Adaptive expansion method and system for SaaS applications based on industrial Internet

Abstract

The present invention belongs to the field of industrial Internet technology, and in particular to an adaptive expansion method and system for SaaS applications based on the industrial Internet. The method realizes intelligent resource scheduling and adaptive expansion of the SaaS application platform when processing user uploaded services by integrating demand prediction, remaining resource monitoring prediction and distributed enhanced resource scheduling model. First, the resource demand within the business cycle is predicted, and the remaining resource status is monitored. Secondly, based on resource demand prediction, remaining completion time and resource distribution changes, distributed enhanced scheduling is used to optimize resource allocation and supplement missing resources in order to minimize total cost. Finally, through continuous testing and evaluation, resource allocation parameters are adjusted to ensure efficient and cost-optimized business processing. The present invention realizes adaptive expansion of SaaS application resources through dynamic monitoring and intelligent scheduling, improves the response speed and resource utilization of the system, and reduces operation and maintenance costs.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial Internet, and in particular to an adaptive expansion method and system for SaaS applications based on industrial Internet. Background Art

[0002] With the development of industrial Internet technology, more and more companies are beginning to adopt the software as a service (SaaS) model to deploy and manage various application systems. Especially in the fields of warehouse management system (SWMS), manufacturing execution system (SMES), and supply chain management (SERP), the SaaS model is widely welcomed for its flexibility, scalability, and low initial investment cost. However, the existing industrial Internet SaaS applications mostly use static resource configuration, resulting in slow response speed and low resource utilization, and most of the existing user needs show seasonal demand changes. For example, the warehouse management corresponding to the warehouse management system in different seasons is very different, which makes its corresponding resource requirements different in different time periods. If only static resource configuration is used, it will lead to insufficient or waste of resources, increase unnecessary expenses, and reduce the quality of service.

[0003] For example, a Chinese patent with authorization announcement number CN112116417B discloses a SaaS model business extensible configuration method, which receives a business extensible configuration request; the extensible configuration request includes an extensible configuration type identifier, an extensible configuration item identifier, and a user identifier; according to the extensible configuration type identifier, a corresponding operation policy is determined; according to the user identifier and the determined operation policy, the extensible configuration item corresponding to the extensible configuration item identifier is processed.

[0004] For example, a Chinese patent application with publication number CN115437807A discloses a method, device and electronic device for calling a SAAS extended application. The method includes: obtaining an application request from a client, determining a target SAAS extended application corresponding to the application request; determining a remote service type corresponding to the target SAAS extended application; based on the remote service type, constructing an extended service reference; and remotely calling the target SAAS extended application through the extended service reference.

[0005] The above existing technologies have the following problems: the existing technologies mostly adopt static resource configuration, resulting in slow response speed and low resource utilization, and most of the existing user demands show seasonal demand changes, which makes the corresponding resource demands different in different time periods. If only static resource configuration is adopted, it will lead to insufficient or waste of resources. For this reason, the present invention provides an adaptive expansion method and system for SaaS applications based on the Industrial Internet. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention proposes an adaptive expansion method and system for SaaS applications based on the industrial Internet. The method realizes intelligent resource scheduling and adaptive expansion of the SaaS application platform when processing user uploaded services by integrating demand prediction, remaining resource monitoring prediction and distributed enhanced resource scheduling model. First, it predicts the resource demand within the business cycle and monitors the remaining resource status. Secondly, based on the resource demand prediction, remaining completion time and resource distribution changes, it optimizes resource allocation and supplements missing resources through distributed enhanced scheduling to minimize the total cost. Finally, through continuous testing and evaluation, the resource allocation parameters are adjusted to ensure efficient and cost-optimized business processing. The present invention realizes adaptive expansion of SaaS application resources through dynamic monitoring and intelligent scheduling, improves the system's response speed and resource utilization, and reduces operation and maintenance costs.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The adaptive expansion method of SaaS applications based on the Industrial Internet includes:

[0009] S1. Receive the corresponding type of processing business uploaded by the user, and obtain the resource demand of the user's current type of processing business in each business cycle according to the corresponding type of processing business uploaded by the user and the corresponding historical processing business data through the configured demand prediction model;

[0010] S2. Configure a remaining resource monitoring and prediction model on the SaaS application platform, and obtain the remaining completion time and distribution state of the remaining resources for the current business processing according to the current usage status data of the SaaS application platform and the status data in the corresponding historical business cycle through the configured remaining resource monitoring and prediction model;

[0011] S3, according to the resource demand in each service cycle of the user's current type of service processing, the remaining completion time of the current service processing, and the distribution state of the remaining resources, the remaining resources are scheduled through the configured distributed enhanced resource scheduling model, and the missing resource amount is obtained;

[0012] S4. Perform cloud service resource supplementation scheduling based on the acquired missing resource amount through the distributed enhanced resource scheduling model;

[0013] S5. Set the evaluation score threshold, repeat S1 to S4 to perform business processing tests, and obtain the resource allocation, completion and cost data of the corresponding business. Evaluate the test process through the configured evaluation algorithm. When the evaluation result is greater than the evaluation score threshold, keep the current resource allocation parameters unchanged. Otherwise, feed back the evaluation score to the distributed enhanced resource scheduling model to adjust the current resource allocation parameters.

[0014] Specifically, the step of obtaining the resource demand of the user's current type of processing service in each service cycle includes:

[0015] S101, obtaining business processing status data corresponding to N processing cycles of M historical application businesses, and preprocessing the obtained data;

[0016] The business processing status data includes business type, business scale, business request timestamp, set business completion timestamp, actual business completion timestamp and business resource demand;

[0017] S102, input the pre-processed business processing status data into the clustering algorithm, and obtain (ng, gk) corresponding to each processing cycle in N processing cycles, where ng represents the number of application services in the g-th clustering space in the n-th processing cycle, and gk represents the business processing status data sequence corresponding to the k-th application service in the g-th clustering space in the n-th processing cycle, where g∈G, k∈K, where G represents the total number of clustering spaces, and K represents the number of application services in each clustering space;

[0018] S103. Construct a demand forecasting model containing G integrated sub-models based on support vector machine and integration algorithm, input (ng, gk) corresponding to the G clustering spaces in the nth processing cycle into each integrated sub-model in the demand forecasting model for training, and output the predicted amount of business processing resource demand for K application businesses in the corresponding clustering space.

[0019] Specifically, the step of obtaining the resource demand of the user's current type of processing service in each service cycle also includes:

[0020] S104, using the predicted business processing resource requirements of K application business types in the output corresponding cluster space and the corresponding actual business processing resource requirements, obtain the sum of the processing resource requirements quantum errors ek corresponding to the K application businesses in the g-th cluster space in the n-th processing cycle The sum of resource demand errors corresponding to G clustering spaces eg,k

[0021]

[0022] S105: Set the sub-error sum threshold and the resource demand error sum threshold. is less than the sum threshold of sub-errors and When the error sum of resource demand is less than the threshold, the training is completed, and the trained demand forecasting model is obtained;

[0023] S106, if or If the corresponding threshold is not met, or Feedback to the demand forecasting model for training until the threshold condition is met, and a trained demand forecasting model is obtained;

[0024] S107, repeating the process of S103-S106, inputting the business processing status data corresponding to N processing cycles into the demand forecasting model for training, and obtaining N trained demand forecasting models;

[0025] S108, obtaining the real-time business processing status data corresponding to the current upload processing business, and judging the processing cycle to which the current upload processing business belongs through the business request timestamp in the real-time business processing status data, and calling the demand prediction model of the corresponding processing cycle according to the processing cycle to which it belongs;

[0026] S109: input the real-time business processing status data corresponding to the current upload processing business into the called demand prediction model to obtain the resource demand in the processing cycle to which the current upload processing business belongs.

[0027] Specifically, the step of obtaining the remaining completion time and the distribution state of the remaining resources of the currently processed business includes:

[0028] S201, obtaining a distribution state quantity sequence of historical resource requirements of all application services at each time point in N processing cycles, a corresponding distribution state quantity sequence of remaining resources, and a difference sequence between the expected service processing time length and the actual service processing time length of each application service in the corresponding cluster space in each processing cycle;

[0029] S202, constructing a remaining resource monitoring prediction model based on a random forest algorithm, inputting the sequence obtained in S201 into the remaining resource monitoring prediction model for training, and obtaining a trained remaining resource monitoring prediction model;

[0030] S203, obtain the remaining processing resources at the current time point, the resource demand in the current processing cycle of the current upload processing business, and the expected business processing time length data of each business, and input the obtained data into the trained remaining resource monitoring and prediction model to predict the remaining completion time of each application business currently being processed and the remaining resource distribution state corresponding to each time point in the corresponding processing cycle.

[0031] Specifically, the distributed enhanced resource scheduling model includes a central resource scheduling sub-model and I edge scheduling sub-models, and I edge scheduling sub-models correspond to I resource servers; the steps of constructing the distributed enhanced resource scheduling model include:

[0032] S301: Based on the resource demand in the current processing cycle of the current upload processing business, the remaining completion time of each business currently being processed, the remaining resource distribution state corresponding to each time point in the processing cycle, and the remaining schedulable resources of each resource server, the current input state is constructed:

[0033]

[0034] in It represents the resource demand of the k-th application service at time t in the n-th processing cycle, and It represents the remaining resources of I resource servers at the current time t. Indicates the calling status of the i-th resource server at time t. When , it means that the i-th resource server has not been called at the current time t; when It means that the i-th resource server is in the calling state at time t. represents the remaining schedulable resources of the i-th resource server at time t, Indicates the remaining completion time of the k-th application being processed in the n-th processing cycle. Indicates the processing priority of the kth application service at time t in the nth processing cycle. represents the cluster similarity space matrix corresponding to K application services in the g-th cluster space in the n-th processing cycle; Indicates the price of a single resource server called in the nth processing cycle. represents the price of the jth uncalled resource server in the nth processing cycle, represents the probability that the i-th resource server in the calling state fails in the n-th processing cycle, Indicates the amount of missing resources corresponding to the k-th application service at the current time t in the n-th processing cycle.

[0035] Specifically, the steps of constructing the distributed enhanced resource scheduling model also include:

[0036] S302, setting the execution action trigger information corresponding to the current moment according to the current input state quantity, including:

[0037] S3021. Setting a cluster matching threshold. When the remaining resources of the i-th resource server at the current time t meet the resource requirements of the application business k and the cluster matching value between the cluster similarity space matrix of the business being processed in the i-th resource server and the business processing status data of the application business k is greater than the cluster matching threshold, triggering the first execution action information, that is, uploading the application business k to the i-th resource server for processing;

[0038] S3022. When the remaining resources of the i-th resource server at the current time t do not meet the resource requirements of the application business k but meet the cluster matching value greater than the cluster matching threshold, the second execution action information is triggered, that is, the remaining resource distribution state of each scheduled resource server within the expected business processing time length of the application business k is monitored through the remaining resource monitoring prediction model, and the remaining resources of each scheduled resource server at each time are scheduled to the i-th resource server;

[0039] S3023. If the resources after scheduling do not meet the resource requirements of the application service k at the current moment, trigger the third execution action information;

[0040] That is, the jth uncalled resource server with the lowest price at the current moment corresponding to the jth edge scheduling submodel is scheduled through the central resource scheduling submodel to supplement resources. At the same time, the remaining completion time of the business being processed in the current resource server is monitored through the remaining resource monitoring and prediction model. When the remaining completion time of the business being processed is less than the expected business processing time length corresponding to the application business k, and the resources released after the business being processed is completed meet the resource demand of the application business k, the timestamp of the corresponding completion time of the processed business is obtained through the remaining resource monitoring and prediction model, and the scheduling of the jth resource server is disconnected through the central resource scheduling submodel at the corresponding completion time of the business. If the resources released after the business being processed is completed do not meet the resource demand of the application business k, the jth resource server is not disconnected.

[0041] Specifically, the steps of constructing the distributed enhanced resource scheduling model also include:

[0042] S3024, setting a resource matching threshold, when the remaining resources of the i-th resource server at the current time t meet the resource requirements of the application business k, but the clustering similarity space matrix of the business being processed in all resource servers and the clustering matching value of the business processing status data of the application business k are less than or equal to the clustering matching threshold, triggering the fourth execution action information;

[0043] That is, the remaining resource distribution state quantity of each scheduled resource server within the expected service processing time length of the application service k is monitored through the remaining resource monitoring prediction model, and the obtained remaining resource distribution state quantity is matched with the resource demand of the application service k. If there is a resource server with a matching score greater than the resource matching threshold, the application service k is uploaded to the resource server corresponding to the maximum matching score for processing. If the matching scores are all less than or equal to the resource matching threshold, the application service k is uploaded to the called resource server with the lowest price within the expected service processing time length of the application service k for processing;

[0044] S3025. When the remaining resources of all resource servers at the current time t do not meet the resource requirements of the application business k and the clustering similarity space matrix of the business being processed in all resource servers and the clustering matching value of the business processing status data of the application business k are both less than or equal to the clustering matching threshold, the fifth execution action information is triggered, that is, the third execution action information is directly triggered, the resource server that has not been called is directly scheduled to supplement the resources, and the remaining resources of all resource servers at each moment are monitored, and it is determined whether to disconnect the resource server scheduled for supplementation according to the monitoring results;

[0045] Specifically, the steps of constructing the distributed enhanced resource scheduling model also include:

[0046] S3026. When the number of currently applied services exceeds one, the sixth execution action information is triggered; that is, the clustering algorithm is used to determine whether the applied services belong to the same clustering space. If they belong to the same clustering space, all the applied services are placed in the same service processing queue. When the remaining resources at the current moment meet the resource requirements of all applied services in the service processing queue, the first execution action information or the fourth execution action information is triggered to process the applied services in the service processing queue.

[0047] S3027, when the remaining resources at the current moment do not meet the resource requirements of all the application services in the service processing queue, trigger the seventh execution action information; that is, directly trigger the second execution action information or the third execution action information to process the application services in the service processing queue;

[0048] S3028. If they do not belong to the same cluster space, trigger the eighth execution action information; that is, put all the application services belonging to the same cluster space in the same application service into the same service processing queue, and mark each queue with the average priority of all the application services in the same queue, and arrange the marked queues in descending order. When the remaining resources at the current moment meet the needs of all service processing queues, allocate resources according to the descending order of the queues.

[0049] S3029, when the remaining resources at the current moment do not satisfy all business processing queues, the remaining resources at the current moment are allocated to the queues arranged in descending order, and the missing resource amount corresponding to the remaining unobtained resource queues is obtained, and the missing resource amount and the missing resource amount are calculated according to the missing resource amount and The resource server is scheduled to be supplemented so that the scheduling cost is minimized while meeting the resource demand, and the S3023 process is repeated. The distribution state of the remaining resources is monitored through the remaining resource monitoring and prediction model, and the resource server for supplementary scheduling is disconnected at the timestamp node that meets the conditions.

[0050] The adaptive expansion system of SaaS applications based on the industrial Internet includes: demand forecasting module, resource monitoring module, resource scheduling module and test feedback module;

[0051] The demand prediction module includes a receiving unit and a prediction unit; the receiving unit is used to receive the corresponding type of processing business uploaded by the user, and pre-process and save the received data;

[0052] The prediction unit is used to obtain the resource demand of the user's current type of processing service in each service cycle through the configured demand prediction model according to the corresponding type of processing service uploaded by the user and the corresponding historical processing service data;

[0053] The resource monitoring module is used to obtain the remaining completion time and the distribution state of the remaining resources of the current processing business based on the current usage status data of the SaaS application platform and the status data in the corresponding historical business cycle through the configured remaining resource monitoring prediction model;

[0054] The resource scheduling module includes a resource scheduling unit and an expansion unit;

[0055] A resource scheduling unit is used to schedule the remaining resources using a distributed enhanced resource scheduling model according to the resource demand, the remaining completion time, and the change in the remaining resource distribution state, and obtain the current missing resource amount;

[0056] The expansion unit is used to supplement and schedule cloud service resources according to the amount of missing resources;

[0057] The test feedback module includes a test evaluation unit; the test evaluation unit is used to obtain an evaluation score in the corresponding test process through a configured evaluation algorithm based on the resource allocation, completion and cost data of the corresponding type of business obtained during the test process.

[0058] A computer-readable storage medium stores computer instructions, which, when executed, execute an adaptive expansion method for SaaS applications based on the industrial Internet.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] In view of the deficiencies in the prior art, the present invention dynamically predicts the resource demand of the user's current type of processing business within the business cycle, and combines the current usage status data and historical status data of the SaaS application platform to predict the distribution state of the remaining resources, and uses a distributed enhanced resource scheduling model to intelligently schedule the remaining resources, thereby effectively solving the problems of slow response speed and low resource utilization caused by static resource configuration in the prior art. Secondly, by analyzing the changes in resource demand of the corresponding type of business in each business cycle, the resources in the corresponding cycle can be dynamically expanded to avoid insufficient or wasted resources. Finally, by obtaining the cost and completion time point of each type of business, and supplementing and scheduling cloud service resources according to the missing resources, it is ensured that the total cost is minimized while meeting the current business processing volume. The present invention not only improves the response speed and resource utilization of the system, but also reduces the operating cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of an adaptive expansion method for SaaS applications based on the industrial Internet according to Embodiment 1 of the present invention;

[0062] Figure 2 This is a module diagram of the adaptive expansion system of SaaS applications based on the industrial Internet in Example 2 of the present invention. DETAILED DESCRIPTION

[0063] Example 1

[0064] See also Figure 1 , an embodiment provided by the present invention: an adaptive expansion method of SaaS application based on industrial Internet, comprising the following steps:

[0065] S1. Receive the corresponding type of processing business uploaded by the user, and obtain the resource demand of the user's current type of processing business in each business cycle based on the corresponding type of processing business uploaded by the user and the corresponding historical processing business data through the configured demand prediction model;

[0066] Furthermore, in this embodiment, the step of obtaining the resource demand of the user's current type of processing service in each service cycle includes:

[0067] S101, obtaining business processing status data corresponding to N processing cycles of M historical application businesses, and preprocessing the obtained data;

[0068] Furthermore, the step of processing cycle acquisition in this embodiment includes:

[0069] Obtain business application volume data of the SaaS application platform for one year, and obtain the month-on-month growth rate of business applications based on the obtained business application volume data through SPSS software;

[0070] According to the obtained month-on-month growth rate of business applications, the time periods corresponding to the same month-on-month growth rate are divided into a processing cycle;

[0071] Furthermore, in this embodiment, the service processing status data includes service type, service scale, service request timestamp, set service completion timestamp, actual service completion timestamp and service resource demand;

[0072] Furthermore, in this embodiment, the application services include services such as warehouse management system (SWMS), manufacturing execution system (SMES) and supply chain management (SERP);

[0073] S102, input the pre-processed business processing status data into the clustering algorithm, and obtain (ng, gk) corresponding to each processing cycle in N processing cycles, where ng represents the number of application services in the g-th clustering space in the n-th processing cycle, and gk represents the business processing status data sequence corresponding to the k-th application service in the g-th clustering space in the n-th processing cycle, where g∈G, k∈K, where G represents the total number of clustering spaces, and K represents the number of application services in each clustering space;

[0074] S103. Construct a demand forecasting model containing G integrated sub-models based on support vector machine and integration algorithm, input (ng, gk) corresponding to the G clustering spaces in the nth processing cycle into each integrated sub-model in the demand forecasting model for training, and output the predicted amount of business processing resource demand for K application businesses in the corresponding clustering space.

[0075] S104, using the predicted business processing resource requirements of K application business types in the output corresponding cluster space and the corresponding actual business processing resource requirements, obtain the sum of the processing resource requirements quantum errors ek corresponding to the K application businesses in the g-th cluster space in the n-th processing cycle The sum of resource demand errors corresponding to G clustering spaces eg,k

[0076]

[0077] S105: Set the sub-error sum threshold and the resource demand error sum threshold. is less than the sum threshold of sub-errors and When the error sum of resource demand is less than the threshold, the training is completed, and the trained demand forecasting model is obtained;

[0078] S106, if or If the corresponding threshold is not met, or Feedback to the demand forecasting model for training until the threshold condition is met, and a trained demand forecasting model is obtained;

[0079] S107, repeating the process of S103-S106, inputting the business processing status data corresponding to N processing cycles into the demand forecasting model for training, and obtaining N trained demand forecasting models;

[0080] S108, obtaining the real-time business processing status data corresponding to the current upload processing business, and judging the processing cycle to which the current upload processing business belongs through the business request timestamp in the real-time business processing status data, and calling the demand prediction model of the corresponding processing cycle according to the processing cycle to which it belongs;

[0081] S109: input the real-time business processing status data corresponding to the current upload processing business into the called demand prediction model to obtain the resource demand in the processing cycle to which the current upload processing business belongs.

[0082] This process obtains the business processing status data of N processing cycles of historical M application services, pre-processes it and inputs it into the clustering algorithm, which can accurately classify similar business types into the same clustering space, thereby improving the accuracy of demand forecasting; secondly, a demand forecasting model containing G integrated sub-models is constructed through support vector machines and integrated algorithms, and the business processing resource demand in each clustering space is predicted, which can effectively capture the resource demand rules of different business types. By setting the sub-error sum threshold and the resource demand error sum threshold, and constantly adjusting the model until the threshold conditions are met, the accuracy and reliability of the prediction model are ensured. Finally, by analyzing the real-time business processing status data of the current upload processing business, judging its processing cycle, and calling the demand forecasting model of the corresponding processing cycle, the resource demand of the current business type can be predicted in real time, further improving the accuracy and response speed of resource scheduling.

[0083] S2. Configure a remaining resource monitoring and prediction model on the SaaS application platform, and obtain the remaining completion time and distribution state of the remaining resources for the current business processing according to the current usage status data of the SaaS application platform and the status data in the corresponding historical business cycle through the configured remaining resource monitoring and prediction model;

[0084] Furthermore, in this embodiment, the step of obtaining the remaining completion time of the currently processed service and the distribution state of the remaining resources includes:

[0085] S201, obtaining a distribution state quantity sequence of historical resource requirements of all application services at each time point in N processing cycles, a corresponding distribution state quantity sequence of remaining resources, and a difference sequence between the expected service processing time length and the actual service processing time length of each application service in the corresponding cluster space in each processing cycle;

[0086] Furthermore, in this embodiment, the expected service processing time length is obtained by the difference between the service request timestamp and the set service completion timestamp;

[0087] S202, constructing a remaining resource monitoring prediction model based on a random forest algorithm, inputting the sequence obtained in S201 into the remaining resource monitoring prediction model for training, and obtaining a trained remaining resource monitoring prediction model;

[0088] S203, obtain the remaining processing resources at the current time point, the resource demand in the current processing cycle of the current upload processing business, and the expected business processing time length data of each business, and input the obtained data into the trained remaining resource monitoring and prediction model to predict the remaining completion time of each application business currently being processed and the remaining resource distribution state corresponding to each time point in the corresponding processing cycle.

[0089] This process can accurately reflect the historical pattern of resource usage by obtaining the historical resource demand distribution state quantity sequence of all application services at each time point in the historical N processing cycles, the corresponding remaining resource distribution state quantity sequence, and the difference sequence between the expected business processing time length and the actual business processing time length of each application service in the corresponding cluster space in each processing cycle. The remaining resource monitoring prediction model constructed using the random forest algorithm can accurately predict the remaining completion time of the current processing business and the distribution state quantity of the remaining resources by training the difference sequence between the expected business processing time length and the actual business processing time length; this process not only improves the accuracy of the prediction, but also enhances the adaptive ability of the system, ensuring the effective utilization and reasonable allocation of resources.

[0090] S3, according to the resource demand in each service cycle of the user's current type of service processing, the remaining completion time of the current service processing, and the distribution state of the remaining resources, the remaining resources are scheduled through the configured distributed enhanced resource scheduling model, and the missing resource amount is obtained;

[0091] Furthermore, the distributed enhanced resource scheduling model in this embodiment includes a central resource scheduling sub-model and I edge scheduling sub-models, and I edge scheduling sub-models correspond to I resource servers;

[0092] Furthermore, the steps of constructing the distributed enhanced resource scheduling model in this embodiment include:

[0093] S301: Based on the resource demand in the current processing cycle of the current upload processing business, the remaining completion time of each business currently being processed, the remaining resource distribution state corresponding to each time point in the processing cycle, and the remaining schedulable resources of each resource server, the current input state is constructed:

[0094]

[0095] in It represents the resource demand of the k-th application service at time t in the n-th processing cycle, and It represents the remaining resources of I resource servers at the current time t. Indicates the calling status of the i-th resource server at time t. When , it means that the i-th resource server has not been called at the current time t; when It means that the i-th resource server is in the calling state at time t. represents the remaining schedulable resources of the i-th resource server at time t, Indicates the remaining completion time of the k-th application being processed in the n-th processing cycle. Indicates the processing priority corresponding to the kth application service at time t in the nth processing cycle. Furthermore, the processing priority in this embodiment is set by setting the service completion timestamp. The earlier the service completion timestamp is, the higher the priority is. represents the cluster similarity space matrix corresponding to K application services in the g-th cluster space in the n-th processing cycle; Indicates the price of a single resource server called in the nth processing cycle.

[0096] represents the price of the jth uncalled resource server in the nth processing cycle, represents the probability that the i-th resource server in the calling state fails in the n-th processing cycle, Indicates the amount of missing resources corresponding to the k-th application service at the current time t in the n-th processing cycle.

[0097] S302, setting the execution action trigger information corresponding to the current moment according to the current input state quantity, including:

[0098] S3021. Setting a cluster matching threshold. When the remaining resources of the i-th resource server at the current time t meet the resource requirements of the application business k and the cluster matching value between the cluster similarity space matrix of the business being processed in the i-th resource server and the business processing status data of the application business k is greater than the cluster matching threshold, triggering the first execution action information, that is, uploading the application business k to the i-th resource server for processing;

[0099] S3022. When the remaining resources of the i-th resource server at the current time t do not meet the resource requirements of the application business k but meet the cluster matching value greater than the cluster matching threshold, the second execution action information is triggered, that is, the remaining resource distribution state of each scheduled resource server within the expected business processing time length of the application business k is monitored through the remaining resource monitoring prediction model, and the remaining resources of each scheduled resource server at each time are scheduled to the i-th resource server;

[0100] S3023. If the resources after scheduling do not meet the resource requirements of the application service k at the current moment, trigger the third execution action information;

[0101] That is, the jth uncalled resource server with the lowest price at the current moment corresponding to the jth edge scheduling submodel is scheduled through the central resource scheduling submodel to supplement resources. At the same time, the remaining completion time of the business being processed in the current resource server is monitored through the remaining resource monitoring and prediction model. When the remaining completion time of the business being processed is less than the expected business processing time length corresponding to the application business k, and the resources released after the business being processed is completed meet the resource demand of the application business k, the timestamp of the corresponding completion time of the processed business is obtained through the remaining resource monitoring and prediction model, and the scheduling of the jth resource server is disconnected through the central resource scheduling submodel at the corresponding completion time of the business. If the resources released after the business being processed is completed do not meet the resource demand of the application business k, the jth resource server is not disconnected.

[0102] S3024, setting a resource matching threshold, when the remaining resources of the i-th resource server at the current time t meet the resource requirements of the application business k, but the clustering similarity space matrix of the business being processed in all resource servers and the clustering matching value of the business processing status data of the application business k are less than or equal to the clustering matching threshold, triggering the fourth execution action information;

[0103] That is, the remaining resource distribution state quantity of each scheduled resource server within the expected service processing time length of the application service k is monitored through the remaining resource monitoring prediction model, and the obtained remaining resource distribution state quantity is matched with the resource demand of the application service k. If there is a resource server with a matching score greater than the resource matching threshold, the application service k is uploaded to the resource server corresponding to the maximum matching score for processing. If the matching scores are all less than or equal to the resource matching threshold, the application service k is uploaded to the called resource server with the lowest price within the expected service processing time length of the application service k for processing;

[0104] S3025. When the remaining resources of all resource servers at the current time t do not meet the resource requirements of the application business k and the clustering similarity space matrix of the business being processed in all resource servers and the clustering matching value of the business processing status data of the application business k are both less than or equal to the clustering matching threshold, the fifth execution action information is triggered, that is, the third execution action information is directly triggered, the resource server that has not been called is directly scheduled to supplement the resources, and the remaining resources of all resource servers at each moment are monitored, and it is determined whether to disconnect the resource server scheduled for supplementation according to the monitoring results;

[0105] S3026. When the number of currently applied services exceeds one, the sixth execution action information is triggered; that is, the clustering algorithm is used to determine whether the applied services belong to the same clustering space. If they belong to the same clustering space, all the applied services are placed in the same service processing queue. When the remaining resources at the current moment meet the resource requirements of all applied services in the service processing queue, the first execution action information or the fourth execution action information is triggered to process the applied services in the service processing queue.

[0106] S3027, when the remaining resources at the current moment do not meet the resource requirements of all the application services in the service processing queue, trigger the seventh execution action information; that is, directly trigger the second execution action information or the third execution action information to process the application services in the service processing queue;

[0107] S3028. If they do not belong to the same cluster space, trigger the eighth execution action information; that is, put all the application services belonging to the same cluster space in the same application service into the same service processing queue, and mark each queue with the average priority of all the application services in the same queue, and arrange the marked queues in descending order. When the remaining resources at the current moment meet the needs of all service processing queues, allocate resources according to the descending order of the queues.

[0108] S3029, when the remaining resources at the current moment do not satisfy all business processing queues, the remaining resources at the current moment are allocated to the queues arranged in descending order, and the missing resource amount corresponding to the remaining unobtained resource queues is obtained, and the missing resource amount and the missing resource amount are calculated according to the missing resource amount and The resource server is scheduled to be supplemented so that the scheduling cost is minimized while the resource demand is met, and the S3023 process is repeated to monitor the distribution state of the remaining resources through the remaining resource monitoring prediction model, and the resource server for supplementary scheduling is disconnected at the timestamp node that meets the conditions;

[0109] S30210. When the i-th resource server fails at the current moment, the ninth execution action information is triggered, that is, the remaining resource distribution state of the non-faulty resource server is monitored through the remaining resource monitoring and prediction model. If the remaining resource distribution state meets the business resource demand of the current faulty resource server, the corresponding processing business will be allocated to the non-faulty resource server. If it does not meet the demand, the central resource scheduling sub-model and the corresponding edge scheduling sub-model will be used to schedule the unused resource server to supplement the resources.

[0110] S303, constructing a current moment joint execution action at=(at1, at2, at3, at4, at5, at6, at7, at8, at9) based on the execution action triggering information corresponding to the current moment, wherein at1 to at9 are the actions executed corresponding to the triggering of the first execution action information to the triggering of the ninth execution action information, respectively;

[0111] Based on the SAC algorithm, a central resource scheduling sub-model and an edge scheduling sub-model are constructed, and the constructed current moment input state quantity, the execution action trigger information corresponding to the current moment, and the current moment joint execution action are input into the distributed enhanced resource scheduling model for training, and the trained distributed enhanced resource scheduling model is obtained. The trained distributed enhanced resource scheduling model is built into the SaaS application platform for adaptive resource scheduling and expansion;

[0112] By setting the cluster matching threshold and resource matching threshold, the model can make fine matches based on the remaining resources of the resource server and the cluster similarity of the business processing status data, ensuring that resource allocation meets both business needs and business types, thereby improving resource utilization and business processing efficiency; secondly, when the current remaining resources of the resource server do not meet the resource requirements of the applied business, the model monitors and schedules resources to ensure that business needs are met in the short term, and selects the optimal resource server for supplementary scheduling through price comparison, effectively reducing the scheduling cost; in addition, the model also considers the failure probability of the resource server, and can migrate the business to a healthy resource server in time, ensuring the continuity and stability of the business; thirdly, for multiple businesses applied for at the same time, the model puts the businesses belonging to the same cluster space into the same queue through the clustering algorithm, and allocates resources according to priority, thereby improving processing efficiency; when the remaining resources are insufficient, the model can give priority to queues with high priority, and schedule supplementary resources to ensure the optimization of resource allocation;

[0113] S4. Based on the acquired missing resource amount, cloud service resource supplementation scheduling is performed through a distributed enhanced resource scheduling model, so that the total cost is minimized while meeting the current business processing volume;

[0114] S5. Set the evaluation score threshold, repeat S1 to S4 to perform business processing tests, and obtain the resource allocation, completion and cost data of the corresponding business. Evaluate the test process through the configured evaluation algorithm. When the evaluation result is greater than the evaluation score threshold, keep the current resource allocation parameters unchanged. Otherwise, feed back the evaluation score to the distributed enhanced resource scheduling model to adjust the current resource allocation parameters.

[0115] Example 2

[0116] See also Figure 2 , another embodiment provided by the present invention: an adaptive expansion system for SaaS applications based on the industrial Internet, comprising: a demand forecasting module, a resource monitoring module, a resource scheduling module and a test feedback module;

[0117] Demand forecasting module, used to receive services uploaded by users and forecast demand resources;

[0118] The demand prediction module includes a receiving unit and a prediction unit; the receiving unit is used to receive the corresponding type of processing business uploaded by the user, and pre-process and save the received data;

[0119] The prediction unit is used to obtain the resource demand of the user's current type of processing service in each service cycle through the configured demand prediction model according to the corresponding type of processing service uploaded by the user and the corresponding historical processing service data;

[0120] The resource monitoring module is used to obtain the remaining completion time and the distribution state of the remaining resources of the current processing business based on the current usage status data of the SaaS application platform and the status data in the corresponding historical business cycle through the configured remaining resource monitoring prediction model;

[0121] The resource scheduling module is used for scheduling the remaining resources and supplementing the missing resources; the resource scheduling module includes a resource scheduling unit and an expansion unit;

[0122] A resource scheduling unit is used to schedule the remaining resources using a distributed enhanced resource scheduling model according to the resource demand, the remaining completion time, and the change in the remaining resource distribution state, and obtain the current missing resource amount;

[0123] The expansion unit is used to supplement and schedule cloud service resources according to the amount of missing resources, ensuring that the total cost is minimized while meeting the current business processing volume;

[0124] The test feedback module is used for test evaluation and feedback optimization of resource scheduling during business processing; the test feedback module includes a test evaluation unit and an evaluation feedback unit; the test evaluation unit is used to obtain the evaluation score of the corresponding test process through the configured evaluation algorithm according to the resource allocation, completion and cost data of the corresponding type of business obtained during the test process;

[0125] The evaluation feedback unit is used to provide feedback optimization and adjustment to the current resource allocation parameters according to the evaluation score and evaluation score threshold in the corresponding test process.

[0126] Example 3

[0127] A computer-readable storage medium stores computer instructions, which, when executed, execute an adaptive expansion method for SaaS applications based on the industrial Internet.

[0128] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements an adaptive expansion method for SaaS applications based on the industrial Internet when executing the computer program.

[0129] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are within the protection of the present invention.

[0130] If the disclosed technical solution involves personal information, the product using the disclosed technical solution has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the disclosed technical solution involves sensitive personal information, the product using the disclosed technical solution has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, when the personal information processing rules are notified by obvious signs / information, the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. An adaptive expansion method for SaaS applications based on the industrial Internet, characterized in that: include: S1. Receive the corresponding type of processing business uploaded by the user, and obtain the resource demand of the user's current type of processing business in each business cycle according to the corresponding type of processing business uploaded by the user and the corresponding historical processing business data through the configured demand prediction model; S2. Configure a remaining resource monitoring and prediction model on the SaaS application platform, and obtain the remaining completion time and distribution state of the remaining resources for the current business processing according to the current usage status data of the SaaS application platform and the status data in the corresponding historical business cycle through the configured remaining resource monitoring and prediction model; S3, according to the resource demand in each service cycle of the user's current type of service processing, the remaining completion time of the current service processing, and the distribution state of the remaining resources, the remaining resources are scheduled through the configured distributed enhanced resource scheduling model, and the missing resource amount is obtained; S4. Perform cloud service resource supplementation scheduling based on the acquired missing resource amount through the distributed enhanced resource scheduling model; S5. Set the evaluation score threshold, repeat S1 to S4 to perform business processing tests, and obtain the resource allocation, completion and cost data of the corresponding business. Evaluate the test process through the configured evaluation algorithm. When the evaluation result is greater than the evaluation score threshold, keep the current resource allocation parameters unchanged. Otherwise, feed back the evaluation score to the distributed enhanced resource scheduling model to adjust the current resource allocation parameters. The step of obtaining the resource demand of the user's current type of processing service in each service cycle includes: Firstly, the business processing status data of N processing cycles of historical M application services are obtained, and the data is pre-processed and then input into the clustering algorithm to classify similar business types into the same clustering space; secondly, a demand prediction model containing G integrated sub-models is constructed through support vector machine and integration algorithm to predict the business processing resource demand in each clustering space, and the model is continuously adjusted until the threshold condition is met by setting the sum threshold of sub-errors and the sum threshold of resource demand errors; finally, the real-time business processing status data of the current uploaded processing business is analyzed to determine the processing cycle to which it belongs, and the demand prediction model of the corresponding processing cycle is called to predict the resource demand of the current business type in real time; The step of obtaining the remaining completion time and the distribution state of the remaining resources of the currently processed business includes: By obtaining the historical resource demand distribution state quantity sequence of all application services at each time point in the historical N processing cycles, the corresponding remaining resource distribution state quantity sequence, and the difference sequence between the expected business processing time length and the actual business processing time length of each application service in the corresponding cluster space in each processing cycle, the remaining resource monitoring and prediction model constructed by the random forest algorithm is trained on the difference sequence between the expected business processing time length and the actual business processing time length, and the remaining completion time of each application service currently being processed and the corresponding remaining resource distribution state quantity at each time point in the corresponding processing cycle are predicted; The distributed enhanced resource scheduling model includes a central resource scheduling sub-model and I edge scheduling sub-models, and the I edge scheduling sub-model corresponds to I resource server; The distributed enhanced resource scheduling model includes a central resource scheduling sub-model and I edge scheduling sub-models, and the I edge scheduling sub-model corresponds to I resource server; the steps of constructing the distributed enhanced resource scheduling model include: By setting the cluster matching threshold and the resource matching threshold, the distributed enhanced resource scheduling model matches the remaining resources of the resource server and the cluster similarity of the business processing status data to ensure that resource allocation meets both business needs and business types; secondly, when the current remaining resources of the resource server do not meet the application business resource needs, the distributed enhanced resource scheduling model monitors and schedules resources to ensure that business needs are met in the short term, and selects the optimal resource server for supplementary scheduling through price comparison; when a resource server fails, the business is migrated to a healthy resource server. For multiple businesses applied for at the same time, the distributed enhanced resource scheduling model uses a clustering algorithm to put businesses belonging to the same cluster space into the same queue, and allocates resources according to priority; when the remaining resources are insufficient, the distributed enhanced resource scheduling model can give priority to resource scheduling compensation to queues with high priority.

2. The method for adaptive expansion of SaaS applications based on the Industrial Internet according to claim 1, characterized in that: The step of obtaining the resource demand of the user's current type of processing service in each service cycle includes: S101, obtaining business processing status data corresponding to N processing cycles of M historical application businesses, and preprocessing the obtained data; The business processing status data includes business type, business scale, business request timestamp, set business completion timestamp, actual business completion timestamp and business resource demand; S102, input the pre-processed business processing status data into the clustering algorithm, and obtain (ng, gk) corresponding to each processing cycle in N processing cycles, where ng represents the number of application services in the g-th clustering space in the n-th processing cycle, and gk represents the business processing status data sequence corresponding to the k-th application service in the g-th clustering space in the n-th processing cycle, where g∈G, k∈K, where G represents the total number of clustering spaces, and K represents the number of application services in each clustering space; S103. Construct a demand forecasting model containing G integrated sub-models based on support vector machine and integration algorithm, input (ng, gk) corresponding to the G clustering spaces in the nth processing cycle into each integrated sub-model in the demand forecasting model for training, and output the predicted amount of business processing resource demand for K application businesses in the corresponding clustering space.

3. The adaptive expansion method of SaaS application based on Industrial Internet as claimed in claim 2, characterized in that: The step of obtaining the resource demand of the user's current type of processing service in each service cycle also includes: S104, using the predicted business processing resource requirements of K application business types in the output corresponding cluster space and the corresponding actual business processing resource requirements, obtain the sum of the processing resource requirements quantum errors ek corresponding to the K application businesses in the g-th cluster space in the n-th processing cycle The sum of resource demand errors corresponding to G clustering spaces eg,k S105: Set the sub-error sum threshold and the resource demand error sum threshold. is less than the sum threshold of sub-errors and When the error sum of resource demand is less than the threshold, the training is completed, and the trained demand forecasting model is obtained; S106, if or If the corresponding threshold is not met, or Feedback to the demand forecasting model for training until the threshold condition is met, and a trained demand forecasting model is obtained; S107, repeating the process of S103-S106, inputting the business processing status data corresponding to N processing cycles into the demand forecasting model for training, and obtaining N trained demand forecasting models; S108, obtaining the real-time business processing status data corresponding to the current upload processing business, and judging the processing cycle to which the current upload processing business belongs through the business request timestamp in the real-time business processing status data, and calling the demand prediction model of the corresponding processing cycle according to the processing cycle to which it belongs; S109: input the real-time business processing status data corresponding to the current upload processing business into the called demand prediction model to obtain the resource demand in the processing cycle to which the current upload processing business belongs.

4. The adaptive expansion method of SaaS application based on Industrial Internet as claimed in claim 3, characterized in that: The step of obtaining the remaining completion time and the distribution state of the remaining resources of the currently processed business includes: S201, obtaining a distribution state quantity sequence of historical resource requirements of all application services at each time point in N processing cycles, a corresponding distribution state quantity sequence of remaining resources, and a difference sequence between the expected service processing time length and the actual service processing time length of each application service in the corresponding cluster space in each processing cycle; S202, constructing a remaining resource monitoring prediction model based on a random forest algorithm, inputting the sequence obtained in S201 into the remaining resource monitoring prediction model for training, and obtaining a trained remaining resource monitoring prediction model; S203, obtain the remaining processing resources at the current time point, the resource demand in the current processing cycle of the current upload processing business, and the expected business processing time length data of each business, and input the obtained data into the trained remaining resource monitoring and prediction model to predict the remaining completion time of each application business currently being processed and the remaining resource distribution state corresponding to each time point in the corresponding processing cycle.

5. The adaptive expansion method of SaaS application based on Industrial Internet as claimed in claim 4, characterized in that: The steps of constructing the distributed enhanced resource scheduling model include: S301: Based on the resource demand in the current processing cycle of the current upload processing business, the remaining completion time of each business currently being processed, the remaining resource distribution state corresponding to each time point in the processing cycle, and the remaining schedulable resources of each resource server, the current input state is constructed: in It represents the resource demand of the k-th application service at time t in the n-th processing cycle, and It represents the remaining resources of I resource servers at the current time t. Indicates the calling status of the i-th resource server at time t. When , it means that the i-th resource server has not been called at the current time t; when It means that the i-th resource server is in the calling state at time t. represents the remaining schedulable resources of the i-th resource server at time t, Indicates the remaining completion time of the k-th application being processed in the n-th processing cycle. Indicates the processing priority of the kth application service at time t in the nth processing cycle. represents the cluster similarity space matrix corresponding to K application services in the g-th cluster space in the n-th processing cycle; Indicates the price of a single resource server called in the nth processing cycle. represents the price of the jth uncalled resource server in the nth processing cycle, pin represents the probability of failure of the i-th resource server in the called state in the nth processing cycle, and ftkn represents the amount of missing resources corresponding to the k-th application service at the current time t in the nth processing cycle.

6. The method for adaptive expansion of SaaS applications based on the Industrial Internet according to claim 5, characterized in that: The step of constructing the distributed enhanced resource scheduling model also includes: S302, setting the execution action trigger information corresponding to the current moment according to the current input state quantity, including: S3021. Setting a cluster matching threshold. When the remaining resources of the i-th resource server at the current time t meet the resource requirements of the application business k and the cluster matching value between the cluster similarity space matrix of the business being processed in the i-th resource server and the business processing status data of the application business k is greater than the cluster matching threshold, triggering the first execution action information, that is, uploading the application business k to the i-th resource server for processing; S3022. When the remaining resources of the i-th resource server at the current time t do not meet the resource requirements of the application business k but meet the cluster matching value greater than the cluster matching threshold, the second execution action information is triggered, that is, the remaining resource distribution state of each scheduled resource server within the expected business processing time length of the application business k is monitored through the remaining resource monitoring prediction model, and the remaining resources of each scheduled resource server at each time are scheduled to the i-th resource server; S3023. If the resources after scheduling do not meet the resource requirements of the application service k at the current moment, trigger the third execution action information; That is, the jth uncalled resource server with the lowest price at the current moment corresponding to the jth edge scheduling submodel is scheduled through the central resource scheduling submodel to supplement resources. At the same time, the remaining completion time of the business being processed in the current resource server is monitored through the remaining resource monitoring and prediction model. When the remaining completion time of the business being processed is less than the expected business processing time length corresponding to the application business k, and the resources released after the business being processed is completed meet the resource demand of the application business k, the timestamp of the corresponding completion time of the processed business is obtained through the remaining resource monitoring and prediction model, and the scheduling of the jth resource server is disconnected through the central resource scheduling submodel at the corresponding completion time of the business. If the resources released after the business being processed is completed do not meet the resource demand of the application business k, the jth resource server is not disconnected.

7. The method for adaptive expansion of SaaS applications based on the industrial Internet according to claim 6, characterized in that: The step of constructing the distributed enhanced resource scheduling model also includes: S3024, setting a resource matching threshold, when the remaining resources of the i-th resource server at the current time t meet the resource requirements of the application business k, but the clustering similarity space matrix of the business being processed in all resource servers and the clustering matching value of the business processing status data of the application business k are less than or equal to the clustering matching threshold, triggering the fourth execution action information; That is, the remaining resource distribution state quantity of each scheduled resource server within the expected service processing time length of the application service k is monitored through the remaining resource monitoring prediction model, and the obtained remaining resource distribution state quantity is matched with the resource demand of the application service k. If there is a resource server with a matching score greater than the resource matching threshold, the application service k is uploaded to the resource server corresponding to the maximum matching score for processing. If the matching scores are all less than or equal to the resource matching threshold, the application service k is uploaded to the called resource server with the lowest price within the expected service processing time length of the application service k for processing; S3025. When the remaining resources of all resource servers at the current moment t do not meet the resource requirements of the application business k and the clustering similarity space matrix of the business being processed in all resource servers and the clustering matching value of the business processing status data of the application business k are less than or equal to the clustering matching threshold, the fifth execution action information is triggered, that is, the third execution action information is directly triggered, and the uncalled resource servers are directly scheduled to supplement the resources, and the remaining resources of all resource servers at each moment are monitored, and it is determined whether to disconnect the resource servers scheduled for supplementation based on the monitoring results.

8. The method for adaptive expansion of SaaS applications based on the industrial Internet according to claim 7, characterized in that: The step of constructing the distributed enhanced resource scheduling model also includes: S3026. When the number of currently applied services exceeds one, the sixth execution action information is triggered; that is, the clustering algorithm is used to determine whether the applied services belong to the same clustering space. If they belong to the same clustering space, all the applied services are placed in the same service processing queue. When the remaining resources at the current moment meet the resource requirements of all applied services in the service processing queue, the first execution action information or the fourth execution action information is triggered to process the applied services in the service processing queue. S3027, when the remaining resources at the current moment do not meet the resource requirements of all the application services in the service processing queue, trigger the seventh execution action information; that is, directly trigger the second execution action information or the third execution action information to process the application services in the service processing queue; S3028. If they do not belong to the same cluster space, trigger the eighth execution action information; that is, put all the application services belonging to the same cluster space in the same application service into the same service processing queue, and mark each queue with the average priority of all the application services in the same queue, and arrange the marked queues in descending order. When the remaining resources at the current moment meet the needs of all service processing queues, allocate resources according to the descending order of the queues. S3029, when the remaining resources at the current moment do not satisfy all business processing queues, the remaining resources at the current moment are allocated to the queues arranged in descending order, and the missing resource amount corresponding to the remaining unobtained resource queues is obtained, and the missing resource amount and the missing resource amount are calculated according to the missing resource amount and The resource server is scheduled to be supplemented so that the scheduling cost is minimized while meeting the resource demand, and the S3023 process is repeated. The distribution state of the remaining resources is monitored through the remaining resource monitoring and prediction model, and the resource server for supplementary scheduling is disconnected at the timestamp node that meets the conditions.

9. An adaptive expansion system for SaaS applications based on the industrial Internet, which is implemented based on the adaptive expansion method for SaaS applications based on the industrial Internet according to any one of claims 1 to 8, characterized in that: include: Demand forecasting module, resource monitoring module, resource scheduling module and test feedback module; The demand prediction module includes a receiving unit and a prediction unit; the receiving unit is used to receive the corresponding type of processing business uploaded by the user, and pre-process and save the received data; The prediction unit is used to obtain the resource demand of the user's current type of processing service in each service cycle through the configured demand prediction model according to the corresponding type of processing service uploaded by the user and the corresponding historical processing service data; The resource monitoring module is used to obtain the remaining completion time and the distribution state of the remaining resources of the current processing business according to the current usage status data of the SaaS application platform and the status data in the corresponding historical business cycle through the configured remaining resource monitoring prediction model; The resource scheduling module includes a resource scheduling unit and an expansion unit; The resource scheduling unit is used to schedule the remaining resources using a distributed enhanced resource scheduling model according to the resource demand, the remaining completion time and the change in the remaining resource distribution state, and obtain the current missing resource amount; The expansion unit is used to perform cloud service resource supplementation scheduling according to the amount of missing resources; The test feedback module includes a test evaluation unit; the test evaluation unit is used to obtain an evaluation score in the corresponding test process through a configured evaluation algorithm based on the resource allocation, completion and cost data of the corresponding type of business obtained during the test process.

10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the adaptive expansion method of SaaS application based on industrial Internet described in any one of claims 1-8 is executed.

Images