Resource management and control method and system for global distributed monitoring system of plant station

Abstract

The invention discloses a resource management and control method and system for a global distributed monitoring system of a plant station, and the resource management and control service achieves the unified access of resources of the global distributed monitoring system and the self-adaptive data collection based on the fluctuation amplitude of a resource load. The data governance service starts a long-short-term memory network prediction restoration and historical data association restoration module as required to realize full-life-cycle data governance and quality guarantee of system resource monitoring data; the intelligent scheduling service adopts a multi-target load balancing and service-driven elastic scaling strategy, so that accurate matching of resource allocation, service requirements and system operation states is realized, the resource utilization rate and the system elasticity are improved, and continuous operation of key services is guaranteed; the global resource view service constructs a global resource visual view, and provides visual global management and control support for operation and maintenance personnel. According to the invention, reasonable arrangement and dynamic scheduling of services can be realized, the utilization rate, elasticity and reliability of system resources are improved, and continuous operation of key businesses of a plant-station global distributed monitoring system is guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of resource management technology for plant-wide distributed monitoring systems, and particularly relates to a resource management method and system for plant-wide distributed monitoring systems. Background Technology

[0002] With the deepening of industrial digital transformation, the deployment scale of distributed monitoring systems in core power grid business scenarios such as power plants and centralized control stations continues to expand, and the various operating resources involved in the systems, such as computing, storage, networks, and substation measurement and control devices, protection devices, and remote control devices, are becoming increasingly complex. In order to adapt to the requirements of high-concurrency data processing, multi-node collaborative operation, and differentiated monitoring business priorities, the complexity of the system operating environment has increased, and higher requirements are placed on the accuracy, real-time performance, and elastic scalability of operating resource management.

[0003] However, existing distributed resource management technologies have many shortcomings, making them difficult to adapt to the actual operational needs of complex substations and centralized control stations: First, they mostly adopt single-node management and static configuration architectures, resulting in high management latency, significant single-point failure risks, weak horizontal scalability, and a lack of cross-safety zone adaptability in large-scale heterogeneous resource scenarios. This makes them unable to meet the differentiated safety requirements of substation zones I to IV, leading to difficult implementation and poor operational stability. Second, their data acquisition modes are rigid, lacking the ability to adapt to multi-source heterogeneous resources, resulting in low transmission and processing efficiency, static data governance, and a disconnect from dynamic resource scheduling, failing to support intelligent decision-making. Third, resource scheduling lacks dynamic adjustment capabilities, making it difficult to match the characteristics and business priorities of different substation safety zones, leading to low resource utilization, insufficient elasticity, and difficulty in ensuring the stable operation of critical businesses. These problems severely restrict the continuous operation guarantee capability of critical substation businesses and hinder the large-scale promotion of distributed monitoring systems. Summary of the Invention

[0004] Purpose of the invention: This invention provides a resource management method and system for a plant-wide distributed monitoring system. By unifying and integrating all resources across the system, a complete plant-level visualization view of all resources is constructed. Relying on intelligent load balancing algorithms and dynamic elastic expansion mechanisms, combined with refined orchestration and dynamic scheduling strategies for distributed monitoring system services, the system resource utilization rate is significantly improved, the elastic expansion capability and stable operation reliability of the system are enhanced, and the continuous and uninterrupted operation of critical plant services is guaranteed.

[0005] Technical solution: This invention provides a resource management method for a plant-wide distributed monitoring system, comprising:

[0006] Set a sampling period, collect resource usage data of each original service in the plant-wide distributed monitoring system according to the sampling period, and store it in the time series database; monitor the load fluctuation of resources in the plant-wide distributed monitoring system, set a sliding window time, calculate the load fluctuation amplitude of each resource within the sliding window time; adjust the collection period according to the load fluctuation amplitude of each resource and the priority of each original service.

[0007] Set up a statistics window to calculate the quality and integrity of resource usage data in the time series database. Based on the quality and integrity of the resource usage data in the time series database and the priority of the original business, determine whether the resource usage data meets the preset standards. If it does not meet the preset standards, perform secondary optimization processing on the resource usage data that does not meet the preset standards, and store the processed resource usage data in the time series database.

[0008] Resource usage data of each original business in each node of the plant-wide distributed monitoring system are obtained from the time series database, and the comprehensive load score of each node is calculated by weighting them. The comprehensive load score weight combination of each node is adjusted according to the business scenario and the preset comprehensive load score weight combination of each business scenario, and the comprehensive load score of each node is updated.

[0009] Calculate the expected resource usage data for the new service. Based on the resource usage data of each original service in the node and the expected resource usage data of the new service, obtain all nodes that meet the preset screening conditions, and deploy the new service to the node with the lowest overall load score.

[0010] Build a unified global resource visualization view to provide a global resource view service.

[0011] Furthermore, the resources include CPU, memory, I / O, database, middleware, and business services; the resource usage data includes CPU utilization, memory usage, I / O traffic, database load, middleware load, and average communication latency of business services; the priority of the services includes high priority and low priority.

[0012] Furthermore, the formula for calculating the load fluctuation amplitude of each resource within the sliding window time is as follows:

[0013] ;

[0014] in, This represents the sampled payload values ​​within the sliding window time t. This represents the peak value of the payload samples within the sliding window time t. This represents the valley value among the payload samples within the sliding window time t. This represents the average value of the payload samples within the sliding window time t;

[0015] The adjustment of the data collection cycle based on the load fluctuation of each resource and the priority of each original service includes:

[0016] When the resource load fluctuation range within the sliding window time t is greater than 10% and the original service is a preset high-priority service, the collection cycle is adjusted to 1 second; when the resource load fluctuation range is less than or equal to 10% or the original service is a preset low-priority service, the collection cycle is adjusted to 5 seconds.

[0017] Furthermore, the quality integrity of resource usage data in the statistical window within the computation time-series database is determined by the following formula:

[0018] ;

[0019] in, This indicates the number of valid data entries within the statistics window. This indicates the total number of data entries that should theoretically be collected within the statistics window.

[0020] Furthermore, the preset standard is that if the resource usage data is for low-priority services, the quality integrity of the resource usage data must be greater than 95%; if the resource usage data is for high-priority services, the quality integrity of the resource usage data must be greater than 98%.

[0021] Furthermore, the secondary optimization process includes:

[0022] If the resource usage data exceeds the preset operating threshold, the system is in a high-load scenario; otherwise, the system is in a low-load scenario.

[0023] If the system is under high load and resource usage data fluctuates or suddenly increases, the Long Short-Term Memory Network (LSTM) temporal prediction and repair module is activated to repair the resource usage data. The LSTM temporal prediction and repair module is based on historical continuous time-series running resource data and learns the temporal change pattern of running resource indicators through an LSTM network to predict, complete, and repair instantaneous missing or abnormal data under high load scenarios.

[0024] If the system is in a low-load scenario and resource usage data is missing, the historical data association and repair module will be activated to repair the resource usage data. The historical data association and repair module, based on the historical running resource database, matches historical similar data corresponding to the current collection time in the low-load scenario, and completes the missing data repair through association deduction.

[0025] Furthermore, the calculation of the overall load score for each node includes:

[0026] Retrieve CPU and memory utilization rates for each node from the time-series database, and count the number of high-priority services deployed on each node. Calculate the percentage of high-priority services deployed on each node. Based on the current CPU and memory utilization rates and the percentage of high-priority services deployed on each node, calculate the weighted overall load score F for each node using the following formula:

[0027] ;

[0028] Where C is the CPU utilization of each node; M is the memory utilization of each node; P is the number of high-priority services deployed on each node; and N is the upper limit of high-priority services that can be deployed on a single node. The percentage of high-priority services deployed on each node; Weights representing CPU utilization; Weights representing memory usage; This indicates the weight of the percentage of high-priority services deployed on each node, and F represents the overall load score.

[0029] Furthermore, adjusting the comprehensive load score weight combination of each node according to the business scenario and the preset comprehensive load score weight combination for each business scenario includes:

[0030] If the business scenario requires prioritizing the stability of high-priority services, configure... =0.3、 =0.2、 =0.5;

[0031] If the business scenario requires prioritizing resource utilization, configure... =0.4、 =0.4、 =0.2;

[0032] If the business scenario requires a balance between stability and resource utilization, configure =0.25、 =0.25、 =0.5.

[0033] Furthermore, obtaining all nodes that meet the preset filtering conditions and deploying the new service to the node with the lowest overall load score includes:

[0034] First, select all nodes that simultaneously satisfy both condition one and condition two to form the first node set;

[0035] Condition 1 is that the number of high-priority services deployed in the node is less than the preset service deployment limit;

[0036] Condition 2 is that the sum of the number of CPU cores used by the services already deployed in the node and the number of CPU cores required by the new services is less than the preset CPU core count threshold of the node.

[0037] If the new service is not within the preset high-priority service scope, the new service will be deployed on the node with the lowest overall load score in the first node set;

[0038] If the new service falls within the preset high-priority service range, then define CPU reservation thresholds and memory reservation thresholds for the new service.

[0039] CPU reserved threshold = base threshold - peak CPU utilization of deployed high-priority services;

[0040] Memory reservation threshold = base threshold - peak memory utilization of deployed high-priority services;

[0041] Nodes that meet the requirements of having a CPU utilization rate lower than the CPU reservation threshold and a memory utilization rate lower than the memory reservation threshold after deploying new services are selected from the first node set to form a second node set. New services within the preset high-priority service scope will be deployed on the node with the lowest overall load score in the second node set.

[0042] This invention also provides a plant-wide distributed monitoring system for resource management, comprising:

[0043] The resource management module is used to set the sampling period, collect resource usage data of each original service in the plant-wide distributed monitoring system according to the sampling period, and store it in the time series database; monitor the load fluctuation of resources in the plant-wide distributed monitoring system, set the sliding window time, calculate the load fluctuation amplitude of each resource within the sliding window time; and adjust the collection period according to the load fluctuation amplitude of each resource and the priority of each original service.

[0044] The data governance module is used to set up a statistics window, calculate the quality and integrity of resource usage data in the time series database within the statistics window, determine whether the resource usage data meets the preset standards based on the quality and integrity of the resource usage data in the time series database and the priority of the original business, and if it does not meet the preset standards, perform secondary optimization processing on the resource usage data that does not meet the preset standards, and store the processed resource usage data in the time series database.

[0045] The intelligent scheduling module is used to obtain resource usage data of each original service in each node of the plant's full-domain distributed monitoring system from the time-series database, and calculate the comprehensive load score of each node by weighting it; and adjust the comprehensive load score weight combination of each node according to the business scenario and the preset comprehensive load score weight combination of each business scenario, and update the comprehensive load score of each node; calculate the expected resource usage data of the new service, and obtain all nodes that meet the preset screening conditions based on the resource usage data of each original service in the node and the expected resource usage data of the new service, and deploy the new service to the node with the lowest comprehensive load score;

[0046] The global resource view display module is used to build a unified global resource visualization view and provide a global resource view service.

[0047] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0048] 1. Deep synergy between systems and methods: Through resource management, data governance, intelligent scheduling, and global resource visualization. Figure 4 The layered collaborative architecture design enables precise support of management and control strategies by system modules. Communication between components is based on the RPC protocol that meets the requirements of substation safety zones I to IV. This not only ensures the security and compliance of cross-safety zone calls, but also strengthens the decoupling and collaboration between services at the communication level, significantly improving the robustness, scalability, and reliability of the overall system architecture.

[0049] 2. High efficiency in data acquisition and governance: By introducing an adaptive sampling and adjustment algorithm based on resource load fluctuations, a dynamic and accurate data foundation is provided for data acquisition and scheduling decisions. Combined with the intelligent switching mechanism between the Long Short-Term Memory (LSTM) time-series prediction repair module and the historical data association repair module, the repair quality and processing efficiency of multi-source heterogeneous data are significantly improved.

[0050] 3. High scheduling accuracy and resource utilization: Through multi-objective load balancing and business-driven elastic scaling strategies, the system achieves precise matching between resource allocation and real-time business needs and system operating status, improving overall resource utilization and system elasticity, and effectively ensuring the continuity and stability of critical business operations. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the resource management method for the plant-wide distributed monitoring system of the present invention;

[0052] Figure 2 This is a schematic diagram of the resource management service collection cycle adjustment logic of the present invention;

[0053] Figure 3This is a logical diagram illustrating the data governance service of the present invention for performing secondary optimization processing on data that does not meet the judgment criteria.

[0054] Figure 4 This is a schematic diagram of the intelligent scheduling service dynamic weight allocation mechanism of the present invention;

[0055] Figure 5 This is a schematic diagram of the resource management system architecture of the plant-wide distributed monitoring system of the present invention. Detailed Implementation

[0056] like Figure 1 As shown, the resource management method for a plant-wide distributed monitoring system according to the present invention includes:

[0057] S1. Set the sampling period. Collect resource usage data of each original service in the plant-wide distributed monitoring system according to the sampling period and store it in the time-series database. The resources include CPU, memory, I / O, database, middleware, and business services. The resource usage data includes CPU utilization, memory usage, I / O traffic, database load, middleware load, and average communication latency of business services. The default sampling period is 5 seconds. The plant-wide distributed monitoring system also includes nodes, which are business deployment servers.

[0058] S2. Monitor the load fluctuations of each resource in the plant's distributed monitoring system in real time, and set a sliding window time to calculate the load fluctuation amplitude of each resource. The formula is:

[0059] ;

[0060] in, This represents the sampled payload values ​​within the sliding window time t. This represents the peak value of the payload samples within the sliding window time t. This represents the valley value among the payload samples within the sliding window time t. This represents the average value of the payload samples within the sliding window time t;

[0061] S3. Based on the load fluctuation range of each resource The data collection cycle is adjusted in real time according to the priority of each original business. Based on the priority of each original business, the original businesses can be divided into high-priority businesses and low-priority businesses. In the plant-wide distributed monitoring system, high-priority businesses generally include database proxy services, permission services, data model services, computing services, real-time data retrieval services, historical data retrieval services, change data aggregation services, alarm services, and unified interface services, etc. Low-priority businesses generally include log collection and archiving services, report generation services, configuration backup and recovery services, historical data cleanup services, operation record query services, system inspection services, and publishing services, etc.

[0062] like Figure 2 As shown, when the resource load fluctuation range within the sliding window time t is greater than 10% and the original service is a high-priority service, the collection cycle is adjusted to 1 second; when the resource load fluctuation range is less than or equal to 10% or the original service is a low-priority service, the collection cycle is adjusted to 5 seconds.

[0063] S4. Set up a statistics window to calculate the quality and integrity of data in the time series database, such as... Figure 3 As shown, the formula for calculating the quality integrity is:

[0064] ;

[0065] in, This indicates the number of valid data entries within the statistics window. This indicates the total number of data entries that should theoretically be collected within the statistics window.

[0066] S5. Based on the quality and integrity of resource usage data in the time-series database and the priority of the original business, determine whether the resource usage data meets the preset standards, such as... Figure 3 As shown;

[0067] The preset standard is:

[0068] If the resource usage data pertains to low-priority services, then the integrity of the resource usage data must be greater than 95%.

[0069] If the resource usage data pertains to high-priority services, then the quality and integrity of the resource usage data must be greater than 98%.

[0070] S6. If the resource usage data in the time-series database does not meet the preset standard, then the resource usage data that does not meet the preset standard will undergo secondary optimization processing, and the processed resource usage data will be stored in the time-series database, such as... Figure 3 As shown;

[0071] If the resource usage data (such as CPU utilization, memory usage, I / O traffic, and other data related to running resources) exceeds a preset operating threshold (such as 50%), the system is in a high-load scenario; otherwise, the system is in a low-load scenario.

[0072] If the system is under high load and resource usage data fluctuates or suddenly increases, the Long Short-Term Memory Network (LSTM) temporal prediction and repair module is activated to repair the resource usage data. The LSTM temporal prediction and repair module is based on historical continuous time-series running resource data and learns the temporal change pattern of running resource indicators through an LSTM network to predict, complete, and repair instantaneous missing or abnormal data under high load scenarios.

[0073] If the system is in a low-load scenario and resource usage data is missing, the historical data association and repair module will be activated to repair the resource usage data. The historical data association and repair module, based on the historical running resource database, matches historical similar data corresponding to the current collection time in the low-load scenario, and completes the missing data repair through association deduction.

[0074] S7. Obtain the CPU utilization and memory utilization of each node from the time-series database, and count the number of high-priority services deployed on each node. Calculate the percentage of high-priority services deployed on each node. Based on the current CPU utilization, memory utilization, and the percentage of high-priority services deployed on each node, calculate the comprehensive load score F for each node using the following weighted average:

[0075] ;

[0076] Where C is the CPU utilization of each node (normalized to the 0-1 range); M is the memory utilization of each node (normalized to the 0-1 range); P is the number of high-priority services deployed on each node; and N is the upper limit of high-priority services that can be deployed on a single node. The percentage of high-priority services deployed on each node; Weights representing CPU utilization; Weights representing memory usage; This indicates the weight of the percentage of high-priority services deployed on each node, and F represents the overall load score; the higher the score, the heavier the load.

[0077] S8. Adjust the comprehensive load score weight combination of each node according to the business scenario and the preset comprehensive load score weight combination for each business scenario, update the comprehensive load score of each node, and achieve the differentiated scheduling goal; the preset comprehensive load score weight combination is as follows: Figure 4 As shown, it includes:

[0078] If the business scenario requires prioritizing the stability of high-priority services, configure... =0.3、 =0.2、 =0.5;

[0079] If the business scenario requires prioritizing resource utilization, configure... =0.4、 =0.4、 =0.2;

[0080] If the business scenario requires a balance between stability and resource utilization, configure =0.25、 =0.25、 =0.5;

[0081] S9. Obtain new business and its expected resource usage data. Based on the resource usage data of each business in the node and the expected resource usage data of the new business, first filter out all nodes that simultaneously meet condition one and condition two to form the first node set.

[0082] Condition 1 is that the number of high-priority services deployed in a node is less than the preset service deployment limit value N (N is the upper limit of high-priority service deployment on a single node, and N is generally set to 3).

[0083] Condition 2 is that the sum of the number of CPU cores used by the services already deployed in the node and the number of CPU cores required by the new services is less than the preset CPU core count threshold of the node.

[0084] If the new service is not within the scope of the preset high-priority services (such as permission services, data model services, computing services, and unified interface services of the plant-wide distributed monitoring system), then the new service will be deployed on the node with the lowest overall load score in the first node set.

[0085] If the new service falls within the preset high-priority service range, i.e., the new service is a high-priority service such as permission service, data model service, computing service, or unified interface service of the plant-wide distributed monitoring system, then a CPU reservation threshold and a memory reservation threshold are defined for the new service; the reservation threshold is dynamically calculated as follows:

[0086] CPU reserved threshold = base threshold - peak CPU utilization of deployed high-priority services;

[0087] Memory reservation threshold = base threshold - peak memory utilization of deployed high-priority services;

[0088] Nodes selected from the first node set that meet the requirements for deploying new services, with both CPU and memory utilization rates below the reserved thresholds, are formed into a second node set. New services within the pre-defined high-priority service scope will be deployed on the node with the lowest overall load score in this second node set. Figure 4 As shown.

[0089] S10. Construct a unified global resource visualization view and provide global resource view services;

[0090] By connecting and integrating the physical and logical resource information managed by the resource management service, the standard and specification data generated by the data governance service, and the decision-making and execution results generated by the intelligent scheduling service, this invention integrates and visualizes the above-mentioned multi-source heterogeneous data to build a unified global resource visualization view, providing intuitive global management and control support for operation and maintenance personnel.

[0091] The comprehensive resource visualization view supports dynamic data analysis and drill-down queries from multiple dimensions such as resource type, physical location, logical relationship, operating status, and load level. It can monitor and visualize abnormal resource status, performance bottlenecks, and policy conflicts in the plant's comprehensive distributed monitoring system in real time, providing comprehensive, accurate, and intuitive data support for resource scheduling decisions.

[0092] All services interact using RPC communication protocols that meet the safety requirements of substation zones I to IV. This invention enables rational service orchestration and dynamic scheduling, improving system resource utilization, resilience, and reliability, and ensuring the continuous operation of critical services in the substation's distributed monitoring system.

[0093] like Figure 5 As shown, the plant-wide distributed monitoring system resource management system of the present invention includes:

[0094] The resource management module is used to set the sampling period, collect resource usage data of each original service in the plant-wide distributed monitoring system according to the sampling period, and store it in the time series database; monitor the load fluctuation of resources in the plant-wide distributed monitoring system, set the sliding window time, calculate the load fluctuation amplitude of each resource within the sliding window time; and adjust the collection period according to the load fluctuation amplitude of each resource and the priority of each original service.

[0095] The data governance module is used to set up a statistics window, calculate the quality and integrity of resource usage data in the time series database within the statistics window, determine whether the resource usage data meets the preset standards based on the quality and integrity of the resource usage data in the time series database and the priority of the original business, and if it does not meet the preset standards, perform secondary optimization processing on the resource usage data that does not meet the preset standards, and store the processed resource usage data in the time series database.

[0096] The intelligent scheduling module is used to obtain resource usage data of each original service in each node of the plant's full-domain distributed monitoring system from the time-series database, and calculate the comprehensive load score of each node by weighting it; and adjust the comprehensive load score weight combination of each node according to the business scenario and the preset comprehensive load score weight combination of each business scenario, and update the comprehensive load score of each node; calculate the expected resource usage data of the new service, and obtain all nodes that meet the preset screening conditions based on the resource usage data of each original service in the node and the expected resource usage data of the new service, and deploy the new service to the node with the lowest comprehensive load score;

[0097] The global resource view display module is used to build a unified global resource visualization view and provide a global resource view service.

Claims

1. A resource management method for a plant-wide distributed monitoring system, characterized in that, include: Set the sampling period, collect resource usage data of each original business in the plant's full-domain distributed monitoring system according to the sampling period, and store it in the time series database; Monitor the load fluctuations of resources in the plant's distributed monitoring system, set a sliding window time, and calculate the load fluctuation amplitude of each resource within the sliding window time; adjust the collection cycle according to the load fluctuation amplitude of each resource and the priority of each original service. Set up a statistics window to calculate the quality and integrity of resource usage data in the time series database. Based on the quality and integrity of the resource usage data in the time series database and the priority of the original business, determine whether the resource usage data meets the preset standards. If it does not meet the preset standards, perform secondary optimization processing on the resource usage data that does not meet the preset standards, and store the processed resource usage data in the time series database. Resource usage data of each original business in each node of the plant-wide distributed monitoring system are obtained from the time series database, and the comprehensive load score of each node is calculated by weighting them. The comprehensive load score weight combination of each node is adjusted according to the business scenario and the preset comprehensive load score weight combination of each business scenario, and the comprehensive load score of each node is updated. Calculate the expected resource usage data for the new service. Based on the resource usage data of each original service in the node and the expected resource usage data of the new service, obtain all nodes that meet the preset screening conditions, and deploy the new service to the node with the lowest overall load score. Build a unified global resource visualization view to provide a global resource view service.

2. The resource management method for a plant-wide distributed monitoring system according to claim 1, characterized in that, The resources include CPU, memory, I / O, database, middleware, and business services; the resource usage data includes CPU utilization, memory usage, I / O traffic, database load, middleware load, and average communication latency of business services; the priorities of the services include high priority and low priority.

3. The resource management method for a plant-wide distributed monitoring system according to claim 2, characterized in that, The formula for calculating the load fluctuation of each resource within the sliding window time is as follows: ； in, This represents the sampled payload value within the sliding window time t. This represents the peak value of the payload samples within the sliding window time t. This represents the valley value among the payload samples within the sliding window time t. This represents the average value of the payload samples within the sliding window time t; The adjustment of the data collection cycle based on the load fluctuation of each resource and the priority of each original service includes: When the resource load fluctuation range within the sliding window time t is greater than 10% and the original service is a preset high-priority service, the collection cycle is adjusted to 1 second; when the resource load fluctuation range is less than or equal to 10% or the original service is a preset low-priority service, the collection cycle is adjusted to 5 seconds.

4. The resource management method for a plant-wide distributed monitoring system according to claim 1, characterized in that, The formula for calculating the quality and integrity of resource usage data in the statistical window within the time-series database is: ； in, This indicates the number of valid data entries within the statistics window. This indicates the total number of data entries that should theoretically be collected within the statistics window.

5. The resource management method for a plant-wide distributed monitoring system according to claim 2, characterized in that, The preset standard is that if the resource usage data is for low-priority services, the quality integrity of the resource usage data must be greater than 95%; if the resource usage data is for high-priority services, the quality integrity of the resource usage data must be greater than 98%.

6. The resource management method for a plant-wide distributed monitoring system according to claim 1, characterized in that, The secondary optimization process includes: If the resource usage data exceeds the preset operating threshold, the system is in a high-load scenario; otherwise, the system is in a low-load scenario. If the system is under high load and resource usage data fluctuates or suddenly increases, the Long Short-Term Memory Network (LSTM) temporal prediction and repair module is activated to repair the resource usage data. The LSTM temporal prediction and repair module is based on historical continuous time-series running resource data and learns the temporal change pattern of running resource indicators through an LSTM network to predict, complete, and repair instantaneous missing or abnormal data under high load scenarios. If the system is in a low-load scenario and resource usage data is missing, the historical data association and repair module will be activated to repair the resource usage data. The historical data association and repair module, based on the historical running resource database, matches historical similar data corresponding to the current collection time in the low-load scenario, and completes the missing data repair through association deduction.

7. The resource management method for a plant-wide distributed monitoring system according to claim 2, characterized in that, The calculation of the overall load score for each node includes: Retrieve CPU and memory utilization rates for each node from the time-series database, and count the number of high-priority services deployed on each node. Calculate the percentage of high-priority services deployed on each node. Based on the current CPU and memory utilization rates and the percentage of high-priority services deployed on each node, calculate the weighted overall load score F for each node using the following formula: ； Where C is the CPU utilization of each node; M is the memory utilization of each node; P is the number of high-priority services deployed on each node; and N is the upper limit of high-priority services that can be deployed on a single node. The percentage of high-priority services already deployed on each node; Weights representing CPU utilization; Weights representing memory usage; This indicates the weight of the percentage of high-priority services deployed on each node, and F represents the overall load score.

8. The resource management method for a plant-wide distributed monitoring system according to claim 7, characterized in that, The adjustment of the comprehensive load score weight combination of each node according to the business scenario and the preset comprehensive load score weight combination for each business scenario includes: If the business scenario requires prioritizing the stability of high-priority services, configure... =0.3、 =0.2、 =0.5; If the business scenario requires prioritizing resource utilization, configure... =0.4、 =0.4、 =0.2; If the business scenario requires a balance between stability and resource utilization, configure =0.25、 =0.25、 =0.

5.

9. The resource management method for a plant-wide distributed monitoring system according to claim 8, characterized in that, The process of obtaining all nodes that meet the preset filtering conditions and deploying new services to the node with the lowest overall load score includes: First, select all nodes that simultaneously satisfy both condition one and condition two to form the first node set; Condition 1 is that the number of high-priority services deployed in the node is less than the preset service deployment limit; Condition 2 is that the sum of the number of CPU cores used by the services already deployed in the node and the number of CPU cores required by the new services is less than the preset CPU core count threshold of the node. If the new service is not within the preset high-priority service scope, the new service will be deployed on the node with the lowest overall load score in the first node set; If the new service falls within the preset high-priority service range, then define CPU reservation thresholds and memory reservation thresholds for the new service. CPU reservation threshold = base threshold - peak CPU utilization of deployed high-priority services; Memory reservation threshold = base threshold - peak memory utilization of deployed high-priority services; Nodes that meet the requirements of having a CPU utilization rate lower than the CPU reservation threshold and a memory utilization rate lower than the memory reservation threshold after deploying new services are selected from the first node set to form a second node set. New services within the preset high-priority service scope will be deployed on the node with the lowest overall load score in the second node set.

10. A resource management system for a plant-wide distributed monitoring system, characterized in that, include: The resource management module is used to set the sampling period, collect resource usage data of each original service in the plant-wide distributed monitoring system according to the sampling period, and store it in the time series database; monitor the load fluctuation of resources in the plant-wide distributed monitoring system, set the sliding window time, calculate the load fluctuation amplitude of each resource within the sliding window time; and adjust the collection period according to the load fluctuation amplitude of each resource and the priority of each original service. The data governance module is used to set up a statistics window, calculate the quality and integrity of resource usage data in the time series database within the statistics window, determine whether the resource usage data meets the preset standards based on the quality and integrity of the resource usage data in the time series database and the priority of the original business, and if it does not meet the preset standards, perform secondary optimization processing on the resource usage data that does not meet the preset standards, and store the processed resource usage data in the time series database. The intelligent scheduling module is used to obtain resource usage data of each original service in each node of the plant's full-domain distributed monitoring system from the time-series database, and calculate the comprehensive load score of each node by weighting it; and adjust the comprehensive load score weight combination of each node according to the business scenario and the preset comprehensive load score weight combination of each business scenario, and update the comprehensive load score of each node; calculate the expected resource usage data of the new service, and obtain all nodes that meet the preset screening conditions based on the resource usage data of each original service in the node and the expected resource usage data of the new service, and deploy the new service to the node with the lowest comprehensive load score; The global resource view display module is used to build a unified global resource visualization view and provide a global resource view service.

Images