported to CMS at the previous time point T1 reaches the warning limit D, the node allocation will be re-evaluated immediately;

[0055] The updated node load level Nlevel is connected to different service levels through linear mapping.

[0056] Preferably, the edge node scheduling system based on the sinking distributed network is characterized by: comprising a load balancing strategy module, a service request processor, a real-time monitoring and feedback adjustment module, a historical data analysis module, an edge node health check module and an adaptive load update mechanism;

[0057] The load balancing strategy module includes a load data collector, an importance factor calculator and a weight allocator;

[0058] The service request processor includes a request receiver, a data type identifier, and a task distributor;

[0059] The real-time monitoring and feedback regulation module includes a real-time monitor and a feedback regulator;

[0060] The historical data analysis module includes a log updater and a fluctuation detector;

[0061] The edge node health check module includes an availability verifier and a service capability evaluator;

[0062] The adaptive load update mechanism includes local resource change sensor, status reporter, warning trigger and service level mapper.

[0063] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein the processor implements the steps of the above method when executing the computer program.

[0064] The disclosed embodiment provides an edge node scheduling method based on a distributed network, including: determining the scheduling strategy of the edge node based on a preset load balancing algorithm; receiving a service request from a terminal device; identifying a suitable service node based on the data type of the service request; allocating the service node to process the service request; and adjusting the scheduling strategy based on real-time monitoring data feedback. The solution of the disclosed embodiment can solve the data consistency problem caused by insufficient data synchronization mechanism in the case of high concurrent requests.

[0065] The problem of increased communication delay due to the lack of adaptive routing protocols in dynamically changing network topologies.

[0066] In the case of resource-constrained edge nodes, the problem of computing resource waste caused by unreasonable task allocation strategy occurs.

[0067] In data-intensive applications, data processing bottlenecks occur due to imperfect data caching mechanisms.

[0068] The risk of data leakage due to insufficient security measures in cross-domain data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions of the exemplary implementation methods of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the embodiments of the present disclosure and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0070] Figure 1 It is a flow chart of the edge node scheduling method and system based on the sinking distributed network. DETAILED DESCRIPTION

[0071] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0072] Next, with reference to the accompanying drawings, the edge node scheduling method and system based on the sinking distributed network of the present invention are described, and its steps include but are not limited to the following parts. First, the method determines which edge nodes are suitable for performing specific services by using a preset load balancing algorithm. Secondly, the system needs to accept and respond to immediate service needs from end users. Then, when it receives a specific type of data request from the user, it can automatically identify which edge node is most suitable to provide the service and make effective scheduling decisions so that these resources can be used reasonably and efficiently. After the service assignment is completed, these existing allocation rules will also be dynamically modified according to the latest network and node conditions, which will help optimize the subsequent service delivery process and efficiency.

[0073] Based on each step of this core concept, we can further explore its specific execution logic and how to solve some major challenges faced by traditional systems through this method. The first step is to develop a set of behavior specifications suitable for node scheduling in edge computing environments by adopting the principle of load balancing; such as consistent hashing or more intelligent artificial neural network methods. The former can help evenly distribute user requests and workloads on multi-server clusters, while the latter uses advanced machine learning technology and models to select a service node that best suits the current network status for each type of data, avoiding traffic overload at a certain location while improving the overall response efficiency of the system.

[0074] For example, in one embodiment, consider a video surveillance system of a smart city, where video surveillance cameras are our front-end devices and the amount of real-time streaming data they generate is extremely large. Therefore, after deploying such a network, during peak hours, hundreds or thousands of terminals will apply to the background for image recognition analysis at the same time. At this time, these tasks are dispersed and assigned to processing servers at different geographical coordinates through a pre-set weight function, which not only helps prevent the pressure of a data center from reaching a critical state, but also shortens the time from receiving a command to generating a result, thereby greatly improving the user experience; in addition, after the service node is determined, it should also be ensured that the information is transmitted from the source without error and delay. This introduces the second principle - it is necessary to establish a reliable information sending and feedback channel so that the consistency of message transmission can always be ensured regardless of the changing network environment or the uncertainty factors caused by the addition of new devices; for example, using a strong consistency election algorithm such as Paxos or Raft in the blockchain, when multiple devices in the network try to share a resource but only one party can succeed, the consensus protocol is enabled to ensure that only a unique instruction is executed in the end to reduce the probability of repeated resource configuration and erroneous operation, thereby achieving a common data standard and completing the work undertaken by each party in a short time even in poor network conditions or node downtime.

[0075] Specifically, continuing with the above surveillance camera example, if there is a temporary power outage in the city one day, servers in certain areas will suddenly go offline. If we continue to use the traditional TCP / IP model to distribute tasks, it is easy to lose the assigned commands and cause global chaos. At this time, we can restart the heartbeat detector in the surrounding unaffected data centers to determine the scope of the fault and then re-send the remaining data sets to be executed to the remaining active computers. The third step is to adjust the previous deployment according to the actual observed system operation status; the idea behind this is that no static solution can remain effective in all occasions, so we need to design an intelligent engine that can automatically learn and update rules to adjust resource allocation at any time to maintain the healthy operation of the entire system and extend the hardware life cycle. For example, we can deploy performance monitoring components for each region and regularly collect data on CPU utilization, memory usage, remaining hard disk capacity, and network congestion, and then hand it over to a trained machine learning framework for online analysis and trend speculation, and output a policy table containing the optimal resource allocation ratio for each day in the next week. Then, these instructions are passed down to each node, and they will allocate resources, open or close unnecessary programs or functions, expand virtual memory space, adjust compression or encryption algorithms, etc. according to the suggestions, thereby reducing the average delay per unit time and alleviating the additional burden caused by sudden peaks. The last point is the interpretation of the fourth rule, that is, when data needs to be transmitted across regions with different permissions, how to build a reliable umbrella to ensure information security. On the one hand, a cache layer is built locally on the edge node to store the most recently used data blocks and to make redundant copies of insensitive data to prevent single-point damage from terminating the business process. At the same time, asymmetric keys are used to sign or encrypt the content on all transmission links to ensure that the content itself cannot be decrypted by eavesdroppers and that a verification mechanism can be used after reaching the target address to ensure that the received data is indeed sent by a trusted authorized person without any tampering or impersonation by other identities. This not only solves the bottleneck of inefficiency caused by low bandwidth, but also avoids the risk of leaking customer privacy information.

[0076] In summary, the node intelligent matching algorithm based on sinking distribution can show excellent flexibility and stability in dealing with sudden massive concurrency, high-frequency changes in path structure and limited infrastructure constraints, while also meeting the current data-intensive interactive application scenarios that put forward new requirements for security performance and speed. The above description is only a specific example for people in this field to understand the technical solution, and is not a way of strictly limiting the scope of the claim. In practical deployment, some of the processes and components can be freely combined or replaced according to different industry application scenarios to achieve more ideal optimization results.

[0077] Next, the present invention describes the scheduling strategy for edge nodes based on a preset load balancing algorithm. This process involves multiple steps: First, the historical load conditions of each edge node must be collected. The collection process here is to obtain the consumption status data of the edge node processing capacity in a certain period of time in the past. For example, the system may collect information such as the task processing time and the number of tasks completed for each edge node in the past hour. This helps to understand the working intensity of different edge nodes in different time periods.

[0078] Next, it is necessary to determine an importance factor α to measure the importance of the historical load collected at different time nodes. The factor α can be set between 0 and 1, and its value determines the contribution of the historical load to the overall scheduling decision. Specifically, in one embodiment, if α is set close to 1, it means that historical load conditions are highly valued, and if the α value is lower, such as 0.3, more attention is paid to real-time or recent data changes. The best choice is adjusted according to the current stability and degree of change of the network. The optimal value of α needs to be determined through testing to balance the impact of real-time changes and past records.

[0079] On this basis, a corresponding weight W(i) is assigned to each historical load condition. Here, W(i) = α*H(i), which represents the historical load impact of the i-th edge node. H(i) in the formula corresponds to the actual load data of node i during the historical period mentioned above. This arrangement is because we want the calculated scheduling strategy to not only reflect the instantaneous working status of the nodes, but also take into account their performance over the past period of time, so as to make a more stable deployment.

[0080] Finally, based on the above analysis and the demand distribution of the current system, an edge node scheduling strategy that can maximize load balance is determined. For example, in a specific scenario, if it is detected that a node that is usually heavily loaded has suddenly reduced its load, and another node that is usually not heavily loaded but currently has a heavy workload, a portion of the new tasks will be directed to the former based on the new α-weighted strategy, thereby alleviating the overall pressure on the system and ensuring the efficient operation of the system.

[0081] The above steps make reasonable adjustments based on a detailed analysis of the previous workload of each node and the current actual situation, so that the load between edge nodes in the entire sinking distributed network is more balanced, thereby improving the service quality and operation efficiency of the network.

[0082] Next, the present invention describes the collection of historical load conditions of each edge node. In this process, first of all, it is necessary to ensure that the log records of the edge nodes are updated after each task is completed. This step involves making detailed records of the operation logs of each edge node, and the content usually covers various aspects of this operation, such as the start time, end time, proportion of system resources used, and other information. These data serve as the basis for evaluating the historical load of each edge node. Then, after accumulating enough log data, it is necessary to focus on extracting information about node resource utilization from these data, especially separating the two key indicators of CPU utilization percentage CPUp(i, t) and memory occupancy ratio Memoryp(i, t) in the historical data.

[0083] Specifically, a specific time point i is used to represent different edge computing nodes, and t is used to mark a certain time point. The CPU utilization percentage refers to the processor usage of node i within time t, which is displayed as a percentage; correspondingly, the memory usage ratio reflects the memory capacity consumed by the same node in the same time period, which is also expressed as a proportion of the maximum available memory. These two indicators are very important for understanding the working pressure of each edge computing node.

[0084] After the above steps are completed, some statistical methods are needed to process the obtained data, with the purpose of evaluating the actual operation performance and fluctuation degree of each edge computing node within a longer time range t. The Δi,t set here represents the change in the CPU usage of node i in the past t time, and Memorydeltai,t is the change in memory usage in the same time period. Both are calculated by subtracting the actual usage value at time node t from the usage value at interval t after a certain time node t. If the result of Δi,t+Memorydeltai,t exceeds a pre-defined positive threshold θ, it is determined that the resource usage fluctuates significantly in time period t. Among them, the parameter t can be adjusted according to the actual business. For example, in the field of network communication, a sufficiently large value can be selected, such as 60 seconds, to cover the complete periodic peaks and troughs to better capture daily patterns, but it should not be set too long to prevent being too rough to accurately capture the immediate resource usage status fluctuations. On the other hand, the setting of parameter θ is directly related to whether to trigger subsequent further actions, such as reallocating tasks or performing other management behaviors. Generally, the value of θ may be a small fraction of the total resource capacity, for example, a suitable number between 5% and 10% of the total CPU or memory utilization increase is selected as the threshold, thereby ensuring that sudden high load problems that may cause service quality degradation are not ignored while also ensuring that overreaction is not made to avoid unnecessary computing resource reallocation costs.

[0085] For example, in one embodiment, if a distributed IoT system deploys a large number of camera monitoring devices and hopes to maintain a good recording and real-time viewing experience. Assuming that t is set to a 5-minute period to check the resource utilization, and θ is set to 7% of the system resource capacity, then every five minutes, it will automatically compare whether the resource consumption during this period has increased or decreased by more than 7%. In this way, not only can the load status of edge devices be monitored efficiently, but also various dynamic demands can be responded to in a timely manner to ensure network stability and service quality.

[0086] The above is a description of the process of collecting and analyzing edge computing node load information. By understanding the specific steps, we can understand its implementation method and provide a reference for similar application scenarios.

[0087] Next, the present invention is described to determine an importance factor α for measuring historical load, including the following steps:

[0088] The dynamic parameter β is set to represent the intensity of the immediate response demand of the network load;

[0089] The static parameter κ is set to reflect the system stability preference weight;

[0090] Then α is determined by finding the average function α = (β + κ) / 2;

[0091] If β / κ>γ (γ represents the target ratio defined by the user), the algorithm parameters are adjusted to favor fast response rather than robustness;

[0092] Under this condition, β represents the system's emphasis on delay sensitivity, while κ focuses on stability and service quality.

[0093] First, set the dynamic parameter β to represent the intensity of the network load's immediate response requirements; β reflects the load conditions in the current network environment and the speed of responding to instantaneous changes. Usually, β can vary between 0 and 1. The larger the value, the higher the requirement for the current system to respond quickly. The optimal value depends on the specific load conditions and business needs.

[0094] or

[0095] β can be obtained as follows:

[0096] Extract the request response time corresponding to each request of the network load and the number of tasks processed per unit time;

[0097] Obtaining a first network load intensity evaluation coefficient according to the request response time corresponding to each request;

[0098] The first network load strength evaluation coefficient is obtained by the following formula:

[0099]

[0100] Among them, F 01 represents the first network load strength evaluation coefficient; n represents the total number of requests sent during network operation; T i represents the request response time corresponding to the i-th request; T c Indicates the preset request response reference time; T b represents the standard deviation of the request response time corresponding to n requests; T max Indicates the maximum request response time corresponding to n requests; M max Indicates the maximum number of tasks processed per unit time corresponding to m unit time; M c It indicates the reference value of the number of tasks processed corresponding to a preset single unit time; sqrt indicates the square root function symbol;

[0101] Obtaining a second network load intensity evaluation coefficient according to the number of tasks processed per unit time combined with the first network load intensity evaluation coefficient;

[0102] The second network load strength evaluation coefficient is obtained by the following formula:

[0103]

[0104] Among them, F 02 represents the second network load strength evaluation coefficient; m represents the number of unit times experienced by the network operation; T max Indicates the maximum request response time corresponding to n requests; M i Indicates the number of tasks processed per unit time; M max Indicates the maximum number of tasks processed per unit time corresponding to m unit time; M c represents the reference value of the number of tasks processed corresponding to a preset single unit time; sqrt represents the square root function symbol; M b represents the standard deviation of the number of tasks processed per m unit time; F 01 represents the first network load intensity evaluation coefficient;

[0105] The dynamic parameter β is obtained by using the first network load strength evaluation coefficient and the second network load strength evaluation coefficient, and the dynamic parameter β is obtained by the following formula:

[0106]

[0107] Among them, β represents the dynamic parameter; F 01 Indicates the first network load strength evaluation coefficient; F 02Indicates the second network load intensity evaluation coefficient.

[0108] Secondly, the static parameter κ is set to reflect the system stability preference weight; the parameter κ is used to quantify the weight of prioritizing the quality of service (SQoS) and the overall stability of the system in a given task environment. κ also varies between 0 and 1. The default setting may be biased towards stability considerations. The optimal setting value depends on the actual needs of the application scenario and the expected service level requirements.

[0109] Then, the mean expression of α = (β + κ) / 2 is calculated by the above parameters β and κ. A simple and direct method of taking the middle value is adopted here to determine the basic standard for measuring the load importance α. This trade-off method reflects the original intention of the design that considers timely response while not ignoring robustness.

[0110] Finally, under the premise of satisfying β / κ>γ, the algorithm is adjusted to prefer the fast response strategy over the stable strategy execution direction. The core of this step is to trigger a switching mechanism for the scheduling method to focus on speed rather than only emphasizing the optimization of stability and reliability when the response demand ratio that is more sensitive than the target setting γ is detected in real time.

[0111] In addition, by extracting the request response time corresponding to each request of the network load and the number of tasks processed per unit time, the technical solution can comprehensively and meticulously capture the instantaneous state of the network load. This helps to more accurately understand the current operating state of the network and provide a solid foundation for subsequent evaluation and parameter setting. The solution introduces the first network load strength evaluation coefficient (F01) and the second network load strength evaluation coefficient (F02), which comprehensively consider the network load from the perspectives of request response time and task processing number. This dual evaluation mechanism can more comprehensively reflect the strength and characteristics of the network load and improve the accuracy and reliability of the evaluation. The introduction of the dynamic parameter β enables the system to make adaptive adjustments based on the current load situation and the speed of responding to instantaneous changes. By combining the first network load strength evaluation coefficient and the second network load strength evaluation coefficient, the solution can calculate the β value that best suits the current network environment in real time, thereby ensuring that the system can run efficiently and stably. By dynamically adjusting the parameter β, the technical solution can ensure that the system can respond quickly and adjust the resource allocation strategy to meet the immediate network needs when facing different load conditions and business needs. This helps to improve the response speed and stability of the system and enhance the user experience. The design of this technical solution has certain flexibility and scalability. As the network environment and business requirements change, the relevant parameters and formulas can be further adjusted and optimized to adapt to new scenarios and challenges. This makes the solution have high practical value and long-term development potential.

[0112] At the same time, the first method enables the evaluator to intuitively understand the intensity of the immediate response demand of the current network load by directly setting the range and meaning of β. This method focuses on grasping the overall trend of the network load. The second method comprehensively evaluates the network load from a more fine-grained perspective by extracting detailed data of the network load (such as request response time and number of task processing) and calculating the first network load intensity evaluation coefficient (F01) and the second network load intensity evaluation coefficient (F02). This method pays more attention to capturing the detailed characteristics of the network load. Both methods allow the value of β or related parameters to be adjusted according to specific load conditions and business needs, thereby improving the flexibility and adaptability of the evaluation. This helps to ensure that the evaluation results can more accurately reflect the actual situation of the current network environment.

[0113] At the same time, by obtaining the dynamic parameter β, the system can understand the changes in network load in real time and dynamically adjust the network parameters as needed. This helps to optimize the configuration of network resources and improve the overall performance of the network. The introduction of the dynamic parameter β enables the system to respond to changes in network load more quickly. When the network load increases, the system can quickly adjust the resource allocation strategy to meet higher performance requirements. This helps to improve the user's network experience. By accurately evaluating the network load and dynamically adjusting the parameter β, the system can better respond to emergencies in the network, such as traffic surges, equipment failures, etc. This helps to enhance the stability and reliability of the network. In addition, the acquisition and application of the dynamic parameter β enables the network to adapt to different load conditions and business needs more intelligently. This helps to promote the development of network intelligence and improve the automation and intelligence level of the network. By obtaining the dynamic parameter β, network administrators can understand the operating status of the network more accurately and provide strong support for network optimization. This helps to reduce network operation and maintenance costs and improve network operation efficiency.

[0114] In summary, this technical solution effectively improves the response speed and stability of the system by accurately evaluating the network load, comprehensively considering the load intensity, and dynamically adjusting the parameter β, providing a strong guarantee for the efficient operation of the network. At the same time, the technical effects of obtaining the dynamic parameter β through the above two methods are mainly reflected in improving the accuracy of network load evaluation, optimizing network resource configuration and performance, and promoting the intelligent development of the network. These effects jointly promote the advancement and development of network technology and provide users with better quality and more efficient network services.

[0115] For example, in one embodiment, several edge computing nodes are arranged in a sinking distribution network. Considering certain specific application scenarios, such as live video streaming services or real-time data analysis tasks of the Internet of Things, the system is extremely concerned about delay energy and therefore sets a relatively high β=0.8 to emphasize the importance of impromptu response performance; at the same time, κ is given a relatively conservative weight of κ=0.3 because of the need to ensure the basic service quality and the coordination and stability between nodes. At this time, a neutral state factor α=(0.8+0.3) / 2=0.55 is obtained by the calculation formula, indicating that for the current system, although the rapid response characteristics are emphasized, the necessary degree of stability consideration is also given to maintain a smooth transition of the overall operation of the system. Subsequently, according to the defined target rate threshold (for example, γ=1.0), since 0.8 / 0.3>1 under the current conditions, it is decided to further adjust the control variables so that the decision logic is more inclined to take immediate response actions to meet the high-priority delay requirements instead of focusing too much on maintaining the invariance of the existing system framework, thereby ensuring that the edge nodes can better adapt to the current network environment and business characteristics, and achieve the goal of both fast and stable.

[0116] Next, the weight allocation when adjusting the scheduling according to the historical load of the present invention includes the following details:

[0117] Understand the impact of historical load H(i) on the current demand D;

[0118] Considering the impact of network transmission efficiency, f2 = 1 / NetDelayi, where NetDelayi is the actual delay from the central server to the edge node i;

[0119] Combine the above two items f=f1+f2 to determine the total factor, and sort the processing priority of each edge node according to this comprehensive score;

[0120] If any two nodes i and j satisfy f(i)f(j)>σ, then i should be considered as a high-priority option for scheduling;

[0121] The value of σ controls the algorithm's ability to distinguish nodes. When σ is small, it will promote more selection results based on fine-grained evaluation.

[0122] The first step listed is to understand the impact f1 of the historical load H(i) on the current demand D. This process involves obtaining the load level H(i) of each edge node i in similar demand scenarios in the past from historical data, and calculating the impact f1 of this load on the change in the current demand. Specifically, f1 can be defined as a measure of the degree to which the load on node i changes its response time and resource utilization when the central node processed a certain type of request in the past. The behavior of this parameter in the current scenario is predicted by analyzing past data. This parameter has a wide range and can be positive or negative; the optimal value depends on the specific application, but a tendency towards zero means that past load has little negative impact on the present.

[0123] Considering the contribution of network transmission efficiency to the scheduling decision, f2 is set to 1 / NetDelayi, where NetDelayi is the actual delay value from the central server to the edge node i. This indicator reflects the transmission efficiency, and the higher it is, the faster the transmission speed and the lower the delay. For example, in one embodiment, if there are two different path options, one with a higher NetDelay (for example, 50ms) and the other with only 5ms, the corresponding efficiency factors will be significantly different. Shorter time brings greater weight, which promotes the system to prefer faster and more efficient routes.

[0124] Combining the above two points, the formula f=f1+f2 is used as the overall evaluation basis for each node. The higher the comprehensive score, the more suitable the node is for this scheduling. Based on the obtained scores, the edge nodes are prioritized, and nodes with high scores are given priority to complete the current operation.

[0125] For any given pair of points (i, j), if the product of their scoring functions is greater than the threshold σ, it means that i has a higher selection right than j and is scheduled first. This condition allows a preferred decision to be made among many candidate targets.

[0126] σ is set as a key threshold for the accuracy of the algorithm's judgment criteria. The specific value depends on the level of detailed classification expected by the end user. By default, it is hoped that the screening will be as detailed as possible, so a small value close to the ideal situation is set. The smaller σ is, the more detailed the screening results are expected; conversely, a looser evaluation mode is preferred. For example, in a specific scenario, setting a lower σ can better capture resource nodes that are not evenly distributed in the network but still have advantages, even if the differences between those nodes are not particularly significant.

[0127] In a practical example, assume that a video streaming service needs to be dispatched in an edge computing environment, and multiple devices are deployed in different geographical areas to act as receiving stations and data distribution nodes. Historical records show that some locations have been affected by network congestion or excessive equipment load (H), and the link bandwidth and delay data (NetDelay) connecting these areas are currently detected in real time. While f1 is calculated using historical data to reflect the impact of past performance on today's expectations, f2 is obtained using network performance to evaluate the instant transmission efficiency. Finally, the two are summed up to generate a comprehensive score (f), which guides how to efficiently and quickly sink the distribution of tasks. When the f of a specific area is multiplied by other candidates and exceeds the pre-agreed sensitivity limit (such as the above conditions), it will be assigned priority to execute the operation command to be issued. This effectively integrates past behavior data and existing communication status into one consideration.

[0128] Next, the steps of combining multiple influences to calculate the total weight of the present invention are described, including: setting a corresponding standardized value ni for each influence item (such as standardized indicator values ​​such as CPU, bandwidth, etc.); selecting a weighted method to calculate the final weighted score Wi; for example, Wi = Σni*weighti; applying a filtering standard F:Wi>λ (where λ represents the minimum effective allocation threshold) to determine that the node is suitable for allocating services; the parameter λ is used to exclude low-efficiency nodes to ensure service quality and performance.

[0129] First, the specific contents of each step are listed. The first step involves setting specific values ​​for each different performance indicator or parameter as a standard. The standard value here, such as ni, includes the central processing unit load (CPU), input and output bandwidth, and any other information that can be used as an evaluation factor, which are adjusted to a state that can be directly compared - usually through normalization to make each indicator directly comparable. For example, if the CPU capacity of an edge device for the maximum processing task is 98% and the system preset maximum processing load is 100% within a certain time window, it can be concluded that the normalized value of the node on the CPU is ni = 0.98. This processing method not only simplifies the subsequent weighted calculation, but also helps to evaluate the data fusion of different dimensions and ensure fair comparison between different nodes.

[0130] Next, we select the appropriate calculation strategy based on the standardized values ​​corresponding to the different influencing items to calculate the final comprehensive score Wi. To illustrate, the standardized data obtained above is multiplied by their respective weighting coefficients weighti, and then accumulated to obtain Wi = Σni*weighti. The weight factor in this process reflects the relative importance of each evaluation item. For example, the CPU influence coefficient may be higher because the system needs to handle tasks with high computing load, or the corresponding weight may be increased to be greater than other items due to intensive network transmission, which leads to stricter requirements on network delay. This weight coefficient should be carefully formulated by domain experts after understanding the overall environmental requirements and operating goals. Its value should be between 0 and 1, and the effect is most optimized when the sum of all weights is 1.

[0131] To further illustrate this, let's use an example. Assuming a scenario is about rationally selecting service terminal nodes for data storage, analysis, and transmission processing in the network, all objects that meet the requirements can be scored according to the above steps and then screened. Only when a device obtains a score Wi greater than the specified threshold λ, it is considered to have sufficient capacity to participate in this resource allocation work; otherwise, it is considered to not meet the corresponding conditions and should be ignored to improve the reliability and operation effect of the entire system.

[0132] The λ threshold here is to avoid the inclusion of low-efficiency hardware facilities to improve efficiency and ensure service quality. Usually, the initial setting of the threshold should be determined according to the actual business scenario and the results of historical data analysis. In the edge computing scenario, it may be set to the sum of the integrals of the minimum values ​​of various parameters required for normal operation of the system. When the system runs stably, it will be gradually increased to continuously eliminate inferior nodes and achieve the purpose of optimizing cycle upgrades. Such a mechanism can effectively avoid the problem of poor overall performance due to hardware limitations, thereby improving the overall service level and meeting the needs of users to still get a satisfactory interactive experience in a changing environment.

[0133] Next, the present invention describes the use of conditions based on multi-dimensional evaluation to screen available service nodes:

[0134] Check whether the comprehensive availability A of the edge endpoint E is at least equal to the minimum acceptance value L defined by the system, indicating that the node E is in a healthy and operational state (A≥L);

[0135] Additionally check that E's effective service processing capacity SC is higher than the minimum processing level M to ensure actual business operation efficiency (SC>M);

[0136] The above formula ensures that only nodes that meet the minimum specifications and service requirements can be finally selected to assign the service request task.

[0137] The method includes the following steps: First, a comprehensive availability test is performed. The test is intended to determine whether the edge node E has sufficient stability and health before executing the task. The parameter A defined here represents the actual measured comprehensive availability of the node at the current moment, and L is a minimum health level threshold defined for the node at the system level. This value L is usually the best lower limit predicted by the system administrator based on historical data and current workload. Its value range can be a decimal between zero and one, and in practical applications it may be preferably 0.6 or above. Such a configuration can ensure that the node is healthy enough to handle upcoming tasks, thereby avoiding task failures caused by sudden node failures.

[0138] After the above steps, further checking of the service processing capacity SC is also a necessary step. The effective processing capacity here refers to the number of services that can be processed stably and continuously under the current network environment and system configuration conditions. In order to ensure that the edge computing network will not be overloaded or delayed beyond expectations when providing services, a minimum processing level M is set as a threshold. Any node with a service processing capacity lower than this standard will be eliminated and will not appear on the candidate list. The value of M is defined based on the maximum utilization rate of all available resources in the network, such as 10,000 requests per second, to ensure that any service point in the network can carry a certain workload.

[0139] To better understand this process, in an example of distributed edge computing implementation, such as automated production equipment in a smart factory, in order to keep these devices running at high performance and respond quickly, and to transmit a large amount of data to the data center for analysis and decision-making through wireless sensor networks, the above two steps must be checked when selecting which device to receive the new data stream. The first is to check the overall health of the edge device, including but not limited to CPU usage (less than 80%), memory utilization (less than 70%), and storage space availability (more than 20%). If a machine A meets these basic conditions, then continue to see if it has requirements that exceed the standard data throughput capacity. If a machine can process 10,000 packets per second in the past without obvious performance degradation, and the system also stipulates that the minimum capacity must be at least 10,000 packets / s to prevent bottlenecks from affecting the production rhythm, then this machine can be considered a suitable choice.

[0140] This multi-dimensional consideration mechanism effectively improves the quality and stability of services, reduces failure rates and the number of redoes, thereby improving user experience and improving overall system efficiency.

[0141] Next, the load update mechanism of the edge node E of the present invention after it enters the range to be scheduled is described as follows: autonomously perceive the local system resource change situation CR, and regularly report it to the central manager CMS; if the increase in the latest CR value compared with the information reported to CMS at the previous time point T1 reaches the warning limit D, then the node allocation will be re-evaluated immediately; the updated node load level Nlevel is connected to different service levels through linear mapping.

[0142] First, after the edge node E enters the range to be scheduled, E autonomously senses its local resource status and records it as the CR value. This process involves measuring multiple indicators including CPU utilization, memory usage, storage space, and network bandwidth utilization to comprehensively assess the quality of resource status. In order to achieve effective management, the edge node sends the CR data to the central manager CMS at a predetermined time interval for filing.

[0143] In one embodiment, assuming that a data center has many edge computing devices participating in workflow task execution support services, whenever an edge node E comes online and is ready to assist in processing loads, that is, enters the waiting-to-be-scheduled state, the node starts the resource self-monitoring process to obtain the current CR.

[0144] Secondly, suppose that the newly measured current load value obtained after a detection cycle has increased by more than the preset ratio D compared with the data uploaded at time T1. This indicates that the resource occupancy has changed significantly or is facing signs of increased pressure, so it is necessary to take immediate measures to adjust. After receiving the updated CR value, CMS immediately initiates a re-evaluation task to consider how to reasonably re-assign the responsibilities of E to adapt to this changing trend and avoid possible overload.

[0145] Specifically, if a node's previous CR report indicated that the CPU utilization was 60%, and then it was detected to have risen to 80%, the increase exceeded the pre-configured threshold of 20%. This means that the system may encounter transient high traffic challenges, so the scheduler is immediately notified to review the service allocation in the area and make corresponding modification commands based on the latest CR information to maintain global operation stability.

[0146] The next step is the adjustment step. The most important point is to associate the adjusted Nlevel, a quantitative index that represents the workload capacity of the edge device, with the actual service assurance level, which is the so-called service quality mapping rule. The purpose is to convert complex background operations into user experience indicators through simple and intuitive means, and help users choose the appropriate connection method according to their specific application scenario requirements.

[0147] For example, in the scenario of network video services, the following sets of standard mapping relationships can be designed, namely lightweight (such as high-definition playback) and heavyweight (such as real-time interactive live broadcast). Whenever the system recognizes that the performance of the E side is approaching the critical level, it automatically lowers the output quality level until Nlevel returns to the normal area, ensuring that video signal stuttering or even interruption does not occur due to insufficient resources. In addition, certain specific conditions can be set to trigger special operations. For example, when it is found that E is under high operating pressure during peak periods, non-critical function options can be temporarily blocked or the amount of new access requests can be restricted to reduce the burden and thus improve the performance of core services. In terms of parameter setting, various complex environmental factors such as the differences in transmission efficiency under different network conditions and the impact of content cache requirements on the system response time need to be considered for comprehensive decision-making to formulate the most acceptable solution strategy for all parties. Therefore, the establishment of mapping rules should be flexibly customized based on the results of actual experimental verification, and there is no absolute unified template applicable to all scenario analysis and problem-solving situations.

[0148] In terms of formulas, a linear relationship representation similar to (Nlevel = alphaCR + beta) can be established. Here, the coefficient (alpha) represents the amplification factor of the impact of each unit change in the CR value on the load level Nlevel, and the constant term (beta) is used to balance and correct the deviation of the starting base value. Theoretically speaking, it is more appropriate to select the interval (0 < alpha < 1) because too high a value will overly amplify the errors caused by small fluctuations, and too small a value will be too insensitive to respond to large-direction changes. Ideally, a compromise between the two needs to be found to achieve the maximum balance of sensitivity and anti-noise interference compatibility. Similarly, there is no fixed recommended value for the base value, which generally depends on the average load expectation target positioning and the allowable deviation range under a specific framework to determine the final value.

[0149] In the present invention, the load data collector is responsible for regularly collecting historical load data such as CPU utilization and memory usage from each edge node. The importance factor calculator calculates α based on the immediate response demand β and the stability preference κ, and adjusts the algorithm bias based on the relationship between β / κ and γ. The weight allocator determines the comprehensive weight W(i) of each edge node based on the historical load H(i), the impact f1 of the current demand D, and the impact f2 of the network transmission efficiency. The request receiver accepts the service request from the terminal device, the data type identifier is used to analyze the data type of the service request and match the appropriate service node. The task allocator selects the appropriate edge node to process the service request based on the comprehensive weight W(i); the real-time monitor is used to continuously monitor the status of all edge nodes and the service execution status. The feedback regulator adjusts the scheduling strategy based on the monitored data to ensure the optimal resources. Source utilization; the log updater is used to update the log records of the edge node after each task is completed. The fluctuation detector is used to analyze the resource usage changes Δi, t and Memorydeltai, t in the past t period to determine whether they exceed the threshold θ; the availability verifier is used to check whether the comprehensive availability A of the edge node E meets the minimum standard L. The service capacity evaluator is used to confirm that the effective service processing capacity SC of the node E is higher than the minimum level M; the local resource change sensor is used to automatically perceive the local system resource changes CR of the edge node E. The status reporter is used to regularly report the latest node status to the central manager CMS. The warning trigger is used to trigger a re-evaluation when the latest CR value increases compared with the information at time T1 and reaches the warning limit D. The service level mapper dynamically adjusts the corresponding service level according to the updated node load level Nlevel.

[0150] The edge node scheduling method based on the sinking distributed network of the present invention includes: using a preset load balancing algorithm in the distributed network to dynamically determine the scheduling strategy of each edge node, which can not only optimize resource allocation, ensure that in the high concurrent request scenario, by dynamically adjusting the data interaction frequency and synchronization interval between edge nodes, thereby effectively avoiding the inconsistency problem caused by insufficient data synchronization mechanism, but also can quickly update the optimal path selection according to the change of the current network topology, reducing the additional communication delay caused by the sudden change of network structure. Secondly, the method classifies and grades the received service requests through an intelligent algorithm, and when the terminal user sends a request, the service node most suitable for this type of workload is automatically matched according to the specific type of the service request for deployment and scheduling, so as to solve the practical difficulties such as improper resource allocation, inefficient task allocation and slow response speed caused by scarce and heterogeneous computing resources at the edge. Furthermore, considering that for large data volume analysis and calculation, storage and reading operations may become a potential efficiency constraint, we also add an efficient local buffering mechanism and an intelligent prediction strategy in this process to reduce the unnecessary time consuming caused by repeated acquisition of the same information resources. At the same time, this patent solution fully considers the requirements of information security and confidentiality. It combines encrypted channels with identity authentication during multi-regional and cross-border remote connections to enhance the security protection performance during transmission and reduce the probability of abnormal leakage of important information. In this way, a safe and reliable new generation of intelligent infrastructure management system is built that can not only meet the requirements of efficient operation in complex network environments but also ensure that the privacy of various users is not violated.

[0151] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the embodiments of the present disclosure. It should be understood that the above description is only the specific implementation method of the embodiments of the present disclosure and is not intended to limit the protection scope of the embodiments of the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present disclosure should be included in the protection scope of the embodiments of the present disclosure.

Claims

1. An edge node scheduling method based on a distributed network is characterized in that: include: Determine the scheduling strategy of edge nodes based on the preset load balancing algorithm; Receiving service requests from terminal devices; identifying an appropriate service node based on the data type of the service request; Allocating the service node to process the service request; Adjust scheduling strategies based on real-time monitoring data feedback.

2. The edge node scheduling method based on the sinking distributed network according to claim 1 is characterized in that: The scheduling strategy for edge nodes based on the preset load balancing algorithm includes: Collect historical load information of each edge node; Determine an importance factor α for measuring historical load; Assign historical load weights, where weight W(i) = α*H(i), H(i) is the historical load value of the i-th edge node; The edge node scheduling strategy is determined based on the current system status and the historical load of the edge nodes.

3. The edge node scheduling method based on the sinking distributed network according to claim 2 is characterized in that: The historical load conditions of each edge node are collected, including: Update the log records of edge nodes after each task is completed; Separate the CPU utilization percentage CPUp(i,t) and memory usage ratio Memoryp(i,t) from the historical data, where i represents the i-th edge node and t represents the time point; Use statistical functions to calculate the CPU utilization change ΔCPUp(i,T) and memory usage change ΔMemoryp(i,T) of each edge node in the past t time period. Specifically: The CPU utilization change is defined as: ΔCPUp(i,T)=CPUp(i,t+T)-ΔCPUp(i,t); The memory usage change is defined as: ΔMemoryp(i,T) = ΔMemoryp(i,t+T) - ΔMemoryp(i,t) If ΔCPUp(i,T)+ΔMemoryp(i,T)>θ, it is considered that the resource usage fluctuation in the time period T is significant, and it is necessary to consider adjusting the load balancing strategy, where θ is a preset threshold used to determine whether the fluctuation range of resource usage reaches the level that affects the load balancing strategy.

4. The edge node scheduling method based on the sinking distributed network according to claim 3 is characterized in that: Determining an importance factor α for weighing historical loads involves the following steps: The dynamic parameter β is set to represent the intensity of the immediate response demand of the network load; The static parameter κ is set to reflect the system stability preference weight; Then α is determined by finding the average function α = (β + κ) / 2; If β / κ>γ, adjust the algorithm parameters to favor rapid response rather than robustness; Under this condition, β represents the system's emphasis on delay sensitivity, while κ focuses on stability and service quality.

5. The edge node scheduling method based on the sinking distributed network according to claim 4 is characterized in that: The details of weight allocation when adjusting scheduling based on historical load include: Understand the impact of historical load H(i) on the current demand D; Considering the impact of network transmission efficiency, f2 = 1 / NetDelayi, where NetDelayi is the actual delay from the central server to the edge node i; Combine the above two items f=f1+f2 to determine the total factor, and sort the processing priority of each edge node according to this comprehensive score; If any two nodes i and j satisfy f(i)f(j)>σ, then i should be considered as a high-priority option for scheduling; The value of σ controls the algorithm's ability to distinguish nodes.

6. The edge node scheduling method based on the sinking distributed network according to claim 5 is characterized in that: The steps to calculate the total weight when combining multiple influences include: Set the corresponding standardized value ni for each influencing item; Select a weighted method to calculate the final weighted score Wi; A filtering criterion F:Wi>λ is applied to determine that the node is suitable for allocating services; The parameter λ is used to exclude low-performance nodes to ensure service quality and performance.

7. The edge node scheduling method based on the sinking distributed network according to claim 6 is characterized in that: Use conditions based on multi-dimensional evaluation to filter available service nodes: Check whether the comprehensive availability A of the edge endpoint E is at least equal to the minimum acceptance value L defined by the system, indicating that the node E is in a healthy and operational state; It is additionally checked that the effective service processing capacity SC of E is higher than the minimum processing level M to ensure the actual business operation efficiency.

8. The edge node scheduling method based on the sinking distributed network according to claim 7 is characterized in that: When edge node E enters the range to be scheduled, its load status update mechanism is: Autonomously sense the local system resource changes CR and regularly report to the central manager CMS; If the increase in the latest CR value compared to the information reported to CMS at the previous time point T1 reaches the warning limit D, the node allocation will be re-evaluated immediately; The updated node load level Nlevel is connected to different service levels through linear mapping.

9. The edge node scheduling system based on the sinking distributed network is characterized by: include, Load balancing strategy module, service request processor, real-time monitoring and feedback adjustment module, historical data analysis module, edge node health check module and adaptive load update mechanism; The load balancing strategy module includes a load data collector, an importance factor calculator and a weight allocator; The service request processor includes a request receiver, a data type identifier, and a task distributor; The real-time monitoring and feedback regulation module includes a real-time monitor and a feedback regulator; The historical data analysis module includes a log updater and a fluctuation detector; The edge node health check module includes an availability verifier and a service capability evaluator; The adaptive load update mechanism includes local resource change sensor, status reporter, warning trigger and service level mapper.

10. A computer device comprising: A memory and a processor; the memory stores a computer program, wherein the processor implements the steps of the method according to any one of claims 1 to 8 when executing the computer program.

Images