Method and device for identifying instance usage status

Abstract

The present application provides a method and device for identifying the usage status of an instance, including: obtaining the operation data of at least one performance metric of a target instance within an identification period, extracting features from the operation data within an identification period for each performance metric to obtain the feature data of each performance metric, calculating the performance index of the target instance from the feature data of the performance metrics, and obtaining the usage status of the target instance based on the performance index of the target instance. The present application processes the operation data of each performance metric of the instance to obtain the performance index of the target instance, and then determines the usage status of the target instance based on the performance index, thereby improving the accuracy of identifying the usage status.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to a method and device for identifying the usage status of instances. Background Art

[0002] An instance is a specific cloud virtual machine (CVM for short), which includes the most basic computing components such as CPU, memory, operating system, network, and disk. It is the basic unit for cloud service providers to provide cloud services, and multiple instances can run in each server. In order to manage numerous instances in a refined manner, cloud service providers need to identify the usage status of instances.

[0003] Generally, when identifying the usage status of an instance, for instances of the same type and the same specification, a fixed threshold is used for usage status identification. When the running data of a certain performance metric is greater than or equal to the set threshold, the usage status of the instance is set to the high-load state; when the running data of a certain performance metric is less than the set threshold, the usage status of the instance is set to the idle state.

[0004] However, generally, the order of magnitude of the running data of each performance metric of an instance reaches the million level. It is necessary to compare the running data with the set threshold one by one, which results in a large amount of data calculation, time-consuming calculation, and requires manual setting of an appropriate threshold. If the threshold is set improperly, it is easy to lead to incorrect identification results. Summary of the Invention

[0005] This application provides a method and device for identifying the usage status of instances to solve the technical problem that the prior art cannot accurately identify the usage status of instances.

[0006] An embodiment of this application provides a method for identifying the usage status of instances, including:

[0007] Obtain the running data of at least one performance metric of a target instance within an identification period;

[0008] For each performance metric, perform feature extraction on the running data within an identification period to obtain the feature data of each performance metric;

[0009] Calculate the performance index of the target instance from the feature data of the performance metric;

[0010] Obtain the usage status of the target instance based on the performance index of the target instance.

[0011] In an embodiment, each identification period includes multiple time intervals; for each performance metric, performing feature extraction on the running data within an identification period to obtain the feature data of each performance metric specifically includes at least one of the following:

[0012] For each performance metric, perform statistical analysis on the operation data within each time interval to obtain the statistical results of the performance metric within each time interval;

[0013] For each performance metric, use a machine learning algorithm to extract features from the operation data within each time interval to obtain the time series feature data of the performance metric within each time interval;

[0014] For each performance metric, use a machine learning algorithm to perform portrait analysis on the operation data within an identification cycle to obtain the portrait data of the target instance.

[0015] In one embodiment, calculate the performance index of the target instance from the feature data of the performance metric, specifically including:

[0016] Use a machine learning algorithm to process the feature data of each performance metric to obtain the weight of each performance metric;

[0017] Calculate the performance index of the target instance based on the weight of each performance metric and the feature data of each performance metric.

[0018] In one embodiment, use a machine learning algorithm to process the feature data of each performance metric to obtain the weight of each performance metric, specifically including:

[0019] Use a machine learning algorithm to process the statistical results of the performance metric within each time interval, the time series feature data of the performance metric within each time interval, the portrait data of the target instance, and the preset initial weight of each performance metric to obtain the weight of each performance metric.

[0020] In one embodiment, calculate the performance index of the target instance based on the weight of each performance metric and the feature data of each performance metric, specifically including:

[0021] For each time interval, perform normalization processing on the statistical results, time series feature data, and portrait data of each performance metric; add the normalized statistical results, time series feature data, and portrait data to obtain the metric data of each performance metric within each time interval;

[0022] Use the weight of each performance metric to perform weighted summation processing on the metric data of each performance metric within each time interval to obtain the performance index of the target instance within each time interval;

[0023] Perform weighted summation on the performance index of the target instance within each time interval and the preset time weight to obtain the performance index of the target instance within the identification cycle.

[0024] In the above technical solution, by extracting features from the performance indices in different time intervals, obtaining the feature data of each performance metric, then performing weighted summation on the feature data of each performance metric to obtain the performance index of the target instance in each time interval, and further performing weighted summation on the performance indices of the target instance in each time interval, the obtained performance index of the target instance can accurately reflect the characteristics of the operation data of the target instance.

[0025] In one embodiment, obtaining the usage status of the target instance according to the performance index of the target instance specifically includes:

[0026] When the performance index of the target instance is less than the first threshold, it is determined that the target instance is in an idle state;

[0027] When the performance index of the target instance is greater than or equal to the first threshold and less than the second threshold, it is determined that the target instance is in a low-load state;

[0028] When the performance index of the target instance is greater than or equal to the second threshold and less than the third threshold, it is determined that the target instance is in a normal-load state;

[0029] When the performance index of the target instance is greater than or equal to the third threshold, it is determined that the target instance is in a high-load state.

[0030] In one embodiment, the method further includes:

[0031] Obtaining the operation data of at least one performance metric of the target instance in the previous recognition cycle and the performance index of the target instance in the previous recognition cycle;

[0032] Using a machine learning algorithm to process the operation data of at least one performance metric in the previous recognition cycle and the performance index of the target instance to obtain the first threshold, the second threshold, and the third threshold.

[0033] In the above technical solution, by using a machine learning algorithm to perform data mining on the operation data of at least one performance metric in the previous recognition cycle and the performance index of the target instance, obtaining the first threshold, the second threshold, and the third threshold, without manual setting of thresholds, and the thresholds can also adapt to changes in the operation data, which can improve the accuracy of the recognition result.

[0034] In one embodiment, obtaining the operation data of at least one performance metric of the target instance in one recognition cycle specifically includes:

[0035] Obtaining the original operation data of at least one performance metric collected by the monitoring point in one recognition cycle;

[0036] Aggregate the raw operation data of at least one performance metric within an identification period, fill in missing values, and clean outliers to obtain the operation data of at least one performance metric of the target instance within an identification period.

[0037] Another embodiment of the present application provides an identification device, including: a processor, and a memory communicatively connected to the processor;

[0038] The memory stores computer-executable instructions;

[0039] The processor executes the computer-executable instructions stored in the memory to implement the identification method for the usage status of the instance provided in the above embodiment.

[0040] Another embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the identification method for the usage status of the instance provided in the above embodiment.

[0041] Another embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the identification method for the usage status of the instance provided in the above embodiment.

[0042] The identification method and device for the usage status of the instance provided in the embodiments of the present application extract feature data of each performance metric by extracting features from the operation data of at least one performance metric within an identification period, and then calculate the performance index of the target instance from the feature data of the performance metric, realizing the conversion of operation data of millions of sizes into the performance index of the target instance, reducing the data calculation amount. Then, according to the performance index of the target instance, the usage status of the target instance is obtained. Because the performance index comprehensively considers more comprehensive operation data of the instance from the time dimension and type dimension, the reliability of the status identification result is increased, and the fluctuations of individual operation data will not have a great interference on the performance index, thus avoiding the generation of false alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0044] Figure 1 It is an identification system for the usage status of the instance provided in an embodiment of the present application;

[0045] Figure 2 It is an identification method for the usage status of the instance provided in another embodiment of the present application;

[0046] Figure 3 It is an identification method for the usage status of the instance provided in another embodiment of the present application;

[0047] Figure 4 A device for identifying the usage status of an example provided in yet another embodiment of the present application;

[0048] Figure 5 Another embodiment of the present application provides an example usage status identification device.

[0049] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0050] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0051] An instance is the basic unit of cloud services provided by cloud service providers, and instances can be divided into general-purpose instances, computing instances, and memory instances according to different configurations. Cloud service providers have millions of servers, and each server also runs multiple instances. Therefore, the number of instances used by cloud service providers to provide services is in the millions. If cloud service providers want to carry out fine-grained management of instances of this magnitude, the basis is to have awareness of the usage status of each instance. For users, understanding the usage status of the instances they purchased is not only conducive to improving service stability and reducing business risks, but is also a prerequisite for reducing costs and increasing benefits. Therefore, how to identify the usage status of millions of instances is the problem that this application is committed to solving.

[0052] When identifying instance usage, fixed thresholds are used for instance identification of the same type and specification. When the operating data for a performance indicator is greater than or equal to the set threshold, the instance's usage status is set to high load. When the operating data for a performance indicator is less than the set threshold, the instance's usage status is set to idle.

[0053] However, there are many types and specifications of instances, and each type and each specification of instance has different performance indicators. Therefore, corresponding running data needs to be stored for each type and each specification of instance, resulting in high data storage costs. It is also necessary to rely on expert experience to set different thresholds for each performance indicator of each type and each specification of instance. Moreover, due to the diversity of instance deployment services, the uncertainty of the fluctuations in the running data of instances makes it easy to misestimate the usage status of instances by setting thresholds based on expert experience.

[0054] This application provides a method and device for identifying the usage status of an instance. By processing the running data of each performance indicator of instances of the same specification and the same type to obtain the performance index of the instance, and identifying the usage status of each instance according to the performance index of the instance, accurate identification of the usage status of the instance can be achieved.

[0055] As Figure 1 shown, an embodiment of this application provides a system for identifying the usage status of an instance. The identification system includes an identification device 100 and multiple servers 200. Multiple instances are running on each server 200, and monitoring points are arranged on each server 200. The running data of each performance indicator of each instance is collected by the monitoring points, and the identification device 100 executes computer instructions to implement the identification method in the following embodiments to process the running data of each performance indicator to determine the usage status of the instance.

[0056] As Figure 2 shown, an embodiment of this application provides a method for identifying the usage status of an instance. The identification method runs on the identification device. The identification method specifically includes the following steps:

[0057] S101. Obtain the running data of at least one performance indicator of a target instance within an identification period.

[0058] In this step, different types and different specifications of instances have different performance indicators. By setting up a monitoring program, the running data of each performance indicator of the target instance running on the server is monitored. For example, each performance indicator of the target instance includes any one or a combination of CPU, memory, network, and disk.

[0059] When the performance indicators of the target instance include CPU, memory, network, and disk, the running data of each performance indicator of the target instance includes the current idle CPU percentage, the current kernel space occupied CPU percentage, the current user space occupied CPU percentage, the amount of used memory, the memory actually used by the user, the number of bytes read per second by the disk, the number of bytes written per second by the disk, the average rate of public network incoming traffic, the average rate of public network outgoing traffic, the average rate of private network incoming traffic, and the average rate of private network outgoing traffic.

[0060] S102. For each performance metric, extract features from the operation data within an identification cycle to obtain the feature data of the performance metric.

[0061] In this step, for each performance metric, use statistical analysis techniques and / or machine learning algorithms to extract features from the operation data of the performance metric within an identification cycle to obtain the feature data of the performance metric.

[0062] S103. Calculate the performance index of the target instance from the feature data of the performance metric.

[0063] In this step, when calculating the feature data of the performance metric, weighted summation can be used, and other calculation methods can also be used for processing.

[0064] When using the weighted summation method for processing, the feature data of each performance metric can reflect the characteristics of the operation data of the performance metric, and the weight setting conditions are configured through a configuration method. For example: the weight setting conditions are specifically that when the feature data shows a certain trend, its weight value is larger. When the feature data shows another trend, its weight value is smaller.

[0065] Use the weight setting conditions to train the machine learning algorithm, and use the trained machine learning algorithm to perform data mining processing on the feature data of each performance metric to obtain the weight of each performance metric. Use the weight of each performance metric to perform weighted summation on the feature data of the performance metric to obtain the performance index of the target instance.

[0066] S104. Obtain the usage status of the target instance based on the performance index of the target instance.

[0067] In this step, set various status conditions, such as: low load condition, idle status condition, normal load condition, and high load condition.

[0068] When the performance index of the target instance meets the idle status condition, determine that the usage status of the target instance is the idle status. When the performance index of the target instance meets the low load condition, determine that the usage status of the target instance is the low load status. When the performance index of the target instance meets the normal load condition, determine that the usage status of the target instance is the normal load status. When the performance index of the target instance meets the high load condition, determine that the usage status of the target instance is the high load status.

[0069] In the above technical solution, the information contained in a large number of operation data is refined into a performance index. Without comparing with thresholds one by one, the usage status of the target instance is determined only based on one performance index, which can reduce the amount of data processing. Moreover, because the performance index comprehensively considers more comprehensive operation data of the instance, the reliability of the status recognition result is increased, and the fluctuation of individual operation data will not cause a large interference to the performance index, thus avoiding the generation of false alarms.

[0070] As Figure 3 shown, another embodiment of the present application provides a method for recognizing the usage status of an instance. This recognition method runs on a recognition device, and the recognition method specifically includes the following steps:

[0071] S201. Obtain the operation data of at least one performance metric of the target instance within one recognition period.

[0072] In this step, each recognition period includes multiple time intervals. For example, if one recognition period is one week and one week includes 7 days, then one week can be divided into 7 time intervals.

[0073] Obtain the original operation data of the performance metric within one recognition period collected by the monitoring point. For example: pull the operation data of performance metrics such as CPU, memory, disk, and network from the database.

[0074] Perform data aggregation, missing value filling, and outlier cleaning on the original operation data of the performance metric within one recognition period to obtain the operation data of each performance metric of the target instance within one recognition period. For example: collect the original operation data for one week and perform data preprocessing such as data aggregation, missing value filling, and outlier cleaning on the original operation data.

[0075] When performing data aggregation, when the operation data stored by the monitoring point is at a 5 - minute scale, it needs to be aggregated into data at an hourly or daily scale. When there is a fault at the monitoring point, data may be missing. At this time, the method of forward filling needs to be used to fill the missing values to facilitate subsequent algorithm calculations. When there is an abnormal collection at the monitoring point, the collected data may contain abnormally large data. At this time, the hypothesis testing method needs to be used to accurately identify and clean these outliers to avoid interference with subsequent algorithms.

[0076] S202. For each performance metric, perform feature extraction on the operation data within one recognition period to obtain the feature data of the performance metric.

[0077] In this step, when performing feature extraction on the operation data, feature extraction can be performed on the operation data within each time interval to obtain the feature data of the performance metric within each time interval.

[0078] In one embodiment, the characteristic data of the performance metrics includes at least one of the following: statistical results, time-series characteristic data, and portrait data.

[0079] In one embodiment, for each performance metric, statistical analysis is performed on the operation data within each time interval to obtain the statistical results of the performance metric within each time interval.

[0080] In one embodiment, for each performance metric, machine learning algorithms are used to extract features from the operation data within each time interval to obtain the time-series characteristic data of the performance metric within each time interval.

[0081] In one embodiment, for each performance metric, machine learning algorithms are used to perform portrait analysis on the operation data within an identification period to obtain the portrait data of the target instance.

[0082] In the above technical solution, the machine learning algorithms are trained so that the machine learning algorithms can perform corresponding data mining to obtain time-series characteristic data or portrait data. The machine learning algorithms include supervised learning, unsupervised learning, reinforcement learning, etc.

[0083] For example: The operation data of each performance metric of a certain target instance includes CPU usage, memory usage, ingress network rate, and egress network rate.

[0084] First, statistical analysis is performed on the CPU usage, memory usage, ingress network rate, and egress network rate of the target instance to obtain the statistical results of the CPU, memory, ingress network rate, and egress network rate. The statistical analysis process includes calculating the minimum value, calculating the average value, calculating the maximum value, calculating the statistical distributions of 25%, 75%, and 95%, etc.

[0085] Machine learning algorithms are used to perform performance metric portraits on the CPU usage, memory usage, ingress network rate, and egress network rate of the target instance, and after in-depth mining, the portrait data of the target instance within the identification period is obtained. The portrait data is a compute-intensive instance, a storage-intensive instance, a read-write-intensive instance, etc. When the characteristic data of the performance metric includes portrait data, the portrait data of each performance metric inherits the portrait data of the target instance.

[0086] Using machine learning algorithms, extract features from the operation data of the target instance and calculate the changes in the operation data of the instance within each time interval. For example, within a certain time interval, the high-water mark operation ratio of the CPU is 20%, the low-water mark operation ratio of the CPU is 30%, and the idle operation ratio of the CPU is 50%. Within a certain time interval, the high-water mark operation ratio of the memory is 15%, the low-water mark operation ratio of the memory is 45%, and the idle operation ratio of the memory is 40%.

[0087] S203. Use machine learning algorithms to process the feature data of each performance metric to obtain the weight of each performance metric.

[0088] In this step, use machine learning algorithms to process the statistical results of the performance metrics within each time interval, the temporal feature data of the performance metrics within each time interval, the portrait data of the target instance, and the preset initial weights of each performance metric to obtain the weights of each performance metric.

[0089] Among them, the preset initial weight of each performance metric is obtained by processing the historical feature data of each performance metric using the entropy weight method. The historical feature data of each performance metric is obtained by extracting features from the operation data within the historical recognition period. Use the historical feature data and preset initial weights of each performance metric to train the machine learning algorithm, and use the trained machine learning algorithm to process the feature data of each performance metric to obtain the weights of each performance metric.

[0090] S204. Calculate the performance index of the target instance based on the weights of each performance metric and the feature data of the performance metrics.

[0091] In this step, the statistical results, temporal feature data, and portrait data of each performance metric are in different data dimensions. When obtaining the performance index of the target instance, it is necessary to first perform normalization processing on the statistical results, temporal feature data, and portrait data of each performance metric to unify the data to the same dimension. For example: all unified to the data range of 0-100.

[0092] In one embodiment, for each time interval, perform normalization processing on the statistical results, temporal feature data, and portrait data of each performance metric. Add the normalized statistical results, temporal feature data, and portrait data to obtain the metric data of each performance metric within each time interval.

[0093] After obtaining the metric data of each performance metric within each time interval, use the weights of each performance metric to perform weighted summation processing on the metric data of each performance metric within each time interval to obtain the performance index of the target instance within each time interval.

[0094] For example: Indicates the index data of the i-th performance metric within the time interval t. The performance index of the instance within the time interval t is calculated according to the following formula:

[0095]

[0096] where N represents the total number of performance metrics, P

[0105] , ,

[0100] ,

[0103] , ,

[0104] , ,

[0101] , i ,

[0106] ,

[0102] , , , , , ,

[0098] , ,

[0099] , t , , t ,

[0097] , represents the performance index of the target instance within the time interval t, and α i represents the weight of the i-th performance metric.

[0097] After obtaining the performance index of the target instance in each time interval, a weighted sum is performed on the performance index of the target instance and the preset time weight in each time interval to obtain the performance index of the target instance within the recognition period.

[0098] For example: The performance index of the target instance within the recognition period is calculated according to the following formula:

[0099]

[0100] where β t represents the weight of the t-th time interval, T represents the total number of time intervals, and PZ represents the performance index of the target instance within the recognition period.

[0101] where the preset time weight is set in advance, and the later the time corresponding to the time interval, the greater the weight of the time interval.

[0102] S205. Obtain the operation data of at least one performance metric of the target instance in the previous recognition period and the performance index of the target instance in the previous recognition period.

[0103] In this step, after obtaining the operation data in the previous recognition period, the operation data in the previous recognition period is processed using the steps of S202 to S204 to obtain the performance index of the target instance in the previous recognition period.

[0104] S206. Use a machine learning algorithm to process the operation data of at least one performance metric in the previous recognition period and the performance index of the instance to obtain a first threshold, a second threshold, and a third threshold.

[0105] In this step, after obtaining the operation data of at least one performance metric in the previous recognition period and the performance index of the target instance, a statistical analysis is performed on the operation data of at least one performance metric in the previous recognition period and the performance index of the target instance to obtain a first threshold, a second threshold, and a third threshold.

[0106] More specifically, perform statistical analysis on the operation data in the last recognition cycle and the performance indexes of the target instances in each time interval to determine the distribution of the operation data and the distribution of the performance indexes, and determine the first threshold, the second threshold, and the third threshold based on the distribution of the operation data and the distribution of the performance indexes.

[0107] S207. Obtain the usage status of the target instance according to the performance index of the target instance and the preset status conditions.

[0108] In this step, meeting the idle state condition specifically includes that the performance index of the target instance is less than the first threshold, and not meeting the idle state condition specifically includes that the performance index of the target instance is greater than or equal to the first threshold.

[0109] Meeting the low-load condition specifically includes that the performance index of the target instance is greater than or equal to the first threshold and less than the second threshold. Not meeting the low-load condition specifically includes that the performance index of the target instance is less than the first threshold, or the performance index of the target instance is greater than or equal to the second threshold.

[0110] Meeting the normal-load condition specifically includes that the performance index of the target instance is greater than or equal to the second threshold and less than the third threshold. Not meeting the normal-load condition specifically includes that the performance index of the target instance is less than the second threshold, or the performance index of the target instance is greater than or equal to the third threshold.

[0111] Meeting the high-load condition specifically includes that the performance index of the target instance is greater than or equal to the third threshold. Not meeting the high-load condition specifically includes that the performance index of the target instance is less than the third threshold.

[0112] When the performance index of the target instance is less than the first threshold, determine that the target instance is in the idle state. When the performance index of the target instance is greater than or equal to the first threshold and less than the second threshold, determine that the target instance is in the low-load state. When the performance index of the target instance is greater than or equal to the second threshold and less than the third threshold, determine that the target instance is in the normal-load state. When the performance index of the target instance is greater than or equal to the third threshold, determine that the target instance is in the high-load state. Among them, the first threshold is less than the second threshold, and the second threshold is less than the third threshold.

[0113] In the above technical solution, the operation data of the target instance is condensed into an index representing the usage of the overall performance metrics of the target instance, i.e., the performance index of the target instance, by using a machine learning algorithm. Subsequently, based on the performance index of the target instance, intelligent identification of the usage status of the target instance is achieved. By performing data mining on the operation data of at least one performance metric in the previous identification cycle and the performance index of the target instance, a first threshold, a second threshold, and a third threshold are obtained. There is no need for manual threshold setting, and the thresholds can also adapt to changes in the operation data, which can improve the accuracy of the identification results.

[0114] As Figure 4 shown, an identification device 300 for the usage status of an instance provided in an embodiment of the present application includes:

[0115] An acquisition module 301, configured to acquire the operation data of at least one performance metric of a target instance in an identification cycle;

[0116] A processing module 302, configured to perform feature extraction on the operation data in an identification cycle for each performance metric to obtain the feature data of each performance metric;

[0117] The processing module 302 is further configured to calculate the performance index of the target instance from the feature data of the performance metrics;

[0118] The processing module 302 is further configured to obtain the usage status of the target instance based on the performance index of the target instance.

[0119] In one embodiment, each identification cycle includes multiple time intervals; the processing module 302 is further configured to:

[0120] For each performance metric, perform statistical analysis on the operation data in each time interval to obtain the statistical result of the performance metric in each time interval;

[0121] For each performance metric, perform feature extraction on the operation data in each time interval by using a machine learning algorithm to obtain the time series feature data of the performance metric in each time interval;

[0122] For each performance metric, perform portrait analysis on the operation data in an identification cycle by using a machine learning algorithm to obtain the portrait data of the target instance.

[0123] In one embodiment, the processing module 302 is specifically configured to:

[0124] Use a machine learning algorithm to process the statistical results of the performance metrics in each time interval, the time series feature data of the performance metrics in each time interval, the portrait data of the target instance, and the preset initial weight of each performance metric to obtain the weight of each performance metric.

[0125] In one embodiment, the processing module 302 is specifically configured to:

[0126] For each time interval, perform normalization processing on the statistical results, temporal feature data, and portrait data of each performance metric; add the normalized statistical results, temporal feature data, and portrait data to obtain the metric data of each performance metric within each time interval;

[0127] Use the weight of each performance metric to perform weighted summation processing on the metric data of each performance metric within each time interval to obtain the performance index of the target instance within each time interval;

[0128] Perform weighted summation on the performance index of the target instance within each time interval and the preset time weight to obtain the performance index of the target instance within the recognition period.

[0129] In one embodiment, the processing module 302 is specifically configured to:

[0130] When the performance index of the target instance is less than the first threshold, determine that the target instance is in an idle state;

[0131] When the performance index of the target instance is greater than or equal to the first threshold and less than the second threshold, determine that the target instance is in a low-load state;

[0132] When the performance index of the target instance is greater than or equal to the second threshold and less than the third threshold, determine that the target instance is in a normal-load state;

[0133] When the performance index of the target instance is greater than or equal to the third threshold, determine that the target instance is in a high-load state.

[0134] In one embodiment, the processing module 302 is further configured to:

[0135] Obtain the operation data of at least one performance metric of the target instance in the previous recognition period and the performance index of the target instance in the previous recognition period;

[0136] Use a machine learning algorithm to process the operation data of at least one performance metric in the previous recognition period and the performance index of the target instance to obtain the first threshold, the second threshold, and the third threshold.

[0137] In one embodiment, the acquisition module 301 is specifically configured to:

[0138] Obtain the original operation data of at least one performance metric collected at the monitoring point within one recognition period;

[0139] Perform data aggregation, missing value filling, and outlier cleaning on the raw operation data of at least one performance metric within an identification period to obtain the operation data of at least one performance metric of the target instance within an identification period.

[0140] As Figure 5 shown, an identification device 400 provided in an embodiment of the present application includes a memory 401 and a processor 402.

[0141] Among them, the memory 401 is used to store computer instructions executable by the processor;

[0142] When the processor executes the computer instructions, it implements each step in the method in the above embodiment. For details, please refer to the relevant descriptions in the foregoing method embodiments.

[0143] Optionally, the above memory 401 can be either independent or integrated with the processor 402. When the memory 401 is independently provided, the identification device further includes a bus for connecting the memory 401 and the processor 402.

[0144] An embodiment of the present application further provides a computer-readable storage medium, in which computer instructions are stored. When the processor executes the computer instructions, it implements each step in the method in the above embodiment.

[0145] An embodiment of the present application further provides a computer program product, including computer instructions. When the computer instructions are executed by the processor, it implements each step in the method in the above embodiment.

[0146] Those skilled in the art will readily think of other implementation schemes of the present application after considering the specification and practicing the invention disclosed herein. The present application aims to cover any variations, uses, or adaptive changes of the present application. These variations, uses, or adaptive changes follow the general principles of the present application and include common knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present application are pointed out by the following claims.

[0147] It should be understood that the present application is not limited to the exact structures already described and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A method for identifying instance usage status, characterized in that: include: Obtaining operating data of at least one performance indicator of the target instance within an identification cycle; For each performance indicator, feature extraction is performed on the operating data within an identification cycle to obtain feature data of each performance indicator; Calculating the characteristic data of the performance indicator to obtain a performance index of the target instance; Obtaining a usage status of the target instance according to a performance index of the target instance; Calculating the characteristic data of the performance indicator to obtain the performance index of the target instance specifically includes: Processing the characteristic data of each performance indicator using a machine learning algorithm to obtain a weight for each performance indicator; The performance index of the target instance is obtained by calculation according to the weight of each performance indicator and the characteristic data of each performance indicator.

2. The identification method according to claim 1, characterized in that Each recognition cycle includes multiple time intervals. For each performance indicator, feature extraction is performed on the operating data within an recognition cycle to obtain feature data for each performance indicator, specifically including at least one of the following: For each performance indicator, statistical analysis is performed on the operating data in each time interval to obtain statistical results of the performance indicator in each time interval; For each performance indicator, a machine learning algorithm is used to extract features from the operating data in each time interval to obtain time series feature data of the performance indicator in each time interval; For each performance indicator, a machine learning algorithm is used to perform a portrait analysis on the operating data within an identification cycle to obtain the portrait data of the target instance.

3. The identification method according to claim 1, characterized in that The characteristic data of each performance indicator is processed using a machine learning algorithm to obtain the weight of each performance indicator, specifically including: A machine learning algorithm is used to process the statistical results of the performance indicator in each time interval, the time series feature data of the performance indicator in each time interval, the portrait data of the target instance and the preset initial weight of each performance indicator to obtain the weight of each performance indicator.

4. The identification method according to claim 3, characterized in that Calculating the performance index of the target instance according to the weight of each performance indicator and the characteristic data of each performance indicator specifically includes: For each time interval, the statistical results, time series feature data, and portrait data of each performance indicator are normalized; and the normalized statistical results, time series feature data, and portrait data are added together to obtain the indicator data of each performance indicator in each time interval; The weight of each performance indicator is used to perform weighted summation on the indicator data of each performance indicator in each time interval to obtain the performance index of the target instance in each time interval; A weighted sum is performed on the performance index of the target instance in each time interval and a preset time weight to obtain the performance index of the target instance in the identification period.

5. The identification method according to claim 1 or 2, characterized in that: Obtaining the usage status of the target instance according to the performance index of the target instance specifically includes: When the performance index of the target instance is less than a first threshold, determining that the target instance is in an idle state; When the performance index of the target instance is greater than or equal to the first threshold and less than a second threshold, determining that the target instance is in a low load state; When the performance index of the target instance is greater than or equal to the second threshold and less than a third threshold, determining that the target instance is in a normal load state; When the performance index of the target instance is greater than or equal to the third threshold, it is determined that the target instance is in a high-load state.

6. The identification method according to claim 5, characterized in that The method further comprises: Obtaining operating data of at least one performance indicator of the target instance in a previous identification cycle and a performance index of the target instance in the previous identification cycle; A machine learning algorithm is used to process the operating data of at least one performance indicator in a previous identification cycle and the performance index of the target instance to obtain the first threshold, the second threshold, and the third threshold.

7. The identification method according to any one of claims 1 or 2, characterized in that: Obtain the operating data of at least one performance indicator of the target instance within an identification cycle, including: Acquiring original operating data of at least one performance indicator collected by the monitoring point within the one identification cycle; Data aggregation, missing value filling and outlier cleaning are performed on the original operating data of at least one performance indicator within the one identification cycle to obtain the operating data of at least one performance indicator of the target instance within the one identification cycle.

8. An identification device comprising: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method for identifying the usage status of any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method for identifying the usage status of an example according to any one of claims 1 to 7.

10. A computer program product comprising a computer program, wherein when the computer program is executed by a processor, the method for identifying the usage status of an example according to any one of claims 1 to 7 is implemented.

Images