A Method and System for Elastic Resource Scheduling in Education Based on Microservice Orchestration

Abstract

This invention discloses a method and system for flexible educational resource scheduling based on microservice orchestration, relating to the field of educational resource scheduling technology. This method acquires educational resource data (including educational support resource data and user behavior data) for a designated educational platform within a set period; analyzes the platform's resource elasticity feature set (including educational resource supply characteristic values ​​and educational demand intensity characteristic values) based on the platform's resource elasticity feature set; and analyzes the platform's educational resource adaptation characteristic values ​​based on the platform's resource elasticity feature set. This invention performs microservice orchestration resource scheduling on the designated educational platform based on the educational resource adaptation characteristic values, effectively avoiding resource scheduling lag and coarse control, thereby enabling the educational platform to maintain refined resource allocation under different load conditions.

Description

Technical Field

[0001] This invention relates to the field of educational resource scheduling technology, specifically to a method and system for flexible educational resource scheduling based on microservice orchestration. Background Technology

[0002] Microservice orchestration is an important technology in cloud computing and distributed systems in recent years. By breaking down complex applications into multiple functionally independent microservices, their running instances are deployed on corresponding service nodes and run in a containerized environment, enabling independent deployment, scaling, and recycling of services. With the help of an orchestration framework, the number, running status, and resource allocation of microservice instances on different service nodes can be dynamically scheduled, thereby achieving elastic scaling when the load fluctuates.

[0003] In the rapid development of educational informatization, online education platforms have undertaken diverse teaching tasks such as homework writing, grading, interaction, and activities. For example, the Zhixietong platform supports students in writing essays and automatic grading, along with actions such as image uploading, interactive comments, and group activities. These tasks exhibit significant volatility: during peak periods of homework submission, exam grading, or activities, the number of user requests surges, leading to a sharp increase in platform resource demand; while during off-peak periods, some resources remain idle. Existing education platforms mostly adopt static configuration or experience-based expansion, making it difficult to dynamically schedule resources based on user behavior and resource supply and demand, which can easily lead to delays due to insufficient resources or waste due to excess resources.

[0004] Existing technology, such as the patent application with publication number CN117215788A, discloses a dynamic elastic scaling method based on microservice load balancing, comprising the following steps: S1 The user defines a workflow through a graphical interface, and the system automatically generates a workflow description file based on the user input and submits it to the workflow parser; S2 The workflow parser parses the workflow description and generates corresponding container startup commands, which are then handed over to the container scheduler; S3 The container scheduler selects a suitable host to run the container based on the host resource usage. This dynamic elastic scaling method based on microservice load balancing uses a workflow auto-scaling model based on CPU utilization. When the workflow is overloaded, it automatically expands the number of containers in the workflow, reducing task waiting time. When the task load decreases, it can scale down the workflow while ensuring no task loss, thus saving resources and costs. By deploying an elastic scaling cluster, it significantly improves the overall computing resource utilization of the business support system and reduces the use of computing resources.

[0005] Based on the above findings, the limitations of existing technologies include at least the following issues: Existing technologies lack analysis of the matching between the supply capacity of educational resources and the intensity of user demand, which can easily lead to lag and coarse scheduling, making it difficult to accurately reflect the true suitability of educational resources. For example, when peak user activity occurs, if analysis is based solely on resource data from service nodes, it may show that demand growth is accumulating but has not yet triggered expansion conditions, easily leading to service node response delays and request backlogs. Furthermore, there is a dynamic correlation between resource carrying capacity and fluctuations in user demand; these fluctuations are sudden and uneven. Existing technologies struggle to comprehensively analyze the supply-demand relationship, easily resulting in idle resources on certain nodes, ultimately leading to an imbalance between resource supply and demand. Therefore, existing technologies cannot detect supply-demand gaps in advance, nor can they promptly recover redundant resources when resources are abundant, and they lack refined control capabilities. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for scheduling elastic educational resources based on microservice orchestration, which solves the problem that existing technologies lack supply and demand matching analysis, resulting in scheduling lag and a lack of fine-grained control.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a microservice orchestration-based method for scheduling elastic educational resources, comprising the following steps: within a set period, acquiring educational resource data of a set educational platform, including educational support resource data and user behavior data; based on the educational resource data of the set educational platform, analyzing the resource elasticity feature set of the set educational platform, including educational resource supply feature values ​​and educational demand intensity feature values; based on the resource elasticity feature set of the set educational platform, analyzing the educational resource adaptation feature values ​​of the set educational platform; and performing microservice orchestration resource scheduling processing on the set educational platform based on the educational resource adaptation feature values.

[0008] Furthermore, the educational support resource data includes the response latency value, processing capacity value, microservice call chain latency fluctuation rate value, energy consumption conversion efficiency value, operating temperature drift rate value, cache hit rate value, and service anomaly rate value of each service node. The specific steps for analyzing and setting the educational resource supply characteristic values ​​of the educational platform are as follows: Based on the educational support resource data of the educational platform, analyze and set the resource carrying capacity service sequence characteristic set of the educational platform, including the resource carrying capacity characteristic value and resource efficiency characteristic value of each service node; based on the resource carrying capacity service sequence characteristic set of the educational platform, analyze and set the educational resource supply characteristic values ​​of the educational platform.

[0009] Furthermore, the specific steps for analyzing the resource carrying capacity service sequence feature set of the setting education platform are as follows: Based on the response latency value, processing capacity value, and microservice call chain latency fluctuation value of each service node of the setting education platform, analyze the resource carrying capacity feature value of the corresponding service node; Based on the energy consumption conversion efficiency value, operating temperature drift rate value, cache hit rate value, and service anomaly rate value of each service node of the setting education platform, analyze the resource efficiency feature value of the corresponding service node.

[0010] Furthermore, the user behavior data includes a homework image dataset, homework submission rate, user interaction activity factor, activity triggering peak factor, and differentiated demand factor. The specific steps for analyzing the educational demand intensity characteristic value of the setting education platform are as follows: Based on the pre-trained homework content parsing model and combined with the homework image dataset of the setting education platform, analyze the intelligent grading resource demand characteristic value of the setting education platform; Based on the homework submission rate, user interaction activity factor, activity triggering peak factor, and differentiated demand factor of the setting education platform, analyze the user behavior intensity characteristic value of the setting education platform, and combined with the intelligent grading resource demand characteristic value, analyze the educational demand intensity characteristic value of the setting education platform.

[0011] Furthermore, the homework image dataset includes several homework image data. The specific steps for analyzing the intelligent grading resource requirement feature values ​​of the setting education platform are as follows: input each homework image data of the setting education platform into the pre-trained homework content parsing model, analyze the content resource mapping feature set in the corresponding homework image, including homework clarity interpretation feature value, grading load feature value, and expression structure breadth feature value; based on the content resource mapping feature set in each homework image of the setting education platform, analyze the intelligent grading resource requirement feature values ​​of the setting education platform.

[0012] Furthermore, the task image data includes the pixel value and two-dimensional coordinates of each pixel in the task image, and the task content parsing model includes an input layer, an image perception layer, a content encoding layer, and an output layer.

[0013] Furthermore, the specific steps for analyzing the content resource mapping feature set of each assignment image on the setting education platform are as follows: In the input layer of the assignment content parsing model, the data of each assignment image on the setting education platform is received and preprocessed; in the image perception layer of the assignment content parsing model, based on the preprocessed data of each assignment image on the setting education platform, the pixel set and text vector of the corresponding assignment image are extracted; in the content encoding layer of the assignment content parsing model, based on the pixel set and text vector of each assignment image on the setting education platform, the content resource requirement feature vector of the corresponding assignment image is extracted; in the output layer of the assignment content parsing model, based on the content resource requirement feature vector of each assignment image on the setting education platform, the content resource mapping feature set of the corresponding assignment image is output.

[0014] Furthermore, the specific steps for calculating the educational resource adaptation feature values ​​of the educational platform are as follows: ;in, To set the educational resource adaptation feature values ​​for the education platform, To define the characteristic value of educational demand intensity for the education platform, To set the characteristic values ​​of educational resource supply for the education platform, The smoothing adjustment coefficients are stored in the database. These are the difference adjustment coefficients stored in the database.

[0015] Furthermore, the specific steps for microservice orchestration and resource scheduling of the designated education platform based on the educational resource adaptation feature value are as follows: the educational resource adaptation feature value of the designated education platform is compared with the preset educational resource adaptation feature range; based on the judgment result, corresponding microservice orchestration and resource scheduling measures are taken for the designated education platform.

[0016] The microservice orchestration-based elastic resource scheduling system for education includes: an education resource data acquisition module, used to acquire education resource data of a designated education platform within a set period, including education support resource data and user behavior data; a resource elasticity feature analysis module, used to analyze the resource elasticity feature set of the designated education platform based on the education resource data, including education resource supply feature values ​​and education demand intensity feature values; an education resource adaptation analysis module, used to analyze the education resource adaptation feature values ​​of the designated education platform based on the resource elasticity feature set; and a microservice orchestration scheduling module, used to perform microservice orchestration resource scheduling processing on the designated education platform based on the education resource adaptation feature values.

[0017] The present invention has the following beneficial effects:

[0018] (1) The microservice orchestration-based educational elastic resource scheduling method collects educational resource data and establishes educational resource supply characteristic value and educational demand intensity characteristic value. Based on this, it generates educational resource adaptation characteristic value to characterize the degree of matching between platform supply and demand. Thus, before a surge in user demand occurs, it can determine whether there is a supply-demand gap based on the dynamic feedback of educational resource adaptation characteristic value, and then take corresponding microservice orchestration resource scheduling measures. This effectively avoids resource scheduling lag and rough control, and enables the educational platform to maintain refined resource allocation under different load conditions. This ensures that user requests do not accumulate during peak periods and avoids waste caused by resource surplus, while improving the accuracy of resource scheduling.

[0019] (2) The educational elastic resource scheduling method based on microservice orchestration collects user behavior data and combines it with a pre-trained homework content parsing model to perform in-depth analysis on the homework image data dataset in the user behavior data, thereby generating intelligent grading resource demand feature values. This enables the platform to comprehensively identify the intensity of user demand, covering both the frequency and volatility at the behavioral level and capturing the real demand for resource consumption at the homework content level. This allows for more accurate analysis of the dynamic changes in demand intensity, thereby achieving more reasonable resource scheduling and ensuring the stability of services such as grading.

[0020] (3) The educational elastic resource scheduling method based on microservice orchestration analyzes educational support resource data, establishes resource carrying capacity characteristic value and efficiency characteristic value, and synthesizes them into educational resource supply characteristic value. This enables the platform to achieve a comprehensive characterization on the supply side, thereby reflecting the carrying capacity of service nodes under high load. By combining supply characteristics with demand characteristics, the platform can make differentiated decisions on the scaling order and degree of different nodes during the microservice orchestration process, prioritize resource compensation for low-efficiency or high-latency nodes, and reasonably reclaim redundant resources of excess nodes. This achieves dynamic balance and refined control of resource allocation and improves the overall operating efficiency of the educational platform.

[0021] (4) The microservice orchestration-based educational elastic resource scheduling system achieves precise adjustment of the supply and demand relationship of educational resources through the synergistic effect between various modules. The educational resource data acquisition module ensures the synchronous acquisition of educational support resource data and user behavior data. The resource elasticity feature analysis module transforms the raw data into supply and demand features that can be directly used for scheduling judgment. On this basis, the educational resource adaptation analysis module conducts a unified evaluation of supply and demand features and generates educational resource adaptation feature values ​​that reflect the degree of supply and demand matching, thereby revealing the balance between resource carrying capacity and demand fluctuations. The microservice orchestration scheduling module performs elastic scaling and task allocation based on the adaptation feature values, realizes dynamic coordination and optimization between service nodes, and improves the agility of overall resource scheduling, enabling the educational platform to maintain efficient operation under high concurrency demand scenarios.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart of the educational elastic resource scheduling method based on microservice orchestration of the present invention.

[0024] Figure 2 This is a schematic diagram illustrating the feature set data of the content resource mapping of the homework image sequence of the education platform in the educational elastic resource scheduling method based on microservice orchestration of the present invention.

[0025] Figure 3 This is a flowchart illustrating the specific steps involved in analyzing and setting the content resource mapping feature set for each assignment image on the education platform in the microservice orchestration-based educational elastic resource scheduling method of this invention.

[0026] Figure 4 This is a block diagram of the educational elastic resource scheduling system based on microservice orchestration of the present invention. Detailed Implementation

[0027] Please see Figure 1This invention provides a technical solution: a microservice orchestration-based method for scheduling elastic educational resources, comprising the following steps: acquiring educational resource data of a designated educational platform (such as a Zhixietong platform) within a set period (e.g., 5 seconds, 10 seconds, 30 seconds, or 1 minute), including educational support resource data and user behavior data; analyzing the resource elasticity feature set of the designated educational platform based on the educational resource data, including educational resource supply feature values ​​and educational demand intensity feature values; analyzing the educational resource adaptation feature value of the designated educational platform based on the resource elasticity feature set (used to characterize the matching degree between platform resource supply and user demand; if the value is large and greater than 1, it indicates that the resource supply is sufficient to meet the current demand; if the value is small and less than 1, it indicates that the resource supply is insufficient and it is difficult to fully meet the demand); and performing microservice orchestration resource scheduling processing on the designated educational platform based on the educational resource adaptation feature value.

[0028] The specific steps for calculating the educational resource adaptation feature values ​​of the educational platform are as follows: ;in, To set the educational resource adaptation feature values ​​for the education platform, To define the characteristic value of educational demand intensity for the education platform, To set the characteristic values ​​of educational resource supply for the education platform, The smoothing adjustment coefficients are stored in the database. The difference adjustment coefficient is stored in the database, and in this embodiment, the smoothing adjustment coefficient is stored in the database. Difference adjustment coefficient The values ​​were 0.001 and 0.500, respectively.

[0029] Specifically, the educational support resource data includes the response latency, processing capacity, microservice call chain latency fluctuation, energy efficiency, operating temperature drift rate, cache hit rate, and service anomaly rate of each service node (i.e., the running unit that hosts microservice instances of the educational platform, such as a container, virtual machine, or computing unit on a physical server). The specific steps for analyzing and setting the educational resource supply characteristic values ​​of the educational platform are as follows:

[0030] Based on the educational support resource data of the established education platform, the resource carrying capacity service sequence feature set of the established education platform is analyzed, including the resource carrying capacity feature value and resource efficiency feature value of each service node. Based on the resource carrying capacity service sequence feature set of the established education platform, the educational resource supply feature value of the established education platform is analyzed. Specifically, the resource carrying capacity feature value and resource efficiency feature value of each service node of the established education platform are weighted to obtain the load adaptation feature value of its corresponding service node. The load adaptation feature values ​​of all service nodes are then averaged to obtain the educational resource supply feature value of the established education platform. This feature value is used to characterize the resource supply level that the education platform can provide within a set period, reflecting the platform's overall support capability for educational services within that period.

[0031] It should be noted that in this implementation example, the weight coefficients of each parameter in the weighted processing can be obtained using sample entropy weights. Taking the weighted processing of load adaptation feature values ​​as an example, the resource carrying capacity feature value and resource efficiency feature value of each service node of the education platform are read and their corresponding information entropy values ​​are extracted. Then, their corresponding information entropy values ​​are transformed using the reciprocal suppression mapping function f(x)=1 / (1+x), such as 1 / (1+information entropy value of resource carrying capacity feature value), and summed to obtain the information entropy sum value. The corresponding transformed information entropy values ​​are then compared with the information entropy sum value to obtain the weight coefficients corresponding to each parameter. It should also be noted that if there is a parameter that is negatively correlated with its corresponding weighted result, before performing entropy weight calculation, the parameter is subjected to directional reverse transformation processing, that is, the original value of the parameter is reverse transformed in the form of 1 / (1+original value of the parameter).

[0032] The response latency value is the average round-trip time for processing user requests within the specified period. It can be obtained by using distributed call chain tracing tools (such as Jaeger, Zipkin, and SkyWalking) to obtain the timestamp when the user request enters the node and the timestamp when the request returns a response. The difference between the two (take the absolute value) is the round-trip latency of a single request. The round-trip latency of all requests within the set period is statistically averaged, and the result is used as the response latency value.

[0033] The processing load value is the processing load level of the service node within a set period. It can be obtained by acquiring the CPU utilization, memory usage, and disk I / O rate at each time point. The CPU utilization is the ratio of the processor used by the service node at the set time point, the memory usage is the ratio of the node's used memory to the total memory at the set time point, and the disk I / O rate is the node's read and write operation rate at the set time point. The above data is obtained by the resource monitoring interface provided by the service node's operating environment. The CPU utilization, memory usage, and disk I / O rate at each time point are standardized and weighted based on the results to extract the processing load value.

[0034] The microservice call chain latency volatility value is the latency stability of service nodes in the microservice call chain within a set period. It can be obtained by using distributed call chain tracing tools (such as Jaeger, Zipkin, and SkyWalking) to record the latency sequence of all call chains, and then processing the standard deviation of the sequence to obtain the result as the microservice call chain latency volatility value.

[0035] The energy conversion efficiency value is the ratio between the effective task processing volume and energy consumption of a service node within a set period. It is used to characterize the degree of matching between node energy consumption and performance. It can be obtained by collecting real-time power at each time point (e.g., through hardware power sensors, energy consumption estimation interfaces of virtualization platforms, or energy consumption statistics tools of container environments), and calculating the total number of processing tasks completed by the service node within the period. The real-time power at each time point within the period is integrated to obtain the total energy consumption of the corresponding service node. The ratio of the total number of processing tasks of the service node to the total energy consumption is then processed, and the result is used as the energy conversion efficiency value.

[0036] The operating temperature drift rate is the average rate of change of the service node temperature over time within a set period. It is used to characterize the thermal stability of the node during operation. It can be obtained by calling temperature sensors (such as motherboard sensors, IPMI interfaces, etc.) to collect the temperature value at each time point, and then using the least squares method to perform linear fitting to obtain the temperature slope value of each service node. The result is used as the operating temperature drift rate value.

[0037] The cache hit rate is the service node's cache utilization efficiency within a set period. It can be used to characterize the degree of response optimization of the node under high concurrency requests. It can be calculated by statistically analyzing the total number of cache accesses (the total number of cache requests initiated by the service node in this period) and the number of hits (the number of cache requests successfully hit by the service node in this period). Both can be statistically analyzed from the cache access logs stored in the database, and the ratio of the number of hits to the total number of accesses is calculated. The result is used as the cache hit rate value.

[0038] The service anomaly rate is the probability of service node calls being abnormal within a set period. It can be calculated by statistically analyzing the total number of service calls within the period (the total number of requests received by the service node within the period, which can be obtained from the service call logs stored in the database or the call count information of the microservice framework) and the number of call failures (such as timeouts, connection errors, etc., which can be obtained from the error logs stored in the database), and then calculating the ratio of the number of failures to the total number of calls, and using the result as the service anomaly rate.

[0039] The specific steps for analyzing the resource carrying capacity service sequence feature set of the education platform are as follows: Based on the response latency, processing capacity, and microservice call chain latency volatility of each service node of the education platform, analyze the resource carrying capacity feature value of the corresponding service node. Specifically, standardize the response latency, processing capacity, and microservice call chain latency volatility of each service node of the education platform (Min–Max interval standardization method can be used). Based on the standardization results, perform weighted processing. In this weighted processing, the standardized response latency and microservice call chain latency volatility are both reciprocally processed, such as 1 / (1+standardized response latency), to obtain the resource carrying capacity feature value of the corresponding service node. This is used to characterize the actual computing power support capability of the education platform within a set period. The higher the resource carrying capacity feature value, the more the platform can support education business requests with faster speed and more stable links.

[0040] Based on the energy conversion efficiency, operating temperature drift rate, cache hit rate, and service anomaly rate of each service node in the educational platform, the resource efficiency characteristics of the corresponding service nodes are analyzed. Specifically, the energy conversion efficiency, operating temperature drift rate, cache hit rate, and service anomaly rate of each service node in the educational platform are standardized (using the Min–Max interval standardization method). The standardized results are then weighted, and during this weighting process, the standardized operating temperature drift rate and service anomaly rate are reciprocally calculated, such as 1 / (1 + standardized operating temperature drift rate), to obtain the resource efficiency characteristics of the corresponding service nodes. These characteristics characterize the resource utilization efficiency of the educational platform within a set period. A higher resource efficiency characteristic value indicates that the platform completes more effective tasks with the same energy consumption while maintaining stable operation.

[0041] In this implementation plan, educational support resource data is broken down into resource carrying capacity characteristic values ​​and resource efficiency characteristic values ​​for each service node. This data is then used to analyze the overall educational resource supply characteristic values ​​of the platform, enabling a precise assessment of the resource supply level. Specifically, indicators such as response latency and call chain fluctuations reflect the processing speed and link stability of nodes, while energy conversion efficiency and temperature drift reveal the utilization rate and operating status of hardware resources. Cache hit rate and service anomaly rate reflect the optimization capability and fault tolerance level under concurrent requests. By unifying these parameters into load adaptation characteristic values, the supply characteristic values ​​can cover both computing power and stability. This allows the platform to make dynamic scheduling based on the comprehensive supply level when facing fluctuations in user demand, thereby making resource allocation more refined and improving the accuracy of supply and demand matching and resource utilization efficiency.

[0042] Specifically, user behavior data includes homework image datasets, homework submission rates, user interaction activity factors, activity triggering peak factors, and differentiated demand factors. The specific steps for analyzing the educational demand intensity characteristics of the setting education platform are as follows: Based on a pre-trained homework content parsing model and combined with the homework image dataset of the setting education platform, analyze the intelligent grading resource demand characteristics of the setting education platform; Based on the homework submission rate, user interaction activity factors, activity triggering peak factors, and differentiated demand factors of the setting education platform, analyze the user behavior intensity characteristics of the setting education platform (i.e., standardize the homework submission rate, user interaction activity factors, activity triggering peak factors, and differentiated demand factors of the setting education platform (using the Min–Max interval standardization method), and weight the results to obtain the user behavior intensity characteristics); Combined with the intelligent grading resource demand characteristics, analyze the educational demand intensity characteristics of the setting education platform, specifically: weight the user behavior intensity characteristics and intelligent grading resource demand characteristics of the setting education platform to obtain the educational demand intensity characteristics of the setting education platform.

[0043] The assignment submission rate is the frequency of user assignment submissions within a set period. It can be obtained by counting the number of assignment submission requests initiated by all users through the writing and grading modules within the period on the education platform (which can be obtained from the upload events recorded in the education platform logs stored in the database).

[0044] The user interaction activity factor is the frequency of user interaction on the education platform within a set period. It can be calculated by counting the number of comments each user makes on the education platform within that period (the number of comments made by users on articles, assignments, or activities, which can be obtained from the comment event logs stored in the database) and the number of group interactions (the number of times users post, reply, like, etc., which can be obtained from the event logs stored in the database, which record fields such as the type of interaction event, user ID, group ID, and timestamp; counting all group interaction events within the set period to obtain the number of group interactions for each user), normalizing the results, weighting them, and summing them up. The result is used as the user interaction activity factor.

[0045] The activity triggering peak factor is the maximum concurrent request value caused by activities on the education platform within a set period. It can be obtained by statistically analyzing the time series of all user requests within the period on the education platform and detecting the maximum instantaneous number of concurrent requests in the time series within the period. The number of concurrent requests can be obtained by statistically analyzing the user request timestamps and session identifiers recorded in the service call logs of the platform stored in the database.

[0046] The differentiated demand factor is the ratio of the number of times users call advanced functions to the number of times they call ordinary functions within a set period. It can be obtained by statistically analyzing the number of times users call advanced functions (such as AI-assisted writing, intelligent essay correction, and writing skill recommendations) and ordinary functions (such as essay submission and comment interaction) within the period on the education platform, and then performing a summation process on the ratio of the two. The result is used as the differentiated demand factor, and the number of calls can be obtained from the call logs and usage records stored in the database.

[0047] The homework image dataset includes several homework images. The specific steps for analyzing the intelligent grading resource demand characteristics of the designated education platform are as follows: Each homework image from the designated education platform is input into a pre-trained homework content parsing model. The content resource mapping feature set of the corresponding homework image is analyzed, including homework clarity feature value, grading load feature value, and expression structure breadth feature value. Based on the content resource mapping feature set of each homework image from the designated education platform, the intelligent grading resource demand characteristics of the designated education platform are analyzed. Specifically, based on the homework clarity feature value, grading load feature value, and expression structure breadth feature value of each homework image from the designated education platform, the evaluation resource demand characteristic value of the corresponding homework image is analyzed (used to characterize the degree of resource demand of the education platform for this single homework during automatic grading). The evaluation resource demand characteristic values ​​of all homework images are summed to obtain the intelligent grading resource demand characteristic value of the designated education platform (used to characterize the overall resource demand intensity of the education platform when batch processing homework within a set period).

[0048] The specific formula for calculating the feature value of the evaluation resource demand in a certain homework image of the educational platform is as follows: ;in, To define the characteristic values ​​for judging resource needs in a specific assignment image on an educational platform, To define clear reading feature values ​​for a specific assignment image on an educational platform, The judgment adjustment coefficients are stored in the database. To set the grading load characteristic value in a specific homework image on the education platform, The batch load adjustment coefficients are stored in the database. To define the breadth feature value of the expressive structure in a specific homework image on an educational platform, This refers to the structure breadth adjustment coefficient stored in the database. And the judgment adjustment coefficients stored in the database , Modify load adjustment coefficient Structural breadth adjustment coefficient The values ​​were 0.328, 0.367, and 0.305, respectively.

[0049] The following is a specific implementation example of calculating the evaluation resource demand feature value in a homework image from a designated education platform. The available data includes the homework clarity interpretation feature value, grading workload feature value, and expression structure breadth feature value from five randomly selected homework images from the designated education platform, as detailed in Table 1 and... Figure 2 As shown:

[0050] Table 1. Example of Feature Set Data for Mapping Content Resources of Homework Image Sequences on an Educational Platform

[0051] Clearly interpret the feature values ​​in the assignment Modify load characteristic value Expressing the breadth of structural features Image 1 of the assignment 0.892 0.432 0.467 Image 2 of the assignment 0.924 0.547 0.573 Image 3 of the assignment 0.859 0.492 0.514 Image 4 of the assignment 0.874 0.673 0.584 Image 5 of the assignment 0.768 0.638 0.726

[0052] The judgment adjustment coefficient stored in the database The value is: 0.328;

[0053] Batch load adjustment coefficients stored in the database The value is: 0.367;

[0054] The structure breadth adjustment coefficient stored in the database The value is: 0.305;

[0055] Substituting the data from Table 1 and the aforementioned coefficients into the specific formula for calculating the characteristic value of the evaluation resource demand in a certain homework image of the educational platform, we obtain:

[0056] The characteristic value of the evaluation resource requirement in the first assignment image of the education platform is set to 0.328×(1-0.892)+0.367×0.432+0.305×0.467≈0.336;

[0057] The characteristic value of the evaluation resource requirement in the second assignment image of the education platform is set to 0.328×(1-0.924)+0.367×0.547+0.305×0.573≈0.400;

[0058] The characteristic value of the evaluation resource requirement in the third homework image of the education platform is set as 0.328×(1-0.859)+0.367×0.492+0.305×0.514≈0.384;

[0059] The characteristic value of the evaluation resource requirement in the fourth homework image of the education platform is set as 0.328×(1-0.874)+0.367×0.673+0.305×0.584≈0.466;

[0060] The characteristic value of the evaluation resource requirement in the fifth homework image of the education platform is set to 0.328×(1-0.768)+0.367×0.638+0.305×0.726≈0.532.

[0061] The task image data includes the pixel value and two-dimensional coordinates of each pixel in the task image. The task content parsing model includes an input layer, an image perception layer, a content encoding layer, and an output layer.

[0062] In this implementation plan, user behavior data is combined with assignment image data to analyze the intensity of educational demand, thereby more accurately reflecting the platform's demand level. Secondly, with the addition of assignment image analysis, the platform can not only identify how many assignments users have submitted, but also the resource consumption required for grading these assignments. Combined with user behavior intensity characteristics, the final intensity of educational demand is generated, making the analysis results more reliable. Furthermore, the obtained intensity of educational demand can provide a precise basis for resource scheduling, enabling the platform to expand resources in advance during peak periods and rationally shrink resources during low loads. This avoids delay accumulation and reduces unnecessary waste, thereby achieving a dynamic balance between supply and demand.

[0063] Specifically, such as Figure 3 As shown, the specific steps for analyzing the content resource mapping feature set of each homework image of the set education platform are as follows: In the input layer of the homework content parsing model, the data of each homework image of the set education platform is received and preprocessed. Specifically, the homework image is normalized in size and noise is removed after normalization. Median filtering and Gaussian filtering can be used for noise processing.

[0064] In the image perception layer of the homework content analysis model, based on the preprocessed homework image data of each homework image from the designated education platform, the pixel set and text vector of the corresponding homework image are extracted. Specifically, text region detection and recognition are performed on each homework image. The convolutional neural network detection method stored in the database can be called (i.e., based on the convolutional neural network, local features of the homework image are extracted layer by layer, including edges, corners and texture distribution, and a multi-scale feature map is generated. In the feature map, a Region Proposal Network (RPN) or Connectivity Component Analysis method is used to generate several candidate text regions. Each candidate region contains possible text information. Non-maximum suppression (NMS) is performed on the candidate text regions to remove redundant candidate boxes and retain text regions with high confidence. Boundary regression is further performed on the retained text regions to adjust the rectangles to precise regions aligned with the text edges to obtain the text regions).

[0065] After region detection and recognition processing, each task image undergoes text extraction processing. The Optical Character Recognition (OCR) network stored in the database can be called. The segmented text regions are sequentially input into the OCR network. The OCR includes convolutional layers, recurrent neural network layers, and CTC (Connectionist Temporal Classification) decoding layers. Spatial features of the text image are extracted in the convolutional layers. The contextual dependencies of the text sequence are learned in the recurrent neural network layers. The feature mapping sequence is transformed into the initial character sequence in the CTC decoding layer. At the same time, the recognition confidence of the initial character sequence is output. The initial character sequence is processed by word embedding and mapped to a multi-dimensional semantic vector space with the recognition confidence to form a text vector. The pixel value and two-dimensional coordinates of each pixel in each task image are marked as the pixel set of the corresponding task image.

[0066] In the content encoding layer of the homework content parsing model, based on the pixel set and text vector of each homework image on the educational platform, the content resource requirement feature vector of the corresponding homework image is extracted. Specifically, for any homework image, the pixel value of each pixel is processed into grayscale to obtain the grayscale pixel value of each pixel, and the homework image is divided into several local windows of fixed size (e.g., 8×8 or 16×16 pixel blocks); the grayscale pixel variance of all pixels in each local window is calculated, and edge detection (e.g., Sobel operator or Canny operator) is performed on the homework image to obtain the edge pixel set (including several edge pixels). Based on the edge pixel set, a pixel connectivity analysis method (e.g., 8-neighborhood search) is used to connect adjacent and continuous edge pixels into several edge chains.

[0067] In each edge chain, the continuity of the edge chain is detected. Specifically, when the distance between two adjacent edge pixels (which can be obtained based on the Euclidean distance formula combined with the two-dimensional coordinates of the two adjacent edge pixels) exceeds a set threshold (e.g., 2 pixels), it is considered a break. The next adjacent edge pixel is marked as the break point and used as the starting point of the edge sub-chain of that edge chain. This process continues until a new break point appears in the edge chain, thus dividing the entire edge chain into several edge sub-chains. The distance between two adjacent edge pixels at the break point is taken as the break length. The total break length of all edge chains and the total edge chain length (i.e., the length of each edge chain) are calculated. The sum of the distances between adjacent edge pixels in the edge chain is calculated and the ratio is processed to obtain the edge breakage rate. The recognition confidence in the text vector is read and weighted with the grayscale pixel variance and the edge breakage rate. The edge breakage rate in the weighted processing is expressed as (1-edge breakage rate) to extract the clear reading features of each task image. This is used to characterize the overall decodeability level of the task image, that is, the degree to which the task image is accurately parsed in the automatic recognition and subsequent grading process. The higher the feature value, the easier it is for the task content to be automatically parsed and recognized, and the lower the computing power and storage resources required by the education platform.

[0068] The text vector is counted character by character to obtain the total number of characters in the text, which is taken as the total word count. Furthermore, the text vector is segmented into sentences based on punctuation marks (such as periods, question marks, and exclamation marks) and sequence labeling rules (i.e., punctuation constraints: when a period, question mark, or exclamation mark is detected, the current position is automatically marked as the sentence end boundary; part-of-speech constraints: when particles, pronouns, or other parts of speech that are likely to cause sentence breakage appear consecutively, candidate markers for sentence boundaries are added at that position). The number of characters in each sentence is counted, and the average number of characters in all sentences is calculated as the average sentence length.

[0069] The characters in the sentence are segmented based on a pre-defined word segmentation dictionary and statistical language model (such as a word segmentation model based on conditional random fields or a word segmentation method based on deep neural networks) to obtain independent word sequences in the sentence. These sequences are then input into a dependency parser (such as a dependency parsing tool based on neural networks). Each word in the sentence is assigned a dependency relation label (such as subject-verb, verb-object, or attributive-head relation), and directional connections are established between the words. (That is, when the dependency relationship between two words is determined, the subordinate word is taken as the child node and the governing word as the parent node, and a one-way connection is established between them, with the connection pointing from the subordinate word to the governing word.) This is used to represent the subject-subordinate relationship in the grammatical structure of a sentence. For example, in the sentence "students writing an essay," the dependency relationship between "students" and "writing" is a subject-predicate relationship, so a connection is established from "students" to "writing." The core predicate in the sentence (i.e., the main verb that occupies the predicate position in the syntactic structure and serves as the center of control for other words, which is neither a subordinate node of other verbs nor a controlling node for multiple nouns or phrases, such as "writing" in "students writing an essay") is used as the root node of the dependency syntax tree. All words that have a direct or indirect dependency relationship with it are added layer by layer according to the directional connection, and other words in the sentence are marked as leaf nodes to form a complete dependency syntax tree.

[0070] A depth-first traversal method is used, starting from the root node and proceeding sequentially to all leaf nodes. The number of nodes in each path is recorded, and the path from the root node to the deepest leaf node (i.e., the word with the longest path length in the dependency syntax tree that does not connect to any other leaf nodes; the path length refers to the total number of nodes traversed from the root node through several intermediate leaf nodes to reach a certain leaf node) is used as the syntactic depth value of the sentence. (For example, in the sentence, "A student wrote an essay: Path 1 is writing, learning...") If a sentence has a length of 2 and a path of 2: write, composition, one article, with a length of 3, then the maximum path length is 3, and the syntactic depth of the sentence is 3. The syntactic depth of all sentences is then averaged to obtain the overall syntactic depth. The total number of words, average sentence length, and overall syntactic depth are standardized, and the results are weighted to extract the grading load feature. This feature is used to characterize the intensity of the demand on the educational platform's processing resources during the automatic grading of assignment texts. The higher the feature value, the longer the composition and the higher the syntactic complexity, and the more processing nodes and memory space the platform needs to allocate.

[0071] Read the word sequence in each sentence and compare it with the writing vocabulary database stored in the database. The writing vocabulary database includes a basic vocabulary subset and an advanced vocabulary subset. Calculate the total coverage of all sentences in the basic vocabulary subset and the advanced vocabulary subset respectively, and then compare it with the total number of words (i.e., the total number of words in all sentences) to obtain the coverage rate.

[0072] Read the dependency syntactic tree in the form of each sentence, identify the sentence pattern type with a preset dependency relation pattern library, and assign corresponding labels. For example, if the syntactic tree contains only one subject-predicate structure and no clause attachment, it is determined as a simple sentence; if the syntactic tree contains multiple subject-predicate structures and there is a situation where a clause depends on the main clause, it is determined as a complex sentence; if a passive marker word (such as "被", "由……所") appears in the syntactic tree, it is determined as a passive sentence; if there are multiple parallel relations in the syntactic tree (such as the subjects or predicates connected by "和", "以及"), it is determined as a compound sentence; count the probability value of each label appearance (such as counting the total number of all simple sentences and performing a ratio process with the total number of sentences), and based on the entropy formula, analyze the probability value of each label appearance to extract the expression complex features, and perform a weighted process with the coverage rate to extract the expression structure breadth features, which are used to characterize the overall expression expansion level of the composition, reflecting the breadth and complexity of the composition in vocabulary usage and sentence pattern organization. The higher this feature value is, the higher the breadth and complexity of the composition in vocabulary and sentence patterns, and the platform needs to call more semantic understanding modules and knowledge bases, increasing the complexity of resource scheduling; and splice the homework clear judgment feature, marking load feature, and expression structure breadth feature into a content resource demand feature vector;

[0073] In the output layer of the homework content parsing model, based on the content resource demand feature vector in each homework image of the set education platform, output the content resource mapping feature set corresponding to each homework image. Specifically: Activate the homework clear judgment feature, marking load feature, and expression structure breadth feature in the content resource demand feature vector through the Sigmoid function respectively to obtain the homework clear judgment feature value, marking load feature value, and expression structure breadth feature value with the result between 0 and 1.

[0074] Among them, the pre-training steps of the homework content parsing model are as follows:

[0075] Obtain an annotated data set, which consists of several homework images and corresponding label data. These label data are annotated by education experts based on actual homework. Each group of samples in the annotated data set includes the pixel values of each homework image, text region information (including characters, words, etc.), and its corresponding true label (such as the true value of the homework clear judgment feature, the true value of the marking load feature, the true value of the expression structure breadth feature, etc.). Each sample in the data set has been manually annotated and preprocessed. The preprocessing steps include operations such as size normalization, noise removal, and grayscale conversion to ensure that the input data can adapt to subsequent deep learning training, and divide the preprocessed annotated data set into a training set, a validation set, and a test set. For example, 80% of the data is used for training, 10% for validation, and 10% for testing.

[0076] The task content parsing model is trained using an image perception layer and a content encoding layer as core components. The training steps are as follows: The task image is input into the model and preprocessed, including grayscale processing, size normalization, and noise removal. Low-level features of the task image, such as edges, corners, and textures, are extracted using a convolutional neural network (CNN). A region proposal network (RPN) is used to detect text regions, and non-maximum suppression (NMS) is used to remove redundant regions. Optical character recognition (OCR) is performed on the segmented text regions. Spatial features are extracted through convolutional layers, and contextual dependencies are captured through a recurrent neural network (RNN). An initial character sequence is generated through a CTC decoding layer. Finally, the text vector and recognition confidence are output. Syntactic analysis is performed on the text vector to extract grading load features, expressive structure breadth features, etc., and comprehensive feature extraction is performed by combining text information with pixel information of the task image.

[0077] Training is performed using optimization algorithms (such as the Adam optimizer) to minimize the loss function (such as cross-entropy loss or mean squared error). The model weights are adjusted using the backpropagation algorithm (BPTT). During training, different features are weighted to ensure that the model can comprehensively consider the resource requirements of multiple dimensions such as the clarity of the assignment images, the complexity of grading, and the breadth of the expressive structure.

[0078] The model's performance is evaluated on the validation set, and hyperparameters (such as learning rate, number of hidden layer units, etc.) are adjusted to improve training effectiveness. By continuously optimizing the model, training error is reduced, and the generalization ability of the model is evaluated using the test set to ensure that the trained model can handle unseen assignment images and generate effective educational content requirements features.

[0079] In this implementation, a pre-trained assignment content parsing model is introduced to perform deep analysis of the assignment image dataset, thereby ensuring the accuracy of subsequent grading and resource allocation. The preprocessing stage of the input layer, through size normalization and filtering noise reduction, effectively reduces image resolution differences and noise interference, making subsequent text detection and recognition clearer and more stable. Then, in the image perception layer, the introduction of text region detection and OCR network enables the platform to accurately extract the student's writing content and convert it into semantic vectors. This not only recognizes the text but also lays the foundation for subsequent grammar, vocabulary, and sentence structure analysis. The content encoding layer combines the pixel features of the image with the text features, enabling the image to truly reflect the length, complexity, and expression level of the essay content. The output layer uniformly forms a content resource mapping feature set, so that each assignment image can be converted into comparable feature values. This allows the platform to assess the difficulty and requirements of individual assignments and to allocate resources reasonably as a whole, thereby ensuring that the grading process is both efficient and accurate.

[0080] Specifically, the steps for microservice orchestration resource scheduling of a designated education platform based on educational resource adaptation feature values ​​are as follows: The educational resource adaptation feature values ​​of the designated education platform are compared with a preset educational resource adaptation feature range. Based on the comparison result, corresponding microservice orchestration resource scheduling measures are taken for the designated education platform. Specifically, if the educational resource adaptation feature value of the designated education platform is lower than the lower limit of the preset educational resource adaptation feature range, it indicates that the platform's resource supply is insufficient to effectively meet the current demand. In this case, the load adaptation feature values ​​of each service node of the designated education platform are sorted in descending order to generate a priority expansion table. Service nodes in the priority expansion table with load adaptation feature values ​​higher than the preset value are then expanded in terms of execution instances to quickly increase the effective computing power supply, such as increasing the number of microservice replicas and CP of the node. If the educational resource adaptation feature value is still lower than the lower limit after the above measures, the container orchestration engine (such as Kubernetes) will be triggered to create more service node instances in the cluster to improve the overall supply capacity of the education platform. It should be noted that the preset educational resource adaptation feature range can be set by the following steps: obtain the educational resource adaptation feature values ​​of the set education platform for several historical periods, and extract the mean value and standard deviation value of the educational resource adaptation features. The upper limit of the educational resource adaptation feature range is the multiple of the mean value and the standard deviation value of the educational resource adaptation features (which can be between 1.0 and 2.0). The lower limit of the educational resource adaptation feature range is the multiple of the mean value and the standard deviation value of the educational resource adaptation features (which can be between 1.0 and 2.0).

[0081] If the educational resource adaptation characteristic value of the educational platform is within the preset educational resource adaptation characteristic range, it indicates that the resource supply and demand are in balance, and there is no need to expand or shrink; the existing resource configuration can be maintained. If the educational resource adaptation characteristic value of the educational platform is higher than the upper limit of the preset educational resource adaptation characteristic range, it indicates that the platform has an oversupply of resources, that is, the platform's resource allocation exceeds the current demand. In this case, the load adaptation characteristic values ​​of each service node of the educational platform are sorted in ascending order to generate a priority scaling-down table. Service nodes in the priority scaling-down table with load adaptation characteristic values ​​lower than the preset load adaptation characteristic value are then scaled down in sequence. This reduces the number of replicas or resource quotas of the node, and the request tasks of low-load nodes are migrated to at least a few high-efficiency nodes for centralized processing. Idle nodes are then released or shut down through microservice orchestration. If there is still an oversupply after node-level scaling-down, the overall number of microservice instances in the cluster is reduced through the container orchestration engine to reduce energy consumption and operational overhead.

[0082] In this implementation plan, corresponding microservice orchestration resource scheduling measures are triggered based on the educational resource adaptation feature values ​​to achieve dynamic linkage between supply and demand status and resource allocation. When resource insufficiency is detected, the system will first sort the nodes according to their load adaptation feature values ​​and prioritize scaling up the nodes that are under the most pressure. This can accurately improve computing power supply and avoid waste caused by blindly scaling up the entire system. If local scaling up still cannot meet the demand, the orchestration engine will start up new nodes to enhance the overall carrying capacity of the platform, thereby enabling the platform to respond quickly during peak periods and avoid grading delays or request backlogs. Conversely, when resources are surplus, low-load nodes will be scaled down first, and tasks will be migrated to more efficient nodes for centralized processing. Then, redundant instances will be gradually released, ultimately reducing the overall number of replicas at the cluster level. This maintains the high efficiency of the platform's operation and enables the educational platform to maintain supply and demand balance and maximize resource utilization under different load conditions.

[0083] Please see Figure 4 This invention provides a technical solution: an educational elastic resource scheduling system based on microservice orchestration, comprising: an educational resource data acquisition module, used to acquire educational resource data of a designated educational platform within a set period, including educational support resource data and user behavior data; a resource elasticity feature analysis module, used to analyze the resource elasticity feature set of the designated educational platform based on the educational resource data of the designated educational platform, including educational resource supply feature values ​​and educational demand intensity feature values; an educational resource adaptation analysis module, used to analyze the educational resource adaptation feature values ​​of the designated educational platform based on the resource elasticity feature set of the designated educational platform; and a microservice orchestration scheduling module, used to perform microservice orchestration resource scheduling processing on the designated educational platform based on the educational resource adaptation feature values.

[0084] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0085] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for scheduling elastic educational resources based on microservice orchestration, characterized in that, Includes the following steps: Within a set period, acquire educational resource data from a designated educational platform, including educational support resource data and user behavior data. The educational support resource data includes the response latency value, processing capacity value, microservice call chain latency fluctuation rate value, energy consumption conversion efficiency value, operating temperature drift rate value, cache hit rate value, and service anomaly rate value for each service node. The user behavior data includes a homework image dataset, homework submission rate value, user interaction activity factor, activity triggering peak factor, and differentiated demand factor. The homework image dataset includes several homework image data. Based on the educational resource data of the designated education platform, the resource elasticity characteristic set of the designated education platform is analyzed, including the characteristic values ​​of educational resource supply and the characteristic values ​​of educational demand intensity, which are as follows: Based on the educational support resource data of the established education platform, the resource carrying capacity service sequence feature set of the established education platform is analyzed, including the resource carrying capacity feature value and resource efficiency feature value of each service node; Based on the feature set of the resource availability service sequence of the education platform, the characteristic values ​​of the education resource supply of the education platform are analyzed. Based on a pre-trained homework content parsing model and combined with the homework image dataset of the designated education platform, the characteristic values ​​of the intelligent grading resource requirements of the designated education platform are analyzed, specifically: Each homework image data from the educational platform is input into a pre-trained homework content parsing model to analyze the content resource mapping feature set in the corresponding homework image, including homework clarity feature value, grading load feature value, and expression structure breadth feature value. Based on the content resource mapping feature set in each homework image of the setting education platform, the intelligent grading resource demand feature value of the setting education platform is analyzed. Based on the homework submission rate, user interaction activity factor, activity triggering peak factor, and differentiated demand factor of the education platform, we analyze the intensity characteristics of user behavior on the education platform, and combine them with the characteristic characteristics of intelligent grading resource demand to analyze the intensity characteristics of educational demand on the education platform. Based on the resource elasticity feature set of the education platform, the resource adaptation feature value of the education platform is analyzed. Based on the educational resource adaptation characteristics, microservice orchestration and resource scheduling are performed on the designated educational platform.

2. The educational elastic resource scheduling method based on microservice orchestration according to claim 1, characterized in that, The specific steps for analyzing and defining the feature set of the resource-enabled service sequence of the education platform are as follows: Based on the response latency, processing capacity, and microservice call chain latency fluctuation of each service node in the education platform, the resource carrying capacity characteristics of the corresponding service nodes are analyzed. Based on the energy conversion efficiency, operating temperature drift rate, cache hit rate, and service anomaly rate of each service node in the educational platform, the resource efficiency characteristics of the corresponding service nodes are analyzed.

3. The educational elastic resource scheduling method based on microservice orchestration according to claim 1, characterized in that, The task image data includes the pixel value and two-dimensional coordinates of each pixel in the task image, and the task content parsing model includes an input layer, an image perception layer, a content encoding layer, and an output layer.

4. The educational elastic resource scheduling method based on microservice orchestration according to claim 3, characterized in that, The specific steps for analyzing and defining the content resource mapping feature set for each assignment image on the education platform are as follows: In the input layer of the homework content parsing model, each homework image data from the designated education platform is received and preprocessed. In the image perception layer of the homework content parsing model, based on the preprocessed homework image data of each homework image from the designated education platform, the pixel set and text vector of the corresponding homework image are extracted. In the content encoding layer of the homework content parsing model, based on the pixel set and text vector in each homework image of the set education platform, the content resource requirement feature vector in the corresponding homework image is extracted. In the output layer of the homework content parsing model, based on the content resource requirement feature vector in each homework image of the set education platform, the content resource mapping feature set of the corresponding homework image is output.

5. The educational elastic resource scheduling method based on microservice orchestration according to claim 1, characterized in that, The specific steps for calculating the educational resource adaptation feature values ​​of the educational platform are as follows: ； in, To set the educational resource adaptation feature values ​​for the education platform, To define the characteristic value of educational demand intensity for the education platform, To set the characteristic values ​​of educational resource supply for the education platform, The smoothing adjustment coefficients are stored in the database. These are the difference adjustment coefficients stored in the database.

6. The educational elastic resource scheduling method based on microservice orchestration according to claim 1, characterized in that, The specific steps for microservice orchestration and resource scheduling of a designated education platform based on educational resource adaptation feature values ​​are as follows: The system will compare the educational resource adaptation feature values ​​of the educational platform with the preset educational resource adaptation feature ranges. Based on the judgment and processing results, corresponding microservice orchestration and resource scheduling measures are taken for the designated education platform.

7. A microservice orchestration-based educational elastic resource scheduling system, used to implement the microservice orchestration-based educational elastic resource scheduling method as described in any one of claims 1-6, characterized in that, include: The education resource data collection module is used to acquire education resource data from a designated education platform within a set period, including education support resource data and user behavior data. The resource elasticity characteristic analysis module is used to analyze the resource elasticity characteristic set of the designated education platform based on the education resource data of the designated education platform, including the characteristic values ​​of education resource supply and the characteristic values ​​of education demand intensity. The educational resource adaptation analysis module is used to analyze the educational resource adaptation feature values ​​of a set of educational platforms based on the resource elasticity feature set of the set educational platforms. The microservice orchestration and scheduling module is used to perform microservice orchestration and resource scheduling on a designated education platform based on educational resource adaptation characteristics.

Images