Equipment alarm threshold determination method and device, equipment and medium

Abstract

The invention provides an equipment alarm threshold determination method and device, equipment and a medium, relates to the technical field of computers, and can be applied to scenes such as cloud technology, artificial intelligence, intelligent traffic and auxiliary driving. Comprising the following steps: acquiring current unit task time consumption for executing a target task by target equipment, a plurality of time sequence-based operation index values corresponding to the current unit task time consumption and a current equipment alarm threshold value; performing duration anomaly detection on the current unit task time consumption to obtain a duration detection result; if the plurality of operation index values are lower than the current equipment alarm threshold value, determining a to-be-verified index value according to the duration detection result and the plurality of operation index values; taking the to-be-verified index value as a target simulation value of a simulator corresponding to the target equipment, performing simulation operation and index abnormity verification, and obtaining an abnormity verification result and a target index value; if the abnormal verification result is verification abnormity, updating a current equipment alarm threshold value according to the target index value; according to the invention, the timeliness and accuracy of threshold adjustment can be remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, apparatus, device and medium for determining a device alarm threshold. Background Art

[0002] In order to ensure the equipment operation safety, task execution efficiency and reasonable configuration of equipment computing power during task execution, it is necessary to set the alarm threshold of equipment indicators. The relevant technology usually determines the indicator value under the task abnormality state as the alarm threshold after task abnormalities such as execution failure or severe execution jam occur during operation, and adjusts the alarm threshold when the above task abnormality occurs again. However, the threshold configuration of this method is based on task abnormalities, and the adjustment is delayed and the accuracy is low, which is not conducive to the equipment operation quality and task research and development efficiency. Summary of the invention

[0003] The present application provides a method, apparatus, device and medium for determining a device alarm threshold, which can significantly improve the accuracy of threshold configuration and ensure the timeliness of threshold adjustment.

[0004] In one aspect, the present application provides a method for determining a device alarm threshold, the method comprising:

[0005] Obtaining the current unit task time consumed by the target device to execute the target task, multiple operation indicator values ​​based on timing corresponding to the current unit task time consumed, and the current device alarm threshold, wherein the target task includes a unit task that is executed cyclically, and the current unit task time consumed is the actual operation time of the unit task currently executed by the target device;

[0006] Performing duration anomaly detection on the time consumed by the current unit task to obtain a duration detection result;

[0007] If multiple operating indicator values ​​are lower than the current device alarm threshold, determine the indicator value to be verified according to the duration detection result and the multiple operating indicator values;

[0008] Taking the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, performing simulation operation of the target task, and performing indicator abnormality verification based on the simulation unit task time consumption of the simulator during the simulation operation to obtain the abnormality verification result and the target indicator value;

[0009] If the abnormal verification result is a verification abnormality, the current device alarm threshold is updated according to the target indicator value.

[0010] Another aspect provides a device for determining a device alarm threshold, the device comprising:

[0011] Acquisition module: used to acquire the current unit task time consumed by the target device to execute the target task, multiple operation index values ​​based on timing corresponding to the current unit task time consumed, and the current device alarm threshold, wherein the target task includes a unit task executed cyclically, and the current unit task time consumed is the actual operation time of the unit task currently executed by the target device;

[0012] Duration detection module: used to perform duration anomaly detection on the duration of the current unit task and obtain a duration detection result;

[0013] Indicator determination module: used to determine the indicator value to be verified according to the duration detection result and the multiple operating indicator values ​​if the multiple operating indicator values ​​are lower than the current device alarm threshold;

[0014] Abnormal verification module: used to use the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, perform simulation operation of the target task, and perform indicator abnormality verification based on the simulation unit task time of the simulator during the simulation operation to obtain abnormal verification results and target indicator value;

[0015] Threshold updating module: used to update the current device alarm threshold according to the target indicator value if the abnormal verification result is a verification abnormality.

[0016] On the other hand, a computer device is provided, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the device alarm threshold determination method as described above.

[0017] On the other hand, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored. The at least one instruction or the at least one program is loaded and executed by a processor to implement the device alarm threshold determination method as described above.

[0018] On the other hand, a server is provided, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the device alarm threshold determination method as described above.

[0019] On the other hand, a terminal is provided, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the device alarm threshold determination method as described above.

[0020] On the other hand, a computer program product or a computer program is provided, which includes computer instructions, and when the computer instructions are executed by a processor, the method for determining a device alarm threshold as described above is implemented.

[0021] The device alarm threshold determination method, apparatus, device, storage medium, server, terminal, computer program and computer program product provided in this application have the following technical effects:

[0022] The technical solution of the present application first obtains the current unit task time generated by the unit task of the target device executing the target task, multiple timing-based operating indicator values ​​corresponding to the current unit task time, and the current device alarm threshold; and performs a duration anomaly detection on the current unit task time to obtain a duration detection result to determine whether there is a time anomaly; compares multiple operating indicator values ​​with the current device alarm threshold, and if multiple operating indicator values ​​are lower than the current device alarm threshold, determines the indicator value to be verified based on the duration detection result and multiple operating indicator values; and then, uses the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, performs simulation operation of the target task, and performs verification based on the simulated unit task time of the simulator during the simulation operation. Perform indicator abnormality verification to obtain the abnormality verification result and the target indicator value; if the abnormality verification result is a verification abnormality, update the current equipment alarm threshold according to the target indicator value; in this way, the detection granularity of the unit task as the running time is combined with the time detection result to determine the possible abnormal indicator value to be verified, and then simulate the running of the target task and the simulation of the indicator value to be verified through the simulator to verify whether the indicator value to be verified may have task execution abnormalities, and then in the case of verification abnormalities, the alarm threshold is updated in time to improve the accuracy and reliability of threshold adjustment. There is no need to rely on the determined task abnormalities for threshold adjustment, so that the adjustment operation is advanced and adapted to business needs, which significantly improves the equipment resource operation quality, efficiency and R&D effectiveness of task execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 is a schematic diagram of an application environment provided by an embodiment of the present application;

[0025] Figure 2 It is a flow chart of a method for determining a device alarm threshold provided in an embodiment of the present application;

[0026] Figure 3 It is a flowchart of another method for determining a device alarm threshold provided in an embodiment of the present application;

[0027] Figure 4 It is a flowchart of another method for determining a device alarm threshold provided in an embodiment of the present application;

[0028] Figure 5 It is a flowchart of another method for determining a device alarm threshold provided in an embodiment of the present application;

[0029] Figure 6 It is a time curve diagram provided by an embodiment of the present application;

[0030] Figure 7 It is a temperature index value floating curve diagram provided in an embodiment of the present application;

[0031] Figure 8 It is a structural framework diagram of a device alarm threshold determination system provided in an embodiment of the present application;

[0032] Fig. 9 It is a schematic diagram of a framework of a device for determining a device alarm threshold provided by an embodiment of the present application;

[0033] Fig.10 It is a hardware structure block diagram of an electronic device for executing a method for determining a device alarm threshold provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0035] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server comprising a series of steps or submodules is not necessarily limited to those steps or submodules clearly listed, but may include other steps or submodules that are not clearly listed or inherent to these processes, methods, products or devices.

[0036] See also Figure 1 , Figure 1 is a schematic diagram of an application environment provided by an embodiment of the present application, such as Figure 1 As shown, the application environment may include at least a first server 01 and a second server 02. The first server 01 and the second server 02 may communicate through any network form, and this application does not impose any specific restrictions. The first server 01 and the second server 02 may include servers respectively. The server in the embodiment of this application may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0037] Specifically, cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or a local area network to achieve data computing, storage, processing, and sharing. Cloud technology can be applied to various fields, such as medical cloud, cloud Internet of Things, cloud security, cloud education, cloud conferencing, artificial intelligence cloud services, cloud applications, cloud calls, and cloud social networking. Cloud technology is based on the cloud computing business model application. It distributes computing tasks on a resource pool composed of a large number of computers, so that various application systems can obtain computing power, storage space, and information services as needed. The network that provides resources is called a "cloud". The resources in the "cloud" are infinitely expandable in the eyes of users, and can be obtained at any time, used on demand, expanded at any time, and paid for by use. As a basic capability provider for cloud computing, a cloud computing resource pool (referred to as a cloud platform, generally referred to as an IaaS (Infrastructure as a Service)) platform will be established, and various types of virtual resources will be deployed in the resource pool for external customers to choose to use. The cloud computing resource pool mainly includes: computing devices (virtualized machines, including operating systems), storage devices, and network devices.

[0038] According to the logical function division, the PaaS (Platform as a Service) layer can be deployed on the IaaS layer, and the SaaS (Software as a Service) layer can be deployed on the PaaS layer, or SaaS can be directly deployed on IaaS. PaaS is a platform for software operation, such as databases, web containers, etc. SaaS is a variety of business software, such as web portals, SMS mass senders, etc. Generally speaking, SaaS and PaaS are upper layers relative to IaaS.

[0039] Specifically, the server mentioned above may include a physical device, which may specifically include a network communication submodule, a processor, a memory, etc., and may also include software running in the physical device, which may specifically include an application program, etc.

[0040] In an embodiment of the present application, the first server end 01 can be a target device for running a target task, and the second server end 02 can be used to obtain the current unit task time consumed by the target device to execute the target task, multiple timing-based operating indicator values ​​corresponding to the current unit task time, and the current device alarm threshold, and perform abnormality verification and determine the target indicator value based on the acquired data to determine whether to update the current device alarm threshold with the target indicator value.

[0041] In one application scenario, the target task can be a model training task, and the unit task can be an iterative training of the model training, such as executing large-scale model training of an artificial intelligence model. The target device can run a GPU. Due to the large number of parameters in the model training task and the high scale of data exchange during the training process, the required GPU computing power will reach more than 10,000 cards. To run the training task on such a large-scale GPU computing power device, it is necessary to clearly define the device alarm thresholds of each indicator dimension to achieve the operability of the computing power. Understandably, the target task can also be other tasks that require cyclic execution of unit tasks.

[0042] In view of this, the present application proposes a device alarm threshold determination scheme, which can be triggered from the perspective of normalized operability of indicator items, realize anomaly detection and verification, and then adjust the threshold to ensure the accuracy of threshold determination and timeliness of adjustment, thereby ensuring the R&D efficiency and computing power operation quality of tasks such as model training.

[0043] Furthermore, it is understandable that Figure 1 What is shown is only an application environment of a method for determining a device alarm threshold value. The application environment may include more or fewer nodes, and the present application does not impose any limitation thereto.

[0044] The application environment involved in the embodiment of the present application, or the first server 01 and the second server 02 in the application environment, etc., can be a distributed system formed by connecting a client and multiple nodes (any form of computing devices in the access network, such as servers and user terminals) through network communication. The distributed system can be a blockchain system, which can provide the above-mentioned target task operation and device alarm threshold determination services.

[0045] The following is an introduction to the technical solution of this application based on the above application environment. The embodiments of this application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, assisted driving, etc. Please refer to Figure 2 , Figure 2 It is a flowchart of a method for determining a device alarm threshold provided in an embodiment of the present application. This specification provides method operation steps such as the embodiment or flowchart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one way of executing the steps among many orders, and does not represent the only order of execution. When the actual system or server product is executed, it can be executed in sequence or in parallel according to the method shown in the embodiment or the accompanying drawings (for example, in a parallel processor or multi-threaded processing environment). Specifically, as Figure 2 As shown, the method may include the following steps S201-S209:

[0046] S201: Obtaining the current unit task consumption of the target device to execute the target task, multiple time-series-based operating indicator values ​​corresponding to the current unit task consumption, and the current device alarm threshold.

[0047] Specifically, the target device can be a first server for running the target task, which can provide computing power or storage services. The target task includes a unit task that is executed cyclically, and the current unit task time is the actual running time of the unit task currently running on the target device; it can be understood that due to factors such as device resource usage and task operation differences, during the execution of the target task, the unit task time of each execution of the unit task is different. Accordingly, the actual running time consumed by the current execution of the unit task is collected in real time to obtain the above-mentioned current unit task time. Specifically, the multiple operation index values ​​based on timing corresponding to the current unit task time are the operation index values ​​of the target device collected at intervals during the current execution of the unit task. It can be understood that the target device can have multiple dimensions of operation indicators, such as temperature, video memory usage, memory usage, etc. Accordingly, during the execution of the current unit task, multiple dimensions of operation index values ​​can be collected, that is, multiple operation index values ​​based on timing corresponding to each indicator dimension corresponding to the current unit task are collected, and each indicator dimension corresponds to a current device alarm threshold. Specifically, the current device alarm threshold can be an alarm threshold determined by operations such as unit task consumption obtained before the execution of the current round of unit tasks, duration anomaly detection of operating indicator values, determination of indicator values ​​to be verified, simulation operation, and indicator anomaly verification, or in the initial state, when the target task is initially executed, the current device alarm threshold can be a set empirical value.

[0048] Exemplarily, multiple temperature index values ​​corresponding to the temperature index and multiple video memory usage index values ​​corresponding to the video memory usage index are collected, and the current device alarm threshold (temperature alarm threshold) corresponding to the temperature index and the current device alarm threshold (video memory usage alarm threshold) corresponding to the video memory usage index are obtained. It can be understood that in the case of multiple indicator dimensions, it is necessary to determine the indicator value to be verified, simulate the operation, verify the indicator abnormality, and determine the target indicator value for each indicator dimension in combination with the duration detection result and the current device alarm threshold corresponding to the indicator dimension, so as to update the current device alarm threshold corresponding to the indicator dimension.

[0049] In one embodiment, the target device is a server with an integrated GPU, the target task can be a model training task, the unit task is an iterative training in the model training process, and the single-step task duration refers to the actual running time to complete an iterative training, which is equivalent to the single-step duration of the model training. The indicator dimensions can include but are not limited to GPU card temperature, video memory usage, etc., and the specific information fields collected can be shown in Table 1 below.

[0050] Table 1

[0051] Time (minutes) Operation index value Single-step task time Time 1 GPU card temperature xxx seconds Time 2 Video memory usage xxx seconds

[0052] Furthermore, taking the temperature operation indicator as an example, the temperature can be set to be collected once every minute. In one iterative training cycle, multiple temperature indicator values ​​are usually collected, and each temperature indicator value is written into the time series database. For example, the field table information configuration written can be shown in Table 2 below. It can be understood that, referring to Figure 8 The real-time time series data such as the current multiple operating index values ​​collected and the historical time series data such as the historically collected operating index values ​​are stored in the time series database.

[0053] Table 2

[0054]

[0055] It can be understood that the indicator dimensions can be determined according to various business scenarios and are not limited to the above examples.

[0056] S203: Perform duration anomaly detection on the duration of the current unit task to obtain a duration detection result.

[0057] It is understandable that when the equipment resources are in a normal state, the time taken to execute a unit task should be within a reasonable range. Through the duration anomaly detection, it is possible to determine whether there are abnormalities such as delays during the execution of the unit task, so as to facilitate subsequent indicator detection and anomaly verification. The duration detection result is used to indicate whether there is a duration anomaly in the current unit task.

[0058] In some embodiments, in combination Figure 3, S203 may include S2031-S2035:

[0059] S2031: The task of obtaining multiple historical units takes time.

[0060] Specifically, there are no abnormalities in the time consumption of multiple historical unit tasks and they are continuous in time sequence with the time consumption of the current unit task; the time consumption of the historical unit task refers to the actual running time of the unit tasks of multiple historical cycles executed before the unit task of the current cycle is executed. For example, if the current cycle is n, the multiple historical cycles can be n-1, n-2, n-3, etc. In addition, the acquired historical unit task time consumption is data without abnormal duration, that is, when the time consumption is collected in real time for duration abnormality detection, no duration abnormality is detected. Exemplarily, the current cycle can be the current iterative training in the model training, and the time consumption of multiple historical unit tasks can refer to the unit task time consumption of each of the several rounds of iterative training before the current iterative training.

[0061] It should be noted that the number of acquired historical unit task time consumptions is greater than or equal to 2, and the specific number value can be determined based on actual needs or the time interval for data collection and storage.

[0062] Specifically, the collected data such as unit task duration, operating index values, etc. can be stored in a time series database. In some cases, a timed loop can be used to store data. There will be historical collected data between two data writes. Accordingly, the multiple historical unit task durations obtained above can be data collected in the interval between two data writes, or can also be historical data that has been stored.

[0063] S2032: Generate a duration curve based on the duration of multiple historical unit tasks and the duration of the current unit task.

[0064] Specifically, the horizontal axis of the duration curve can be time, and the vertical axis can be the unit task duration. It can be understood that the unit task duration of unit tasks executed in different cycles changes in real time, and the duration of multiple unit tasks adjacent in time series can form a duration curve with amplitude changes. Specifically, it is also possible to form a corresponding relationship between the duration curve and the corresponding operating indicator values ​​collected in the time series. Taking the temperature indicator dimension as an example, please refer to Figure 6 In the figure, the horizontal axis of the coordinate system is time, and the vertical axis height represents the size of the data index. The upper curve in the coordinate system is a duration curve formed by the time consumption of three adjacent unit tasks in time series (solid points), and the solid points on the right side of the curve are the current unit task time consumption; the bar chart below the curve represents the temperature index value, and the historical unit task time consumption and the current unit task time consumption on the left and middle correspond to the collection of 4, 3 and 5 temperature index values ​​respectively.

[0065] S2033: Determine the amplitude increment corresponding to the time consumption of the current unit task based on the duration curve.

[0066] Specifically, the amplitude increment refers to the increment between the current peak value corresponding to the current unit task time and the reference peak value, and the reference peak value is determined based on at least one historical peak corresponding to the current peak. In some embodiments, the wave where the current unit task time is located on the duration curve is determined, and then the peak value of the wave is determined as the current peak value, and the peak value of the previous wave adjacent to the wave corresponding to the current unit task time is determined as the reference peak value, such as Figure 6 In the example, the current unit task time is the current peak value, the historical unit task time on the left is the reference peak value, the difference between the two peaks is calculated to obtain the peak value increment, and the ratio between the peak value increment and the reference peak value is determined as the amplitude value increment. In other embodiments, when there are multiple historical peaks before the peak corresponding to the current unit task time, the multiple historical peak values ​​can be added and averaged to obtain the historical peak value.

[0067] S2034: If the amplitude increment is greater than or equal to the preset amplitude value, it is determined that the duration detection result is abnormal in the time consumption of the current unit task.

[0068] Specifically, the preset amplitude value can be set based on the actual needs of the target task, such as setting the preset amplitude value based on the task delay tolerance. The greater the delay tolerance, the greater the preset amplitude value, and vice versa. Exemplarily, the preset amplitude value can be 50%. When the amplitude value-added is higher than the preset amplitude value, it is determined that there is a time-consuming anomaly.

[0069] S2035: If the amplitude increment is less than the preset amplitude value, it is determined that the duration detection result is that the duration of the current unit task is normal.

[0070] Specifically, if the time consumption of the current unit task is normal, the time consumption of the current unit task can be used as the historical data of the time consumption of the next unit task.

[0071] Specifically, after the data is written into the time series database, refer to Figure 8 , the anomaly detection model of the real-time detection module can be input to perform the above-mentioned duration anomaly detection processing. If a time-consuming anomaly is detected, the current unit task duration of the anomaly and its corresponding collected operating indicator value are output and stored to facilitate subsequent threshold updates or alarm judgments.

[0072] like Figure 6 In the case where the time consumption of the current unit task is abnormal, the corresponding temperature index value can be packaged using the configuration shown in Table 3 below to be transmitted to the detection interface to trigger subsequent processing operations.

[0073] Table 3

[0074]

[0075] In this way, the duration curve is generated by combining the historical duration data, and the abnormal point is judged by combining the amplitude appreciation, so as to optimize the matching of abnormal duration detection and the time-consuming fluctuation of task execution, avoid the interference of time-consuming normal fluctuations, improve the accuracy of abnormal detection, and reduce the false detection rate.

[0076] S205: If the multiple operating indicator values ​​are lower than the current device alarm threshold, determine the indicator value to be verified according to the duration detection result and the multiple operating indicator values.

[0077] Specifically, multiple operating indicator values ​​are compared with the current device alarm threshold. If each operating indicator value is less than or equal to the current device alarm threshold, the alarm condition is not met temporarily, and the indicator value to be verified corresponding to the operating indicator value needs to be verified to determine whether the current device alarm threshold needs to be updated. If any one of them is higher than the current device alarm threshold, it indicates that there is a risk, and the process of determining whether to generate device alarm information is executed. It should be noted that the indicator value to be verified is less than the current device alarm threshold.

[0078] S207: Use the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, perform simulation operation of the target task, and perform indicator abnormality verification based on the simulation unit task time of the simulator during the simulation operation to obtain the abnormality verification result and the target indicator value.

[0079] Specifically, refer to Figure 8 ,The duration detection results output by the anomaly detection model, the ,abnormal current unit task time consumption and its corresponding collected ,operation index values ​​can be used as the input of the ,experimental verification module to perform threshold comparison, ,determination of the index values ​​to be verified, simulation operation and ,indicator anomaly verification, so as to obtain the anomaly verification results and ,target index values.

[0080] In some embodiments, the simulator running in the experimental verification module can be implemented based on the task data, floating-point matrix and heterogeneous computing card of the constructed target task. Specifically, the task data is based on floating-point matrix operations to simulate the target task and implements simulation operation based on heterogeneous computing cards to simulate the operation of the target task on the target device. In one embodiment, the target task is a model training task, and the task data is sample data of the constructed training task. The sample data usually uses floating-point matrix operations, and the generated sample data is stored in a database for continuous reuse to perform simulation of model training.

[0081] Specifically, the target simulation value is the value that the corresponding simulation index is expected to reach during the operation of the simulator. After obtaining the index value to be verified, it is input into the experimental verification module, and the simulator of the experimental verification module is called to simulate and run the target task based on the task data, and control its simulation running index value to gradually reach the index value to be verified, such as simulating and running on a heterogeneous computing card based on a floating-point matrix, and making the temperature of the heterogeneous computing card reach the temperature index value to be verified.

[0082] Furthermore, during the simulation operation, the simulation unit task time consumption of the simulation operation of each unit task of the simulator is detected, and whether there is an abnormality in the simulation unit task time consumption is combined with the verification of whether there is an abnormality, and an abnormality verification result is obtained. The abnormality verification result is used to indicate whether the proportion of time consumption abnormalities in the simulation unit task time consumption of each round of unit tasks reaches a preset proportion during the simulation operation in which the simulation operation index value gradually reaches the index value to be verified. If so, the abnormality is verified, and if not, there is no abnormality. Specifically, the target index value refers to the target simulation value corresponding to the state of the simulation unit task time consumption with a time consumption abnormality exceeding a preset proportion when the verification abnormality is determined. The target index value is less than or equal to the index value to be verified.

[0083] S209: If the abnormal verification result is a verification abnormality, update the current device alarm threshold according to the target indicator value.

[0084] Specifically, if the verification is abnormal, indicating that the current device alarm threshold can no longer meet the alarm detection requirements, it is updated to the target indicator value, that is, the target indicator value is determined as the updated device alarm threshold. It can be understood that the target indicator value is less than or equal to the indicator value to be verified, and then the target indicator value is less than the current device alarm threshold.

[0085] In the embodiment of the present application, after S209, the method further includes:

[0086] S211: If the abnormal verification result is normal, maintain the current device alarm threshold.

[0087] Specifically, if the verification is normal, the threshold value is not updated, and the subsequent unit task time consumption and operation index values ​​are continuously detected and verified.

[0088] Specifically, after obtaining the latest device alarm threshold, the above steps will be executed repeatedly during the operation of the target task to achieve real-time detection and experimental verification, and to continuously update and correct the device alarm threshold. In some cases, the device alarm threshold can be further corrected according to the acceptable degree of latency of the target task. If the latency tolerance is high, the device alarm threshold can be increased. If the latency tolerance is low, the device alarm threshold can be reduced to adapt to different task execution and service requirements.

[0089] In summary, the detection granularity of unit task running time is combined with the time detection result to determine the possible abnormal values ​​of the indicator to be verified, and then the simulator is used to simulate the running of the target task and the simulation of the indicator value to be verified to verify whether there may be task execution abnormalities in the indicator value to be verified. Then, in the case of verification abnormalities, the alarm threshold is updated in time to improve the accuracy and reliability of threshold adjustment. There is no need to rely on the determined task abnormalities for threshold adjustment, so that the adjustment operation is pre-placed and adapted to business needs, which significantly improves the equipment resource operation quality, efficiency and R&D effectiveness of task execution.

[0090] Based on some or all of the above-mentioned implementation methods, the indicator value to be verified is an operating indicator value determined based on an abnormal time consumption state. Accordingly, in some embodiments, S205 may include S301: If the duration detection result indicates that the current unit task is abnormally time-consuming, the peak operating indicator value among multiple operating indicator values ​​is determined as the indicator value to be verified.

[0091] Specifically, when it is detected that the current unit task is abnormally time-consuming, the operating state of the target device may be abnormal, and the corresponding multiple operating indicator values ​​collected may be at risk or abnormal. Then, the highest operating indicator value among the multiple operating indicator values ​​is determined as the indicator value to be verified, so as to verify whether the target task operation is abnormal in the current state of the target device. The duration detection result of the unit task is used as an indication condition for abnormal verification, thereby improving the timeliness of threshold updates and detection, and thereby improving the accuracy of threshold setting.

[0092] Taking the temperature index as an example, the highest temperature index value among multiple temperature index values ​​is determined as the index value to be verified.

[0093] In some other embodiments, in combination Figure 4 , S205 includes S303-S307:

[0094] S303: If the duration detection result indicates that the current unit task duration is abnormal, multiple historical operation indicator values ​​corresponding to the historical unit task durations are obtained;

[0095] S305: Determine a peak historical operation index value among a plurality of historical operation index values;

[0096] S307: Determine an operation index value that is higher than the peak historical operation index value among the multiple operation index values ​​as the index value to be verified.

[0097] Specifically, the duration detection result of the historical unit task time consumption in S303 is that the time consumption is normal, and multiple historical operation index values ​​are normal index values. The historical operation index values ​​corresponding to one or more historical unit task time consumptions with normal time consumption can be used for calculation, and the number of historical unit task time consumptions or the number of historical operation index values ​​used can be set based on actual needs. In some embodiments, one or more historical unit task time consumptions with normal time consumptions that are closest to the current unit task time consumption in terms of timing can be used.

[0098] Furthermore, the peak historical operating index value refers to the value with the largest value among multiple historical operating index values. Based on this value, an operating index value higher than this value is determined as the index value to be verified. It can be understood that there are multiple operating index values ​​higher than the peak historical operating index value, and multiple index values ​​to be verified are obtained accordingly, and then abnormal verification is performed for each index value to be verified respectively. If the abnormal verification results corresponding to more than one index value to be verified are successively determined to be verification abnormalities, and more than one target index value is obtained, the current device alarm threshold is updated according to the target index value with the smallest value.

[0099] In some cases, if the multiple operating index values ​​are all lower than the peak historical operating index value, the peak operating index value among the multiple operating index values ​​is determined as the index value to be verified.

[0100] Taking the temperature operation index as an example, if the current unit task takes an abnormal amount of time, the corresponding multiple temperature index values ​​are subtracted from the multiple historical temperature index values ​​of the normal historical unit task time to achieve the screening of the index value to be verified. Exemplarily, the set of multiple temperature index values ​​is (T1, T2, T3), and the set of multiple historical temperature index values ​​is (T4, T5, T6). If T4>T5>T6, T1>T3>T2, and T1>T4>T3, T1 is determined as the index value to be verified. If T4>T1, T1 can be determined as the index value to be verified based on peak screening. If T1>T3>T4>T2, T1 and T3 are determined as the index values ​​to be verified.

[0101] In this way, an indicator comparison benchmark is generated by combining the historical operating indicator values ​​with normal time consumption collected historically, so as to screen out the indicator value to be verified from multiple operating indicator values ​​when the current unit task time consumption is abnormal, so as to carry out subsequent simulation verification, which is beneficial to improve the adaptability of the indicator value to be verified and the current target device resources, and improve the accuracy of the indicator value to be verified.

[0102] Furthermore, if the indicator value to be verified is an operating indicator value under an abnormal time-consuming state of the current unit task, S207 may specifically include S401-S411:

[0103] S401: Control the simulator to simulate and run the target task, and control the simulation index value of the simulator to reach the index value to be verified.

[0104] Specifically, the target task can be simulated and run on a heterogeneous computing card based on a floating-point matrix, and the simulated index value of the heterogeneous computing card can be gradually made to reach the index value to be verified.

[0105] S403: In the process of the simulation index value approaching the index value to be verified, if an abnormal simulation unit task time consumption is detected, the simulation index value to be verified corresponding to the abnormal simulation unit task time consumption is obtained.

[0106] Specifically, the process of approaching the index value to be verified refers to the process in which the simulation index value of the simulator gradually increases to the index value to be verified during the repeated simulation operation of the unit task. In this stage, if it is detected that the simulation unit task time consumption generated by the simulation operation of any round of unit tasks is abnormal, then the multiple simulation index values ​​collected in the unit task operation round that generated the abnormal simulation unit task time consumption are obtained, and then the simulation index value to be verified is determined based on the multiple simulation index values. Specifically, if the abnormal simulation unit task time consumption is detected, the following step S413 is executed.

[0107] It can be understood that the time-consuming abnormal detection of the simulation unit task is similar to the time-consuming abnormal detection of the current unit task mentioned above, and the method for determining the simulation index value to be verified is similar to the method for determining the index value to be verified in the aforementioned S301-S307, which will not be repeated here.

[0108] S405: Control the simulator to use the simulation index value to be verified as the target simulation value of the simulator, perform simulation operation of the target task until the preset number of simulations is reached, and perform duration anomaly detection on the simulation unit task time during each simulation operation to obtain the first simulation detection result of each simulation operation.

[0109] Specifically, the target simulation value of the simulator is reset to the simulation index value to be verified, the target task is executed, and the simulation run is repeated for a preset number of times. The goal of each simulation run is for the simulation index to reach the simulation index value to be verified, and the simulation unit task time generated by the simulation execution of each unit task in the process is detected. The first simulation detection result is used to indicate whether there is an abnormal simulation unit task time consumption in the process of each simulation index reaching the simulation index value to be verified. If so, the first simulation detection result indicates that the simulation unit task time consumption is abnormal, otherwise, it indicates that the simulation unit task time consumption is normal.

[0110] S407: If the results indicating abnormal time consumption of the simulation unit tasks in each of the first simulation detection results reach a preset proportion, the abnormal verification result is determined to be a verification abnormality.

[0111] Specifically, the number of simulation unit task time-consuming anomalies in the first simulation detection results of the preset number of simulations is counted. If the proportion of this number in the preset number of simulations is greater than or equal to the preset proportion, the abnormal verification result is determined to be a verification abnormality. The preset proportion can be set based on actual needs, such as the delay tolerance setting based on the target task, and the delay tolerance forecast, the larger the preset proportion, the smaller it is. Exemplarily, the preset proportion can be 80%.

[0112] S409: Determine the simulation index value to be verified as the target index value.

[0113] It can be understood that the simulation indicator value to be verified is less than the indicator value to be verified determined above. In this way, during the simulation operation, further abnormality detection is performed below the indicator value to be verified to avoid missed detection of abnormal values ​​caused by excessively high indicator values ​​to be verified, ensure the accuracy of threshold updates, and improve the sensitivity of equipment alarms.

[0114] S411: If the results indicating abnormal time consumption of the simulation unit tasks in each first simulation detection result do not reach the preset proportion, execute the following step S413 to continue controlling the simulator to use the indicator value to be verified as the target simulation value of the simulator, and perform simulation operation of the target task until the preset number of simulations is reached, and perform duration abnormality detection on the simulation unit task time consumption during each simulation operation, obtain the operation of the third simulation detection result of each simulation operation, and then obtain the abnormal verification result corresponding to the indicator value to be verified.

[0115] In some embodiments, S207 may further include S413-S417:

[0116] S413: In the process of the simulation index value approaching the index value to be verified, if no abnormal simulation unit task time consumption is detected, continue to control the simulator to use the index value to be verified as the target simulation value of the simulator, perform simulation operation of the target task, and perform duration abnormality detection on the simulation unit task time consumption during the simulation operation to obtain a second simulation detection result;

[0117] S415: If the second simulation detection result is that the simulation unit task consumes abnormal time, determine the abnormal verification result as a verification abnormality;

[0118] S417: Determine the indicator value to be verified as the target indicator value.

[0119] Specifically, if the value of the indicator to be detected is verified to have a duration abnormality, the value of the indicator to be detected is determined to be the target indicator value, the threshold is updated, and the threshold correction is implemented to improve the detection sensitivity of the equipment task operation, which is conducive to the timely adjustment of the equipment computing power and the improvement of the equipment's business service quality. In addition, in the case of undetected simulated indicator values ​​to be verified, the abnormality verification corresponding to the indicator value to be verified is performed to avoid threshold omissions and improve the reliability of the threshold determination and adjustment plan.

[0120] Furthermore, S207 may further include S419-S423:

[0121] S419: If the second simulation detection result shows that the simulation unit task time consumption is normal, continue to control the simulator to use the indicator value to be verified as the target simulation value of the simulator, perform simulation operation of the target task until the preset number of simulations is reached, and perform duration abnormality detection on the simulation unit task time consumption during each simulation operation to obtain a third simulation detection result of each simulation operation;

[0122] S421: If the time consumption of any simulation unit task in each of the third simulation detection results is abnormal, determine the abnormal verification result as a verification abnormality;

[0123] S423: Determine the index value to be verified as the target index value;

[0124] Specifically, the specific implementation method of S419 is similar to the aforementioned S405 and will not be repeated here. After determining that any unit task simulation execution has a duration abnormality, the verification abnormality is determined, and the indicator value to be verified is determined as the target indicator value to facilitate threshold update. In this way, in the case that no duration abnormality is detected once, multiple rounds of false alarm verification are performed on the indicator value to be verified determined under the abnormal time-consuming state of the current unit task to verify the accuracy of the simulation abnormality detection link and avoid threshold leakage.

[0125] Furthermore, in some embodiments, if no abnormal time consumption of the simulation unit task occurs in any of the third simulation detection results, the current round of verification process is terminated and the current device alarm threshold is maintained.

[0126] In some other embodiments, if no abnormal time consumption of the simulation unit task appears in each of the third simulation detection results, the next indicator prediction in time sequence is performed based on multiple operation indicator values ​​to obtain the predicted indicator value; if the predicted indicator value is lower than the current device alarm threshold, the predicted indicator value is determined as the indicator value to be verified, and then the predicted indicator value is verified by simulation operation abnormality to obtain the abnormal verification result. The above-mentioned indicator prediction, abnormal verification of the predicted indicator value and determination of the target indicator value are performed using the following S309-S311, and S425-S431 specific implementation methods.

[0127] The above describes the method for determining the abnormality verification and target indicator value when the duration detection result is abnormal for the unit task duration. In other embodiments, when the duration detection result is normal for the unit task duration, combined with Figure 5 , S205 may also include S309-S313:

[0128] S309: If the duration detection result indicates that the current unit task takes a normal amount of time, the next indicator in the time series is predicted based on the multiple operation indicator values ​​to obtain a predicted indicator value;

[0129] S311: If the predicted index value is lower than the current device alarm threshold, the predicted index value is determined as the index value to be verified;

[0130] S313: If the predicted index value is higher than the current device alarm threshold, end this round of verification and maintain the current device alarm threshold.

[0131] Specifically, when the current unit task takes normal time and multiple operation index values ​​are lower than the current device alarm threshold, indicating that the operation status of the target device and the execution status of the target task are normal, the index prediction operation is performed.

[0132] In some cases, an indicator value curve is generated by combining multiple operating indicator values, and an indicator value line segment is formed based on the latest operating indicator value in the time series and its adjacent operating indicator values, and the slope of the line segment is determined as the slope of the curve, thereby predicting the predicted indicator value of the next acquisition time.

[0133] It is also possible to predict multiple predicted index values ​​for redundant verification and improve the reliability of prediction. Accordingly, determine two or more adjacent operating index values ​​that are adjacent to the latest operating index value in the time series, connect each adjacent operating index value with the latest operating index value in the time series, obtain the index value line segment corresponding to each adjacent operating index value, and then determine the slope of the curve of each index value line segment to obtain the predicted index value of the next acquisition time corresponding to each index value line segment, and input the data of the obtained multiple predicted index values ​​that are lower than the current equipment alarm threshold into the abnormal verification module for simulation verification.

[0134] In other cases, the current benchmark indicator value corresponding to the current unit task time consumption is determined based on multiple running indicator values. The current benchmark indicator value can be a peak running indicator value among the multiple running indicator values, or can be the summed average of the multiple running indicator values; and at least one historical unit task time consumption adjacent to the current unit task time consumption sequence and the historical benchmark indicator value corresponding to each historical unit task time consumption are obtained; the number of executions of the unit task is used as the horizontal coordinate, and the indicator value is used as the vertical coordinate to form a floating curve of the historical benchmark indicator value and the current benchmark indicator value, with reference to Figure 7, get the slope of the curve, and then predict the indicator corresponding to the next unit task round, and get the predicted indicator value corresponding to the time consumption of the next unit task. Figure 7 In the figure, the rightmost point is the current benchmark indicator value. Similar to the predicted indicator value of the next collection time mentioned above, the slope of the line segment formed by the previous historical benchmark indicator value and the current benchmark indicator value can be determined as the slope of the curve, and then the predicted indicator value can be obtained. It is also possible to combine multiple historical benchmark indicator values ​​adjacent to the current benchmark indicator value, respectively form line segments with the current benchmark indicator value, and then obtain multiple curve slopes to predict multiple predicted indicator values.

[0135] In one embodiment, the peak and trough values ​​of the curve wave adjacent to the current benchmark index value can also be used to form line segments with the current benchmark index value to obtain two slopes, and then perform index prediction to obtain another predicted index value.

[0136] In one embodiment, the slope of the curve can be determined using the following formula, where k is the slope of the curve, Y is Y i is the time taken for the current unit task, X i is the current benchmark indicator value, Y i-m The time taken for the historical unit task, X i-m It is the historical benchmark indicator value corresponding to the time consumption of historical unit tasks.

[0137]

[0138] In this way, when no abnormal duration or operating indicators are detected, the indicators to be verified are determined through prediction, and then it is judged whether threshold adjustment is needed, so as to realize indicator verification and threshold adjustment judgment of the whole cycle of task operation, further improve the sensitivity of threshold adjustment, and ensure the safety of task operation and the reasonable configuration of computing power.

[0139] Accordingly, when the indicator value to be verified is a predicted indicator value, S207 may include:

[0140] S425: Control the simulator to use the predicted index value as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform time anomaly detection on the simulation unit task time consumed during each simulation operation to obtain a third simulation detection result of each simulation operation;

[0141] S427: if the results indicating abnormal time consumption of the simulation unit task in each of the third simulation detection results reach a preset proportion, determining the abnormal verification result as a verification abnormality;

[0142] S429: Determine the predicted index value as the target index value;

[0143] S431: If the results indicating abnormal time consumption of the simulation unit tasks in each of the third simulation detection results do not reach a preset proportion, the abnormal verification result is determined to be normal verification, the current round of verification is ended, and the current device alarm threshold is maintained.

[0144] Specifically, S425-S427 are similar to the aforementioned S405-S407, and are not described in detail here. If the abnormal verification result for the predicted index value is normal, it indicates that the current round does not detect the device alarm threshold that needs to be updated, and the current device alarm threshold is maintained.

[0145] In this way, combined with simulation experiments, multi-stage simulation anomaly verification is carried out on the simulation indicator value to be verified, the indicator value to be verified corresponding to the detected abnormal duration, and the predicted indicator value under normal duration, so as to determine different target indicator values ​​and determine whether the alarm threshold needs to be updated based on the verification results of each stage. On the premise of realizing the threshold adjustment judgment of the entire task execution cycle, it is ensured that an accurate update threshold is generated, which significantly reduces the false alarm rate and alarm missed screening rate, and improves the detection accuracy.

[0146] It can be understood that the preset ratio in S431 may be the same as or different from the preset ratio in the aforementioned S407. In some cases, the preset ratio in S431 is higher. For example, it may be 100%, that is, the predicted index value is determined as the target index value only when the simulation unit task time consumption is abnormal each time.

[0147] The device alarm threshold determined based on the above method will repeatedly execute the above detection and verification steps during the subsequent task operation to achieve iterative update and correction of the device alarm threshold. The preset amplitude value and preset ratio of the duration detection can be adjusted based on the delay tolerance of the unit task to adjust the sensitivity of the threshold update, thereby achieving adaptive adjustment of resource configuration such as device computing power to meet the service quality requirements of different businesses.

[0148] In one embodiment, taking temperature as an example, the process of determining the target index value may be as follows: if multiple temperature index values ​​corresponding to the current unit task time consumption are all lower than the current device temperature alarm threshold, and the current unit task time consumption is detected to be abnormal, then determine the target index value T to be verified from the multiple temperature index values. i , input the simulator to verify the simulation anomaly; i To simulate the target simulation value corresponding to the temperature index, the floating-point matrix is ​​continuously operated on the heterogeneous computing card to increase the card temperature of the heterogeneous computing card to T i ; If reaching T i During the experiment, the abnormal simulation unit task time was detected, and the T corresponding to the abnormal simulation unit task time was determined. i-m , then T i-mAs the target simulation value, multiple rounds of simulation run experiments with predicted simulation times are performed, and T is confirmed based on the obtained multiple first simulation test results. i-m Is the abnormality of the value a false alarm? If not, T i-m Determine the target indicator value; if it is a false alarm, continue to follow T i The target simulation value is set, and the experimental process is executed. If the simulation unit task is detected to be abnormal, T i Determine the target indicator value. If no abnormal simulation unit task time consumption is detected, continue to execute multiple rounds of T i The simulation is run until multiple rounds of simulation runs of the predicted number of simulations are completed to obtain multiple third simulation test results; if it is confirmed according to the multiple third simulation test results that T i If the abnormal value is not a false positive, then T i Determine the target indicator value; if it is a false alarm, it is necessary to perform several more rounds at T(i) to verify the accuracy of the detection link. If the verification of several rounds does not show abnormal simulation unit task time consumption, it is necessary to predict the next predicted indicator value according to the slope pattern. Based on the fluctuation trend of the temperature slope, the predicted temperature value T that may be abnormal is predicted. i+1 , if T i+1 After running several rounds of simulation experiments and verification, if the simulation unit task time consumption is abnormal, it will be determined as the target indicator value. Otherwise, this round of experimental verification will be ended, and the next time the unit task time consumption is detected, the cyclic detection and simulation experiment verification will be carried out to iteratively verify and update the equipment temperature alarm threshold.

[0149] In parallel, if the multiple temperature index values ​​corresponding to the current unit task time are all lower than the current device temperature alarm threshold, and the current unit task time is detected to be normal, the above-mentioned next predicted index value is executed to obtain the predicted temperature value T i+1 ', to conduct corresponding verification, this process is consistent with the previous predicted temperature value T i+1 ' is the same as the verification method, so I will not go into details.

[0150] The above describes a method for determining a threshold value when multiple operating indicator values ​​are lower than the current device alarm threshold, that is, when the alarm judgment is not reached. In the embodiment of the present application, there may also be a situation where the operating indicator value is higher than the current device alarm threshold, indicating that there is a risk in the task operation. Accordingly, the method may also include:

[0151] S213: If any of the multiple operating indicator values ​​is higher than the current device alarm threshold, and the duration detection result indicates that the current unit task is abnormally time-consuming, a device alarm message is generated.

[0152] Specifically, if the operating index value is higher than the current device alarm threshold and the current unit task takes an abnormal amount of time, indicating that the target device has an abnormality during the operation of the target task, an alarm is processed to trigger the adjustment of the device computing power configuration and improve the safety of device operation.

[0153] Furthermore, in some embodiments, the method may further include S215-S221:

[0154] S215: If any of the multiple operation index values ​​is higher than the current device alarm threshold, and the duration detection result indicates that the time consumption of the current unit task is normal, the operation index value higher than the current device alarm threshold is determined as an abnormal index value;

[0155] S217: Control the simulator to use the abnormal index value as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform duration abnormality detection on the duration of the simulation unit task of the simulator during each simulation operation to obtain a fourth simulation detection result of each simulation operation;

[0156] S219: if the results indicating abnormal time consumption of the simulation unit task in each of the fourth simulation detection results reaches a preset proportion, generating device alarm information;

[0157] S219: If the results indicating abnormal time consumption of the simulation unit tasks in each of the fourth simulation detection results do not reach a preset proportion, end this round of verification without performing alarm processing.

[0158] Specifically, if the operating index value is higher than the current device alarm threshold, and the current unit task takes normal time, indicating that there may be a false alarm, then the abnormal index value is simulated for abnormal verification, which is similar to the aforementioned S405-S407, except that if the preset ratio is reached, an alarm is directly processed, and if it is not reached, no alarm is issued, and the data corresponding to the abnormal index value can be written into the log and marked as abnormal for manual investigation. In this way, multiple rounds of simulation verification are performed on the suspected abnormal indicators to reduce the probability of false alarms and improve the accuracy of alarms, so as to continue to execute the target task, which is conducive to reducing the task execution time and improving the utilization rate of equipment computing power.

[0159] In summary, this solution is based on the collected time-series-based operating indicator data and unit task time consumption, through real-time detection of the anomaly detection model, and performs experimental verification on the detected or predicted anomalies to determine whether to update the threshold or output different target indicator values ​​based on different verification situations to update the equipment alarm threshold, and continuously verify and update the correction during the subsequent task operation. Compared with the existing method of adjusting according to the feedback of task execution anomalies, the acquired alarm threshold is more accurate, which is conducive to improving resource operation efficiency; and the solution of this application continuously corrects the alarm threshold, which is conducive to reducing the probability of false alarms and reducing The amount of data on operation and maintenance problems can be reduced, thereby alleviating the cost investment in operation and maintenance manpower; at the same time, by building a simulator to perform simulation anomaly verification, false alarms or weak impact indicators on target tasks such as large model training can be distinguished, thereby reducing the amount of related alarms during the execution of target tasks such as large model training, and continuing to execute tasks, which is conducive to shortening the task execution time (such as large model training time) and improving the computing power utilization rate during task execution; in addition, from the perspective of operation services, this application constructs an anomaly detection and simulation anomaly verification framework to correct the accuracy of the alarm threshold, which is conducive to improving the service quality of computing power operation and maintenance and alleviating the operational pressure of operation and maintenance personnel.

[0160] The present application embodiment also provides a device 800 for determining a device alarm threshold. Fig. 9 As shown, Fig. 9 A schematic diagram of the structure of a device alarm threshold determination apparatus provided in an embodiment of the present application is shown, and the apparatus may include the following modules.

[0161] Acquisition module 10: used to acquire the current unit task time consumed by the target device to execute the target task, multiple operation index values ​​based on time series corresponding to the current unit task time consumed, and the current device alarm threshold, wherein the target task includes a unit task executed cyclically, and the current unit task time consumed is the actual operation time of the unit task currently executed by the target device;

[0162] Duration detection module 20: used to perform duration anomaly detection on the duration of the current unit task and obtain a duration detection result;

[0163] Index determination module 30: used to determine the index value to be verified according to the duration detection result and the multiple operating index values ​​if the multiple operating index values ​​are lower than the current device alarm threshold;

[0164] Abnormal verification module 40: used to use the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, perform simulation operation of the target task, and perform indicator abnormality verification based on the simulation unit task time of the simulator during the simulation operation to obtain the abnormality verification result and the target indicator value;

[0165] Threshold updating module 50: used to update the current device alarm threshold according to the target indicator value if the abnormal verification result is a verification abnormality.

[0166] In some embodiments, the duration detection module 20 includes:

[0167] Historical time-consuming acquisition submodule: used to obtain the time-consuming of multiple historical unit tasks. The time-consuming of multiple historical unit tasks does not have any abnormalities and is continuous with the time-consuming of the current unit task in terms of time sequence.

[0168] Duration curve generation submodule: used to generate a duration curve based on multiple historical unit task durations and the current unit task duration;

[0169] Amplitude increment determination submodule: used to determine the amplitude increment corresponding to the time consumption of the current unit task based on the duration curve;

[0170] Time-consuming judgment submodule: used to determine that the duration detection result is abnormal time consumption of the current unit task if the amplitude value increase is greater than or equal to the preset amplitude value.

[0171] In some embodiments, the indicator determination module 30 may include a first verification value determination submodule: for determining a peak operating indicator value among multiple operating indicator values ​​as the indicator value to be verified if the duration detection result indicates that the current unit task is abnormally time-consuming.

[0172] In some other embodiments, the indicator determination module 30 may include:

[0173] Historical indicator acquisition submodule: used to obtain multiple historical operation indicator values ​​corresponding to the historical unit task duration if the duration detection result indicates that the current unit task duration is abnormal;

[0174] A peak index determination submodule: used to determine a peak historical operation index value among a plurality of historical operation index values;

[0175] The second verification value determination submodule is used to determine an operation index value higher than a peak historical operation index value among multiple operation index values ​​as an index value to be verified.

[0176] Accordingly, in some embodiments, the abnormality verification module 40 can be specifically used to: control the simulator to simulate and run the target task, and control the simulation index value of the simulator to reach the index value to be verified; in the process of the simulation index value approaching the index value to be verified, if an abnormal simulation unit task time is detected, obtain the simulation index value to be verified corresponding to the abnormal simulation unit task time; control the simulator to use the simulation index value to be verified as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform duration abnormality detection on the simulation unit task time during each simulation operation to obtain the first simulation detection result of each simulation operation; if the result indicating abnormal simulation unit task time in each first simulation detection result reaches a preset proportion, determine that the abnormality verification result is a verification abnormality; and determine the simulation index value to be verified as the target indicator value.

[0177] In some embodiments, the abnormal verification module 40 can also be specifically used for: in the process of the simulation index value approaching the index value to be verified, if no abnormal simulation unit task time is detected, continue to control the simulator to use the index value to be verified as the target simulation value of the simulator, perform simulation operation of the target task, and perform duration abnormality detection on the simulation unit task time during the simulation operation to obtain a second simulation detection result; if the second simulation detection result is that the simulation unit task time is abnormal, determine the abnormal verification result as a verification abnormality; and determine the index value to be verified as the target index value.

[0178] Furthermore, the abnormality verification module 40 can also be specifically used for: if the second simulation detection result shows that the simulation unit task duration is normal, continue to control the simulator to use the indicator value to be verified as the target simulation value of the simulator, perform simulation operation of the target task until the preset number of simulations is reached, and perform duration abnormality detection on the simulation unit task duration during each simulation operation to obtain a third simulation detection result of each simulation operation; if each third simulation detection result indicates that the duration of any simulation unit task is abnormal, determine that the abnormality verification result is a verification abnormality; and determine the indicator value to be verified as the target indicator value.

[0179] In some embodiments, the indicator determination module 30 may further include:

[0180] Index prediction submodule: used to predict the next index in the time series based on multiple running index values ​​to obtain the predicted index value if the duration detection result indicates that the current unit task takes normal time;

[0181] The third verification value determination submodule is used to determine the predicted index value as the index value to be verified if the predicted index value is lower than the current device alarm threshold.

[0182] Accordingly, in some embodiments, the abnormality verification module 40 can be specifically used to: control the simulator to use the predicted index value as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform duration abnormality detection on the simulation unit task time during each simulation operation to obtain a third simulation detection result of each simulation operation; if the result indicating abnormal duration of the simulation unit task in each third simulation detection result reaches a preset proportion, determine that the abnormality verification result is a verification abnormality; and determine the predicted index value as the target index value.

[0183] In some embodiments, the device also includes an alarm information generation module: used to generate equipment alarm information if any of the multiple operating indicator values ​​is higher than the current equipment alarm threshold and the duration detection result indicates that the current unit task is abnormally time-consuming.

[0184] In some embodiments, the apparatus further comprises:

[0185] Abnormal index determination module: used for determining the operating index value higher than the current equipment alarm threshold as an abnormal index value if any of the multiple operating index values ​​is higher than the current equipment alarm threshold and the duration detection result indicates that the current unit task takes normal time;

[0186] The abnormality verification module 40 is also used to control the simulator to use the abnormality index value as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform time abnormality detection on the simulation unit task time of the simulator during each simulation operation to obtain a fourth simulation detection result of each simulation operation;

[0187] The alarm information generation module is also used to generate device alarm information if the results indicating abnormal time consumption of the simulation unit task in each fourth simulation detection result reaches a preset proportion.

[0188] It should be noted that the above device embodiment and method embodiment are based on the same implementation method.

[0189] An embodiment of the present application provides a device, which may be a terminal or a server, including a processor and a memory, wherein at least one instruction or at least one program is stored in the memory, and the at least one instruction or the at least one program is loaded and executed by the processor to implement a device alarm threshold determination method or a neural network training method as provided in the above method embodiment.

[0190] The memory can be used to store software programs and modules. The processor executes various functional applications and abnormality detection by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory can also include a memory controller to provide the processor with access to the memory.

[0191] The method embodiments provided in the embodiments of the present application can be executed in electronic devices such as mobile terminals, computer terminals, servers or similar computing devices. Fig.10 1 is a hardware structure block diagram of an electronic device for a device alarm threshold determination method or a neural network training method provided in an embodiment of the present application. Fig.10 As shown, the electronic device 900 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 910 (the processor 910 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 930 for storing data, and one or more storage media 920 (such as one or more mass storage devices) for storing application programs 923 or data 922. Among them, the memory 930 and the storage medium 920 can be short-term storage or permanent storage. The program stored in the storage medium 920 may include one or more modules, each of which may include a series of instruction operations in the electronic device. Furthermore, the central processing unit 910 can be configured to communicate with the storage medium 920 and execute a series of instruction operations in the storage medium 920 on the electronic device 900. The electronic device 900 may also include one or more power supplies 960, one or more wired or wireless network interfaces 950, one or more input and output interfaces 940, and / or, one or more operating systems 921, such as Windows Server TM , Mac OS X TM , Unix TM , LinuxTM, FreeBSDTM, etc.

[0192] The input / output interface 940 may be used to receive or send data via a network. The specific example of the network may include a wireless network provided by a communication provider of the electronic device 900. In one example, the input / output interface 940 includes a network adapter (Network Interface Controller, NIC), which may be connected to other network devices via a base station so as to communicate with the Internet. In one example, the input / output interface 940 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0193] It can be understood by those skilled in the art that Fig.10 The structure shown is only for illustration and does not limit the structure of the above electronic device. Fig.10 More or fewer components as shown, or with Fig.10 Different configurations shown.

[0194] An embodiment of the present application also provides a computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to an anomaly detection method in a method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement the anomaly detection method provided by the above method embodiment.

[0195] Optionally, in this embodiment, the storage medium may be located in at least one of the multiple network servers of the computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0196] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the above-mentioned various optional implementations.

[0197] The device alarm threshold determination method, apparatus, equipment, storage medium, server, terminal and program product provided by the above-mentioned present application, the technical solution of the present application first obtains the current unit task time generated by the unit task of the target device to execute the target task, multiple timing-based operating indicator values ​​corresponding to the current unit task time and the current device alarm threshold; and performs a duration anomaly detection on the current unit task time to obtain a duration detection result to determine whether there is a time anomaly; compares multiple operating indicator values ​​with the current device alarm threshold, if multiple operating indicator values ​​are lower than the current device alarm threshold, determines the indicator value to be verified based on the duration detection result and multiple operating indicator values; and then uses the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, performs simulation operation of the target task, and based on During the simulation process, the simulator's simulated unit task consumes time to verify the indicator anomaly, and obtains the anomaly verification result and the target indicator value; if the anomaly verification result is a verification anomaly, the current equipment alarm threshold is updated according to the target indicator value; in this way, the unit task is used as the detection granularity of the running time, and the duration detection result is combined to determine the possible abnormal indicator value to be verified, and then the simulator simulates the running target task and the simulation of the indicator value to be verified to verify whether the indicator value to be verified may have task execution anomalies, and then in the case of verification anomalies, the alarm threshold is updated in time, improving the accuracy and reliability of threshold adjustment, without relying on the determined task anomalies for threshold adjustment, so that the adjustment operation is advanced and adapted to business needs, significantly improving the equipment resource operation quality, efficiency and task execution R&D effectiveness.

[0198] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of the present application are described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0199] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0200] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware, or may be accomplished by instructing the relevant hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0201] The above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for determining a device alarm threshold, characterized in that: The method comprises: Obtaining the current unit task time consumed by the target device to execute the target task, multiple operation indicator values ​​based on timing corresponding to the current unit task time consumed, and the current device alarm threshold, wherein the target task includes a unit task that is executed cyclically, and the current unit task time consumed is the actual operation time of the unit task currently executed by the target device; Performing duration anomaly detection on the time consumed by the current unit task to obtain a duration detection result; If multiple operating indicator values ​​are lower than the current device alarm threshold, determine the indicator value to be verified according to the duration detection result and the multiple operating indicator values; Taking the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, performing simulation operation of the target task, and performing indicator abnormality verification based on the simulation unit task time consumption of the simulator during the simulation operation to obtain the abnormality verification result and the target indicator value; If the abnormal verification result is a verification abnormality, the current device alarm threshold is updated according to the target indicator value.

2. The method according to claim 1, characterized in that The abnormal duration detection of the current unit task time is performed to obtain the duration detection result, which includes: Acquire multiple historical unit task durations, where the multiple historical unit task durations are not abnormal and are continuous with the current unit task duration in terms of time sequence; Generate a duration curve based on the multiple historical unit task durations and the current unit task duration; Determine the amplitude increment corresponding to the time consumption of the current unit task based on the duration curve; If the amplitude increment is greater than or equal to the preset amplitude value, it is determined that the duration detection result is that the current unit task consumes an abnormal amount of time.

3. The method according to claim 1, characterized in that If the multiple operating indicator values ​​are lower than the current device alarm threshold, determining the indicator value to be verified according to the duration detection result and the multiple operating indicator values ​​includes: If the duration detection result indicates that the current unit task consumes abnormal time, a peak operating index value among the multiple operating index values ​​is determined as the index value to be verified.

4. The method according to claim 1, characterized in that: If the multiple operating indicator values ​​are lower than the current device alarm threshold, determining the indicator value to be verified according to the duration detection result and the multiple operating indicator values ​​includes: If the duration detection result indicates that the current unit task is abnormally time-consuming, multiple historical operation indicator values ​​corresponding to the historical unit task durations are obtained; Determining a peak historical operating indicator value among the plurality of historical operating indicator values; An operation index value among the multiple operation index values ​​that is higher than the peak historical operation index value is determined as the index value to be verified.

5. The method according to any one of claims 1 to 4, characterized in that The method of using the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, performing simulation operation of the target task, and performing indicator abnormality verification based on the simulation unit task time consumption of the simulator during the simulation operation, and obtaining the abnormality verification result and the target indicator value includes: Controlling the simulator to simulate and run the target task, and controlling the simulation index value of the simulator to reach the index value to be verified; In the process of the simulation index value approaching the index value to be verified, if an abnormal simulation unit task time consumption is detected, obtaining the simulation index value to be verified corresponding to the abnormal simulation unit task time consumption; Control the simulator to use the simulation index value to be verified as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform duration anomaly detection on the duration of the simulation unit task during each simulation operation to obtain a first simulation detection result of each simulation operation; If the results indicating abnormal time consumption of the simulation unit task in each of the first simulation detection results reach a preset proportion, determining the abnormal verification result as a verification abnormality; The simulation index value to be verified is determined as the target index value.

6. The method according to claim 5, characterized in that The method further comprises: In the process of the simulation index value approaching the index value to be verified, if no abnormal simulation unit task time consumption is detected, continue to control the simulator to use the index value to be verified as the target simulation value of the simulator, perform simulation operation of the target task, and perform duration abnormality detection on the simulation unit task time consumption during the simulation operation to obtain a second simulation detection result; If the second simulation detection result is that the simulation unit task consumes abnormal time, determining that the abnormal verification result is a verification abnormality; The to-be-verified indicator value is determined as the target indicator value.

7. The method according to claim 6, characterized in that The method further comprises: If the second simulation detection result is that the simulation unit task time consumption is normal, continue to control the simulator to use the to-be-verified indicator value as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform duration abnormality detection on the simulation unit task time consumption during each simulation operation to obtain a third simulation detection result of each simulation operation; If the third simulation detection results indicate that any simulation unit task consumes abnormal time, determining the abnormal verification result as a verification abnormality; The to-be-verified indicator value is determined as the target indicator value.

8. The method according to any one of claims 1 to 4, characterized in that If the multiple operating indicator values ​​are lower than the current device alarm threshold, determining the indicator value to be verified according to the duration detection result and the multiple operating indicator values ​​also includes: If the duration detection result indicates that the current unit task consumes normal time, predicting the next indicator in the time series based on the multiple operation indicator values ​​to obtain a predicted indicator value; If the predicted index value is lower than the current device alarm threshold, the predicted index value is determined as the index value to be verified.

9. The method according to claim 8, characterized in that The method of using the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, performing simulation operation of the target task, and performing indicator abnormality verification based on the simulation unit task time consumption of the simulator during the simulation operation, and obtaining the abnormality verification result and the target indicator value includes: Control the simulator to use the prediction index value as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform duration anomaly detection on the duration of the simulation unit task during each simulation operation to obtain a third simulation detection result of each simulation operation; If the results indicating abnormal time consumption of the simulation unit task in each of the third simulation detection results reach a preset proportion, determining that the abnormal verification result is a verification abnormality; The predicted index value is determined as the target index value.

10. The method according to any one of claims 1 to 4, characterized in that The method further comprises: If any one of the multiple operating indicator values ​​is higher than the current device alarm threshold, and the duration detection result indicates that the current unit task is abnormally time-consuming, a device alarm message is generated.

11. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: If any of the multiple operating index values ​​is higher than the current device alarm threshold, and the duration detection result indicates that the current unit task takes normal time, the operating index value higher than the current device alarm threshold is determined as an abnormal index value; Control the simulator to use the abnormal index value as the target simulation value of the simulator, perform simulation operation of the target task until a preset number of simulations is reached, and perform duration abnormality detection on the duration of the simulation unit task of the simulator during each simulation operation to obtain a fourth simulation detection result of each simulation operation; If the results of each of the fourth simulation detection results indicating that the simulation unit task time consumption is abnormal reaches a preset proportion, a device alarm message is generated.

12. A device for determining an alarm threshold of an equipment, characterized in that: The device comprises: Acquisition module: used to acquire the current unit task time consumed by the target device to execute the target task, multiple operation index values ​​based on timing corresponding to the current unit task time consumed, and the current device alarm threshold, wherein the target task includes a unit task executed cyclically, and the current unit task time consumed is the actual operation time of the unit task currently executed by the target device; Duration detection module: used to perform duration anomaly detection on the duration of the current unit task and obtain a duration detection result; Indicator determination module: used to determine the indicator value to be verified according to the duration detection result and the multiple operating indicator values ​​if the multiple operating indicator values ​​are lower than the current device alarm threshold; Abnormal verification module: used to use the indicator value to be verified as the target simulation value of the simulator corresponding to the target device, perform simulation operation of the target task, and perform indicator abnormality verification based on the simulation unit task time of the simulator during the simulation operation to obtain abnormal verification results and target indicator value; Threshold updating module: used to update the current device alarm threshold according to the target indicator value if the abnormal verification result is a verification abnormality.

13. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the device alarm threshold determination method according to any one of claims 1 to 11.

14. A computer device, characterized in that: The device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the device alarm threshold determination method according to any one of claims 1 to 11.

Images