A multi-device collaborative intelligent monitoring method and platform for data centers
By randomly selecting tasks from the data center for verification, abnormalities in device operation logic and performance are identified, and health weights are adjusted. This resolves erroneous calculations caused by device hardware failures and improves the accuracy and reliability of the monitoring system.
Patent Information
- Application Number
- CN202611277894.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-21
- Publication Date
- 2026-11-17
AI Technical Summary
Data centers cannot detect erroneous calculation results in a timely manner when equipment hardware fails, affecting the accuracy and reliability of the monitoring system.
By acquiring a set of monitoring tasks, a portion of the tasks are randomly selected as verification tasks and assigned to at least two devices for execution. The calculation results are collected and compared to determine if the device's calculation logic and performance are abnormal. The device's health weight is then adjusted and the task allocation is reduced.
It enables proactive verification of the device's operational logic, preventing erroneous calculation results from being mixed in, improving the accuracy and reliability of the monitoring system, and realizing the quantitative assessment of the device's health status.
Smart Images

Figure CN123123387A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of equipment monitoring technology, and in particular relates to a method and platform for intelligent monitoring of data centers with multi-device collaboration. Background Technology
[0002] Currently, in multi-device collaborative monitoring systems, data centers only perform simple aggregation, storage, or display of the returned results. At the same time, they passively monitor basic indicators such as CPU utilization, memory usage, and network bandwidth of devices through independently deployed performance monitoring components. When high device load or network timeout is detected, maintenance personnel manually adjust task allocation parameters or temporarily remove abnormal nodes.
[0003] In existing technologies, data centers rely solely on load balancing algorithms for task allocation, lacking an active verification mechanism for the correctness of device calculation results. When a device produces erroneous calculation results due to hardware failure, the data center cannot detect it in time, seriously affecting the accuracy and reliability of the monitoring system. Summary of the Invention
[0004] The purpose of this invention is to provide a multi-device collaborative intelligent monitoring method for data centers, aiming to solve the problem in the prior art where, when a device produces an erroneous calculation result due to hardware failure, the data center cannot detect it in time, which seriously affects the accuracy and reliability of the monitoring system.
[0005] This invention is implemented as follows: a multi-device collaborative intelligent monitoring method for data centers, the method comprising: Obtain the set of monitoring tasks to be processed, randomly select a portion of the tasks from the set as verification tasks, and assign the verification tasks to at least two devices for execution. Each device executes its assigned task and returns the calculation result data. Data transmission speed parameters and calculation time parameters of each device during task execution are collected and reported. Receive and compare the calculation result data returned by each device performing the same verification task. If the comparison results are inconsistent, it is determined that there is an abnormality in the calculation logic of the participating device. Compare the data transmission speed parameter and the calculation time parameter with the preset threshold. If any parameter is lower than the preset threshold, it is determined that the device has a performance abnormality. Based on the judgment results, when the target device is determined to have abnormal computing logic or performance, the health weight of the target device is reduced, and the data transmission allocation or computing load of the target device in subsequent task allocation is reduced according to the reduced health weight.
[0006] Preferably, the steps of each device executing its assigned task and returning the calculation result data, and collecting and reporting the data transmission speed parameters and calculation time parameters of each device during task execution, include: Receive task data packets, record the first timestamp when reception is complete, and parse the task data packets to obtain the content to be processed; The calculation result data is generated based on the content to be calculated, and the calculation time parameter is calculated in real time during the calculation process. During the data transmission process, the average transmission rate is calculated as the data transmission speed parameter. The calculation result data, calculation time parameters, data transmission speed parameters, and corresponding device and task identifiers are associated and encapsulated to generate a status reporting data packet, which is then sent to the data center.
[0007] Preferably, the steps of each device executing its assigned task and returning the calculation result data, and collecting and reporting the data transmission speed parameters and calculation time parameters of each device during task execution, include: Receive status report data packets and parse the calculation result data, data transmission speed parameters, calculation time parameters, device identifier and task identifier from the status report data packets; Based on the parsed task identifier, the system performs a horizontal comparison of multiple calculation result data belonging to the same verification task but corresponding to different device identifiers, and outputs the devices with abnormal calculation logic. At the same time, the system compares the data transmission speed parameter and the calculation time parameter with their respective preset thresholds to output the devices with abnormal performance. Devices exhibiting operational logic or performance abnormalities are marked as abnormal devices, and their device identifiers and corresponding abnormal types are recorded in the abnormal device list.
[0008] Preferably, the step of adjusting the health weight of the target device when it is determined that the target device has an operational logic abnormality or performance abnormality, and reducing the data transmission allocation or operational load of the target device in subsequent task allocation based on the adjusted health weight, includes: Read the device identifier and anomaly type of each abnormal device, determine the downward adjustment range of the health weight based on the anomaly type, adjust the weight, and obtain the updated health weight; Calculate the task quota ratio based on the updated health weight, and reduce the number of tasks assigned to the target device in subsequent task allocations according to the task quota ratio. In subsequent task allocation, tasks are assigned to each device according to the task quota ratio. After the health weight of the target device is restored to the normal threshold, the data transmission allocation and computing load of the target device are gradually increased to the preset normal level.
[0009] Preferably, the health weight is calculated as follows: in, For the health weight after the device update, This is the baseline health weight value for the device; For the device in the The base score for the corresponding anomaly type in the anomaly determination; For the first The time decay factor for each anomaly determination corresponds to an anomaly determination that is closer to the current time. The larger the value; For the device in the Performance rating after each successful task execution; For the first The reward coefficient for each normal execution.
[0010] Another objective of this invention is to provide a multi-device collaborative intelligent monitoring platform for data centers, the platform comprising: The task sampling module is used to acquire a set of monitoring tasks to be processed, randomly select a portion of the tasks from the monitoring task set as verification tasks, and assign the verification tasks to at least two devices for execution. The operation parameter acquisition module is used for each device to execute the assigned tasks and return the calculation result data. It also collects and reports the data transmission speed parameters and calculation time parameters of each device during the task execution process. The anomaly detection module is used to receive and compare the calculation result data returned by each device performing the same verification task. If the comparison results are inconsistent, it is determined that there is an anomaly in the calculation logic of the participating device. The data transmission speed parameter and the calculation time parameter are compared with preset thresholds. If any parameter is lower than the preset threshold, it is determined that the device has a performance anomaly. The weight adjustment module is used to adjust the health weight of a target device when it is determined that the target device has abnormal operation logic or performance based on the judgment result. Then, based on the adjusted health weight, the data transmission allocation or operation load of the target device in subsequent task allocation is reduced.
[0011] Preferably, the operating parameter acquisition module includes: The data parsing unit is used to receive task data packets, record the first timestamp when the reception is completed, and parse the task data packets to obtain the content to be processed. The transmission parameter calculation unit is used to generate calculation result data based on the content to be calculated, and to calculate the time consumption parameter in real time during the calculation process. During the data transmission process, it calculates the average transmission rate as the data transmission speed parameter. The data encapsulation unit is used to associate and encapsulate the calculation result data, calculation time parameters, data transmission speed parameters, and corresponding device identifiers and task identifiers, generate a status reporting data packet, and send the status reporting data packet to the data center.
[0012] Preferably, the anomaly detection module includes: The data extraction unit is used to receive status reporting data packets and parse the calculation result data, data transmission speed parameters, calculation time parameters, device identifier and task identifier from the status reporting data packets. The abnormal device determination unit is used to perform a horizontal comparison of multiple calculation result data belonging to the same verification task but corresponding to different device identifiers based on the parsed task identifier, and output the device with abnormal calculation logic; at the same time, it compares the data transmission speed parameter and the calculation time parameter with their respective preset thresholds to output the device with abnormal performance. The data recording unit is used to mark devices with abnormal operation logic or performance as abnormal devices, and record the device identifier of the abnormal device and its corresponding abnormal type in the abnormal device list.
[0013] Preferably, the weight adjustment module includes: The weight update unit is used to read the device identifier and anomaly type of each abnormal device, determine the downward adjustment range of the health weight based on the anomaly type, adjust the weight, and obtain the updated health weight. The task quota calculation unit is used to calculate the task quota ratio based on the updated health weight, and reduce the number of tasks assigned to the target device in subsequent task allocation according to the task quota ratio. The equipment monitoring unit is used to allocate a corresponding amount of tasks to each device according to the task quota ratio in subsequent task allocation. After the health weight of the target device recovers to the normal threshold, the data transmission allocation and computing load of the target device are gradually increased to the preset normal level.
[0014] Preferably, the health weight is calculated as follows: in, For the health weight after the device update, This is the baseline health weight value for the device; For the device in the The base score for the corresponding anomaly type in the anomaly determination; For the first The time decay factor for each anomaly determination corresponds to an anomaly determination that is closer to the current time. The larger the value; For the device in the Performance rating after each successful task execution; For the first The reward coefficient for each normal execution.
[0015] This invention achieves proactive verification of the correctness of device operation logic by comparing operation results, effectively avoiding the problem of erroneous operation results being mixed into the monitoring system due to single device hardware failure, software defects or data pollution, significantly improving the accuracy and reliability of multi-device collaborative monitoring; and realizing a comprehensive quantitative assessment of device health status. Attached Figure Description
[0016] Figure 1 A flowchart illustrating a multi-device collaborative intelligent monitoring method for data centers, provided as an embodiment of the present invention; Figure 2 This is an architecture diagram of a multi-device collaborative intelligent monitoring system for data centers, provided as an embodiment of the present invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0018] like Figure 1 The diagram shown is a flowchart of a multi-device collaborative intelligent monitoring method for data centers provided by an embodiment of the present invention. The method includes: S100: Obtain the set of monitoring tasks to be processed, randomly select a portion of the tasks from the set as verification tasks, and assign the verification tasks to at least two devices for execution.
[0019] In this step, a set of monitoring tasks to be processed is obtained through the data center. The monitoring task set contains multiple independent monitoring tasks, each with a unique task identifier. The monitoring tasks specifically include tasks for collecting device operating status, reading environmental parameters, analyzing image frames, and parsing log data. After obtaining the monitoring task set, a portion of the tasks are randomly selected from the set as verification tasks according to a preset sampling ratio. The remaining tasks that are not selected are considered regular tasks. The sampling ratio is set to 5%~20%, and is dynamically adjusted according to the scale of connected devices. The more devices connected, the lower the sampling ratio. For tasks marked as verification tasks, the data center synchronously distributes the task to at least two devices for parallel execution. That is, the same verification task is distributed to two or more devices for independent computation. For regular tasks, the data center distributes them to a single device for execution according to a preset load balancing strategy. When allocating tasks, the data center allocates them according to the real-time health weight of each device. Devices with higher health weights receive more task quotas.
[0020] S200: Each device executes its assigned task and returns the calculation result data. It also collects and reports the data transmission speed parameters and calculation time parameters of each device during the task execution process.
[0021] In this step, after receiving the task data packet from the data center, each device first records the first timestamp when the reception is completed. Then, it parses the task data packet to obtain the content to be calculated, performs the corresponding calculation based on the content to be calculated, and generates the calculation result data. During the execution of the task, the device records the start time of the calculation as the second timestamp and the end time of the calculation as the third timestamp. The calculation time parameter is calculated based on the second and third timestamps. During the data transmission process, the device collects the current instantaneous data transmission rate according to the preset sampling period and calculates the average transmission rate as the data transmission speed parameter based on the instantaneous transmission rate at each sampling time. Each device associates and encapsulates the calculation result data, the calculation time parameter, the data transmission speed parameter, and the corresponding device identifier and task identifier to generate a status reporting data packet, and sends the status reporting data packet to the data center.
[0022] S300 receives and compares the calculation result data returned by each device performing the same verification task. If the comparison results are inconsistent, it is determined that the device involved has an abnormal calculation logic. The data transmission speed parameter and the calculation time parameter are compared with preset thresholds. If any parameter is lower than the preset threshold, it is determined that the device has a performance abnormality.
[0023] In this step, the data center receives status report data packets uploaded by each device, which include calculation result data, data transmission speed parameters, calculation time parameters, device identifier, and task identifier. Based on the parsed task identifier, the data center performs a horizontal comparison of multiple calculation result data belonging to the same verification task but corresponding to different device identifiers. If the multiple calculation result data are completely consistent, it is determined that the calculation logic of each device participating in the verification task is normal; if there are discrepancies between the multiple calculation result data and other data, it is determined that the device has an abnormal calculation logic. Meanwhile, the data center compares the data transmission speed parameters reported by each device with the preset speed threshold, and compares the computation time parameters reported by each device with the preset time threshold. If the transmission speed parameter of any device is lower than the preset speed threshold or the computation time parameter is higher than the preset time threshold, the device is determined to have a performance abnormality. Devices with computational logic abnormalities or performance abnormalities are marked as abnormal devices, and the device identifier of the abnormal device and its corresponding abnormality type are recorded in the abnormal device list.
[0024] S400: Based on the judgment result, when it is determined that the target device has abnormal operation logic or performance, the health weight of the target device is reduced, and the data transmission allocation or operation load of the target device in subsequent task allocation is reduced according to the reduced health weight.
[0025] In this step, the device identifier and corresponding exception type of each abnormal device are read from the list of abnormal devices. The reduction range of the health weight is determined according to the exception type. The reduction range for the operation logic exception is greater than that for the performance exception. The data center subtracts the corresponding reduction range from the current health weight of the target device to obtain the updated health weight. Based on the updated health weight, the data center calculates the task quota ratio of the target device in subsequent task allocation. The lower the health weight, the smaller the task quota ratio. The data center allocates a corresponding amount of tasks to each device according to the task quota ratio. When the abnormal device continuously executes tasks normally for a preset number of times in subsequent execution, the data center gradually increases the data transmission allocation and operation load of the device to the normal level, realizing the automatic recovery of the device.
[0026] In a preferred embodiment of the present invention, the steps of each device executing its assigned task and returning computation result data, and collecting and reporting data transmission speed parameters and computation time parameters of each device during task execution, include: S201: Receive the task data packet, record the first timestamp when the reception is completed, and parse the task data packet to obtain the content to be processed.
[0027] In this step, after the data center sends a task data packet to the target device, the device receives the task data packet. The moment the task data packet is completely received is recorded as the first timestamp. Taking the edge node with device number NODE-07 as an example, the task data packet received by this device contains information such as task identifier TASK-1043, task type label "temperature anomaly detection", data content to be processed, and check bits. The moment NODE-07 device receives the last data frame is the first timestamp. NODE-07 device unpacks and parses the task data packet, verifies the data integrity through hash check, and reads the task identifier, task type, and data content to be processed from the data packet. The data content to be processed is the temperature sensor's collected value. The device caches the task identifier and task type locally.
[0028] S202 generates calculation result data based on the content to be calculated, and calculates the calculation time parameter in real time during the calculation process. During the data transmission process, the average transmission rate is calculated as the data transmission speed parameter.
[0029] In this step, after parsing the task data packet, the device begins computation, recording the start time of the computation as the second timestamp. When the computation is completed, the current time is recorded as the third timestamp. The computation time parameter of the task is calculated based on the difference between the third and second timestamps. Taking the NODE-07 device performing a temperature anomaly detection task as an example, after parsing, the temperature anomaly detection algorithm is started at 14:32:18.410. The temperature data points in the computation content are processed one by one, and threshold comparison logic, sliding window smoothing filtering, and anomaly score calculation are performed. The device completes all computations at 14:32:18.50 (third timestamp), so the computation time parameter is 90 milliseconds. After the computation is completed, the result data feedback process is started, and the computation result data is reported to the data center. During the data feedback process, the instantaneous data transmission rate of the current network is collected in real time, and the average value is calculated to obtain the data transmission speed parameter.
[0030] S203 associates and encapsulates the calculation result data, calculation time parameters, data transmission speed parameters, and corresponding device identifiers and task identifiers to generate a status reporting data packet, and sends the status reporting data packet to the data center.
[0031] In this step, the device reads the task identifier of the current task and the device identifier of the device itself, associates them with the calculation result data, calculation time parameters and data transmission speed parameters, and encapsulates them into a status reporting data packet. After the status reporting data packet is assembled, the device sends the status reporting data packet to the receiving port designated by the data center. After receiving the status reporting data packets reported by each device, the data center classifies and collects the data according to the task identifier and device identifier in the data packet.
[0032] In a preferred embodiment of the present invention, the steps of each device executing its assigned task and returning computation result data, and collecting and reporting data transmission speed parameters and computation time parameters of each device during task execution, include: S301: Receive status report data packet, and parse the calculation result data, data transmission speed parameters, calculation time parameters, device identifier, and task identifier from the status report data packet.
[0033] In this step, the data center continuously receives status report data packets from each device. After verifying the integrity of the received data packets, it extracts the calculation result data, data transmission speed parameters, calculation time parameters, device identifier, and task identifier from the data packets according to a predefined communication protocol. For example, when the data center receives a data packet reported by NODE-07, it parses out the device identifier as "NODE-07", the task identifier as "TASK-1043", the calculation result data as a JSON-formatted temperature anomaly determination result, the calculation time parameter as 90 milliseconds, and the data transmission speed parameter as 2.325 MB / s. Similarly, when the data center subsequently receives a data packet reported by NODE-12 with the same task identifier "TASK-1043", it parses out the device identifier as "NODE-12", the calculation result data as no anomaly detected, the calculation time parameter as 45 milliseconds, and the data transmission speed parameter as 4.8 MB / s.
[0034] S302, based on the parsed task identifier, performs a horizontal comparison of multiple calculation result data belonging to the same verification task but corresponding to different device identifiers, and outputs the device with calculation logic abnormality; at the same time, it compares the data transmission speed parameter and calculation time parameter with their respective preset thresholds to output the device with performance abnormality.
[0035] In this step, the data center groups and aggregates all reported data. For data groups with multiple device identifiers under the same task identifier, it reads all the calculation result data belonging to that task and performs consistency judgment. For example, for task TASK-1043, the data center reads the calculation result data of NODE-07, with the anomaly flag bit true, and the calculation result data of the abnormal sensor SEN-07 and NODE-12, with the anomaly flag bit false. After comparing the two field by field, the results are found to be inconsistent. Based on this, it is determined that there is a difference in the calculation result of NODE-07, and NODE-07 is output as a device with an abnormal calculation logic. Meanwhile, the data center independently performs performance anomaly detection: it compares the data transmission speed parameters of each device with preset speed thresholds, and compares the computation time parameters with preset time thresholds. For example, if the data transmission speed parameter of NODE-07 is 2.325MB / s, which is higher than 1.5MB / s, and the computation time parameter is 90 milliseconds, which is lower than 100 milliseconds, then the performance dimension is judged as normal; if the data transmission speed parameter of a device is lower than 1.5MB / s or its computation time parameter is higher than 100 milliseconds, then the device is judged to have performance anomalies.
[0036] S303, mark the device with abnormal operation logic or performance as abnormal device, and record the device identifier of the abnormal device and its corresponding abnormal type in the abnormal device list.
[0037] In this step, the data center categorizes all devices with computational logic anomalies or performance anomalies as abnormal devices. For example, NODE-07 is determined to have a computational logic anomaly because its computational result data is inconsistent with NODE-12. The data center adds a new record to the abnormal device list, which includes the device identifier NODE-07 and the anomaly type - computational logic anomaly. If the same device also has performance anomalies, a performance anomaly tag is added to the anomaly type field. This abnormal device list is dynamically updated after each round of monitoring tasks is completed, and the updated abnormal device list is used as input data for subsequent health weight adjustment steps.
[0038] As a preferred embodiment of the present invention, the step of adjusting the health weight of the target device when it is determined that the target device has an operational logic abnormality or performance abnormality, and reducing the data transmission allocation or operational load of the target device in subsequent task allocation based on the adjusted health weight, includes: S401: Read the device identifier and anomaly type of each abnormal device, determine the downward adjustment range of the health weight based on the anomaly type, adjust the weight, and obtain the updated health weight.
[0039] In this step, The data center reads the device identifier and corresponding anomaly type of each abnormal device from the list of abnormal devices, determines the downward adjustment of the health weight based on the anomaly type, and determines the calculation result error due to abnormal calculation logic. The calculation method of the health weight is as follows: in, For the health weight after the device update, This is the baseline health weight value for the device; For the device in the The base score for the corresponding anomaly type in the anomaly determination; For the first The time decay factor for each anomaly determination corresponds to an anomaly determination that is closer to the current time. The larger the value; For the device in the Performance rating after each successful task execution; For the first The reward coefficient for each successful execution; Taking NODE-07 as an example, this device is marked as having "abnormal computational logic" due to its computational results. Assuming this aberration occurred during the most recent task execution, the time decay factor... The value is 1, representing the baseline health weight. The base score is 100 points. If the score is 20, and the device has no other abnormal records or normal reward accumulation prior to this round, then the health weight updated by NODE-07 will be... Points. If NODE-07 is also marked as having performance anomalies, the corresponding performance anomaly points will be deducted, i.e. If a device is judged to have a performance abnormality solely due to low transmission speed without any logical errors, then... point.
[0040] S402, calculate the task quota ratio based on the updated health weight, and reduce the tasks assigned to the target device in subsequent task allocations according to the task quota ratio.
[0041] In this step, the task quota ratio for each device in subsequent task allocation is calculated based on its updated health weight. The data center determines the task quota ratio according to a preset tiered scheduling strategy: when the health weight is greater than or equal to 80 points, the device maintains a full task quota; when the health weight is between 60 and 80 points, the device reduces its task quota by a first ratio; when the health weight is between 40 and 60 points, the device reduces its task quota by a second ratio, which is greater than the first ratio; when the health weight is below 40 points, the device's task quota is reset to zero, and new task allocation is suspended. Taking NODE-07 as an example, its updated health weight is 80 points, which is at the full threshold. The data center sets its task quota ratio to 80% of the full quota. Device NODE-05's health weight drops to 55 points due to three consecutive performance anomalies, falling into the 40-60 point range. Therefore, its task quota ratio is reduced to 40% of the full quota. In this way, the lower the health weight, the smaller the task quota ratio, and the fewer tasks the target device is allocated in subsequent task allocation.
[0042] S403, in subsequent task allocation, allocate a corresponding amount of tasks to each device according to the task quota ratio. After the health weight of the target device is restored to the normal threshold, gradually increase the data transmission allocation and computing load of the target device to the preset normal level.
[0043] In this step, the adjusted task quota ratio is distributed, and tasks are allocated to each device according to its assigned quota. For example, if NODE-07's current task quota ratio is 80% of the full quota, then in the next round of allocation of 100 monitoring tasks, NODE-07 will be allocated 80% of the tasks it should have been allocated according to its original weight. The remaining tasks will be shared by other healthy devices. Once NODE-07 has executed tasks normally for a preset number of consecutive times during subsequent executions, the data center will gradually increase the data transmission and computing load of the device until it returns to normal levels. Specifically, for every 5 consecutive rounds of normal task execution, the health weight increases by 5 points, and the task quota ratio is increased by one level, until the health weight rises above the normal threshold, and the device returns to its full task quota. If the device experiences an anomaly again during the recovery process, the adjustment process will be restarted.
[0044] like Figure 2 As shown in the figure, a multi-device collaborative intelligent monitoring platform for data centers is provided in an embodiment of the present invention. The platform includes: The task sampling module 100 is used to acquire a set of monitoring tasks to be processed, randomly select a portion of the tasks from the monitoring task set as verification tasks, and assign the verification tasks to at least two devices for execution.
[0045] The operation parameter acquisition module 200 is used for each device to execute the assigned tasks and return the calculation result data, and to collect and report the data transmission speed parameters and calculation time parameters of each device during the task execution process.
[0046] In this platform, the parameter acquisition module 200 includes: The data parsing unit is used to receive task data packets, record the first timestamp when the reception is completed, and parse the task data packets to obtain the content to be processed. The transmission parameter calculation unit is used to generate calculation result data based on the content to be calculated, and to calculate the time consumption parameter in real time during the calculation process. During the data transmission process, it calculates the average transmission rate as the data transmission speed parameter. The data encapsulation unit is used to associate and encapsulate the calculation result data, calculation time parameters, data transmission speed parameters, and corresponding device identifiers and task identifiers, generate a status reporting data packet, and send the status reporting data packet to the data center.
[0047] The anomaly identification module 300 is used to receive and compare the calculation result data returned by each device performing the same verification task. If the comparison results are inconsistent, it is determined that the device has an abnormal calculation logic. The data transmission speed parameter and the calculation time parameter are compared with preset thresholds. If any parameter is lower than the preset threshold, it is determined that the device has a performance abnormality.
[0048] In this platform, the anomaly detection module 300 includes: The data extraction unit is used to receive status reporting data packets and parse the calculation result data, data transmission speed parameters, calculation time parameters, device identifier and task identifier from the status reporting data packets. The abnormal device determination unit is used to perform a horizontal comparison of multiple calculation result data belonging to the same verification task but corresponding to different device identifiers based on the parsed task identifier, and output the device with abnormal calculation logic; at the same time, it compares the data transmission speed parameter and the calculation time parameter with their respective preset thresholds to output the device with abnormal performance. The data recording unit is used to mark devices with abnormal operation logic or performance as abnormal devices, and record the device identifier of the abnormal device and its corresponding abnormal type in the abnormal device list.
[0049] The weight adjustment module 400 is used to adjust the health weight of the target device when it is determined that the target device has abnormal operation logic or performance based on the judgment result, and reduce the data transmission allocation or operation load of the target device in subsequent task allocation based on the adjusted health weight.
[0050] In this platform, the weight adjustment module 400 includes: The weight update unit is used to read the device identifier and anomaly type of each abnormal device, determine the downward adjustment range of the health weight based on the anomaly type, adjust the weight, and obtain the updated health weight. The task quota calculation unit is used to calculate the task quota ratio based on the updated health weight, and reduce the number of tasks assigned to the target device in subsequent task allocation according to the task quota ratio. The equipment monitoring unit is used to allocate a corresponding amount of tasks to each device according to the task quota ratio in subsequent task allocation. After the health weight of the target device recovers to the normal threshold, the data transmission allocation and computing load of the target device are gradually increased to the preset normal level.
[0051] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A multi-device collaborative intelligent monitoring method for data centers, characterized in that, The method includes: Obtain the set of monitoring tasks to be processed, randomly select a portion of the tasks from the set as verification tasks, and assign the verification tasks to at least two devices for execution. Each device executes its assigned task and returns the calculation result data. Data transmission speed parameters and calculation time parameters of each device during task execution are collected and reported. Receive and compare the calculation result data returned by each device performing the same verification task. If the comparison results are inconsistent, it is determined that there is an abnormality in the calculation logic of the participating device. Compare the data transmission speed parameter and the calculation time parameter with the preset threshold. If any parameter is lower than the preset threshold, it is determined that the device has a performance abnormality. Based on the judgment results, when the target device is determined to have abnormal computing logic or performance, the health weight of the target device is reduced, and the data transmission allocation or computing load of the target device in subsequent task allocation is reduced according to the reduced health weight.
2. The intelligent monitoring method for data centers with multi-device collaboration according to claim 1, characterized in that, The steps of each device executing its assigned task and returning the calculation result data, and collecting and reporting the data transmission speed parameters and calculation time parameters of each device during task execution, include: Receive task data packets, record the first timestamp when reception is complete, and parse the task data packets to obtain the content to be processed; The calculation result data is generated based on the content to be calculated, and the calculation time parameter is calculated in real time during the calculation process. During the data transmission process, the average transmission rate is calculated as the data transmission speed parameter. The calculation result data, calculation time parameters, data transmission speed parameters, and corresponding device and task identifiers are associated and encapsulated to generate a status reporting data packet, which is then sent to the data center.
3. The intelligent monitoring method for data centers with multi-device collaboration according to claim 1, characterized in that, The steps of each device executing its assigned task and returning the calculation result data, and collecting and reporting the data transmission speed parameters and calculation time parameters of each device during task execution, include: Receive status report data packets and parse the calculation result data, data transmission speed parameters, calculation time parameters, device identifier and task identifier from the status report data packets; Based on the parsed task identifier, the system performs a horizontal comparison of multiple calculation result data belonging to the same verification task but corresponding to different device identifiers, and outputs the devices with abnormal calculation logic. At the same time, the system compares the data transmission speed parameter and the calculation time parameter with their respective preset thresholds to output the devices with abnormal performance. Devices exhibiting operational logic or performance abnormalities are marked as abnormal devices, and their device identifiers and corresponding abnormal types are recorded in the abnormal device list.
4. The intelligent monitoring method for data centers with multi-device collaboration according to claim 1, characterized in that, The step of adjusting the health weight of a target device when it is determined that the target device has a logical or performance abnormality, and then reducing the data transmission allocation or computational load of the target device in subsequent task allocation based on the adjusted health weight, includes: Read the device identifier and anomaly type of each abnormal device, determine the downward adjustment range of the health weight based on the anomaly type, adjust the weight, and obtain the updated health weight; Calculate the task quota ratio based on the updated health weight, and reduce the number of tasks assigned to the target device in subsequent task allocations according to the task quota ratio. In subsequent task allocation, tasks are assigned to each device according to the task quota ratio. After the health weight of the target device is restored to the normal threshold, the data transmission allocation and computing load of the target device are gradually increased to the preset normal level.
5. The intelligent monitoring method for data centers with multi-device collaboration according to claim 1, characterized in that, The health weight is calculated as follows: in, For the health weight after the device update, This is the baseline health weight value for the device; For the device in the The base score for the corresponding anomaly type in the anomaly determination; For the first The time decay factor for each anomaly determination corresponds to an anomaly determination that is closer to the current time. The larger the value; For the device in the Performance rating after each successful task execution; For the first The reward coefficient for each normal execution.
6. A multi-device collaborative intelligent monitoring platform for data centers, characterized in that, The platform includes: The task sampling module is used to acquire a set of monitoring tasks to be processed, randomly select a portion of the tasks from the monitoring task set as verification tasks, and assign the verification tasks to at least two devices for execution. The operation parameter acquisition module is used for each device to execute the assigned tasks and return the calculation result data. It also collects and reports the data transmission speed parameters and calculation time parameters of each device during the task execution process. The anomaly detection module is used to receive and compare the calculation result data returned by each device performing the same verification task. If the comparison results are inconsistent, it is determined that there is an anomaly in the calculation logic of the participating device. The data transmission speed parameter and the calculation time parameter are compared with preset thresholds. If any parameter is lower than the preset threshold, it is determined that the device has a performance anomaly. The weight adjustment module is used to adjust the health weight of a target device when it is determined that the target device has abnormal operation logic or performance based on the judgment result. Then, based on the adjusted health weight, the data transmission allocation or operation load of the target device in subsequent task allocation is reduced.
7. The multi-device collaborative intelligent monitoring platform for data centers according to claim 6, characterized in that, The operating parameter acquisition module includes: The data parsing unit is used to receive task data packets, record the first timestamp when the reception is completed, and parse the task data packets to obtain the content to be processed. The transmission parameter calculation unit is used to generate calculation result data based on the content to be calculated, and to calculate the time consumption parameter in real time during the calculation process. During the data transmission process, it calculates the average transmission rate as the data transmission speed parameter. The data encapsulation unit is used to associate and encapsulate the calculation result data, calculation time parameters, data transmission speed parameters, and corresponding device identifiers and task identifiers, generate a status reporting data packet, and send the status reporting data packet to the data center.
8. The multi-device collaborative intelligent monitoring platform for data centers according to claim 6, characterized in that, The anomaly detection module includes: The data extraction unit is used to receive status reporting data packets and parse the calculation result data, data transmission speed parameters, calculation time parameters, device identifier and task identifier from the status reporting data packets. The abnormal device determination unit is used to perform a horizontal comparison of multiple calculation result data belonging to the same verification task but corresponding to different device identifiers based on the parsed task identifier, and output the device with abnormal calculation logic; at the same time, it compares the data transmission speed parameter and the calculation time parameter with their respective preset thresholds to output the device with abnormal performance. The data recording unit is used to mark devices with abnormal operation logic or performance as abnormal devices, and record the device identifier of the abnormal device and its corresponding abnormal type in the abnormal device list.
9. The multi-device collaborative intelligent monitoring platform for data centers according to claim 6, characterized in that, The weight adjustment module includes: The weight update unit is used to read the device identifier and anomaly type of each abnormal device, determine the downward adjustment range of the health weight based on the anomaly type, adjust the weight, and obtain the updated health weight. The task quota calculation unit is used to calculate the task quota ratio based on the updated health weight, and reduce the tasks assigned to the target device in subsequent task allocation according to the task quota ratio. The equipment monitoring unit is used to allocate a corresponding amount of tasks to each device according to the task quota ratio in subsequent task allocation. After the health weight of the target device recovers to the normal threshold, the data transmission allocation and computing load of the target device are gradually increased to the preset normal level.
10. The multi-device collaborative intelligent monitoring platform for data centers according to claim 6, characterized in that, The health weight is calculated as follows: in, For the health weight after the device update, This is the baseline health weight value for the device; For the device in the The base score for the corresponding anomaly type in the anomaly determination; For the first The time decay factor for each anomaly determination corresponds to an anomaly determination that is closer to the current time. The larger the value; For the device in the Performance rating after each successful task execution; For the first The reward coefficient for each normal execution.
