CN111367779A - Computer performance index abnormity judgment method - Google Patents
Computer performance index abnormity judgment method Download PDFInfo
- Publication number
- CN111367779A CN111367779A CN202010211662.9A CN202010211662A CN111367779A CN 111367779 A CN111367779 A CN 111367779A CN 202010211662 A CN202010211662 A CN 202010211662A CN 111367779 A CN111367779 A CN 111367779A
- Authority
- CN
- China
- Prior art keywords
- performance index
- value
- data
- current moment
- difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3409—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Debugging And Monitoring (AREA)
Abstract
The invention relates to a computer performance index abnormity judging method, which comprises the steps of obtaining a data value of a performance index at the current moment, obtaining a period component corresponding to each historical moment in a past preset time period, calculating an average value of the period components corresponding to the historical moments to obtain a period component average value, calculating a difference absolute value of the period component average value and the data value of the performance index at the current moment, comparing the difference absolute value with an error threshold value, realizing whether the performance index is abnormal or not according to the difference absolute value and the error threshold value, and finally displaying a judgment result of whether the performance index is abnormal or not on a corresponding display screen in a text form. According to the method, whether the performance index at the current moment is abnormal or not is judged according to the magnitude relation between the absolute value of the difference value and the error threshold, automatic abnormal judgment is achieved, the accuracy is greatly improved compared with the existing extensive judgment mode, and the situation of misjudgment is not prone to occurring.
Description
Technical Field
The invention relates to a method for judging the abnormality of computer performance indexes.
Background
With the development of information technology, various applications and websites come out endlessly, and higher requirements are put on the performance of computers. In the running process of the computer, in order to monitor the running state of the computer, each performance index of the computer needs to be monitored, and when abnormal data occurs, relevant countermeasures can be taken in time to avoid more serious faults, so that the monitoring of each performance index of the computer, the real-time acquisition of the running state of the computer and the abnormal judgment are necessary steps for ensuring the normal running of the computer. Generally, monitored performance metrics include CPU utilization, throughput, response time, and the like.
The realization process of the existing abnormity judgment method of the computer performance index is relatively extensive, and the method roughly comprises the following steps: and acquiring index data of the performance index, comparing the index data with a related threshold value, and judging the abnormality according to a comparison result. Although the abnormality determination method can perform abnormality determination, the accuracy of the determination is poor, and erroneous determination is likely to occur.
Disclosure of Invention
The invention aims to provide a computer performance index abnormity judgment method, which is used for solving the problems that the abnormity judgment method is poor in accuracy and misjudgment is easy to occur.
In order to solve the problems, the invention adopts the following technical scheme:
a computer performance index abnormity judgment method comprises the following steps:
acquiring a data value of a performance index at the current moment;
denoising initial time sequence data of performance indexes in a past preset time period through a multilayer self-encoder, and outputting corresponding target time sequence data; the past preset time period comprises a plurality of historical moments corresponding to the current moment; the initial time sequence data is a sequence of data values of the performance indexes arranged according to a time sequence at each historical moment in the past preset time period;
decomposing the target time sequence data to obtain periodic components corresponding to the historical moments respectively;
calculating the average value of the periodic components corresponding to the historical moments respectively to obtain the average value of the periodic components;
calculating the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic components according to the average value of the periodic components and the data value of the performance index at the current moment;
comparing the absolute value of the difference value with an error threshold value;
if the absolute value of the difference is larger than or equal to the error threshold, judging that the performance index at the current moment is abnormal; if the absolute value of the difference is smaller than the error threshold, judging that the performance index at the current moment is not abnormal;
displaying the judgment result of whether the performance index at the current moment is abnormal or not by a corresponding display screen in a text form;
wherein, the calculation process of the error threshold value comprises the following steps:
calculating the difference absolute value of every two data values in the data values of the performance indexes at the historical moments and the data values of the performance indexes at the current moments to obtain corresponding initial difference absolute values;
and calculating the average value of the initial difference absolute values, wherein the obtained average value is the error threshold.
Optionally, the denoising, by the multi-layer self-encoder, the initial time series data of the performance index in the past preset time period, and outputting corresponding target time series data includes:
inputting the initial timing data into the multi-layer self-encoder;
carrying out multilayer hidden layer coding on the initial time sequence data through a coder in the multilayer self-coder to obtain a low-dimensional feature vector;
and performing multilayer hidden layer decoding on the low-dimensional feature vector through a decoder in the multilayer self-encoder, and outputting the target time sequence data.
Optionally, the decomposing the target time series data to obtain periodic components corresponding to the historical moments respectively includes:
generating a periodic subsequence corresponding to each historical moment according to the target time sequence data;
performing smooth regression on each periodic subsequence to obtain a smooth result corresponding to each periodic subsequence;
and removing the low-pass quantity in each smoothing result to obtain the periodic component corresponding to each historical moment.
The invention has the beneficial effects that: the method comprises the steps that initial time sequence data of performance indexes in a past preset time period are subjected to noise reduction through a multilayer self-encoder, corresponding target time sequence data are output, then the target time sequence data are decomposed to obtain periodic components corresponding to historical moments respectively, the average value of the periodic components is obtained through calculation, the average value of the difference value of the initial time sequence data and the average value of the periodic components is obtained through calculation in combination with the data value of the performance indexes at the current moment and the average value of the periodic components, whether the performance indexes at the current moment are abnormal or not is judged according to the size relation between the absolute value of the difference value and an error threshold value, automatic abnormal judgment is achieved, and compared with an existing rough judgment mode, accuracy is greatly improved, and misjudgment is not; moreover, the error threshold is not set randomly, but is determined by the data value of the performance index at each historical moment and the data value of the performance index at the current moment, and the judgment is performed by combining the error threshold, so that the judgment accuracy can be improved, and the error threshold is not a fixed value, but is changed along with the time, so that the effectiveness of the error threshold is improved, and the accuracy of the abnormal judgment is further improved; in addition, the initial time sequence data is denoised by the multilayer self-encoder, so that the degree of noise interference in the periodic component decomposition process can be reduced, the obtained periodic component is more accurate, the accuracy of the average value of the periodic component is ensured, the accuracy of the absolute value of the difference value of the data value of the performance index at the current moment is ensured, and the accuracy of abnormal judgment is finally ensured.
Drawings
In order to more clearly illustrate the technical solution of the embodiment of the present invention, the drawings needed to be used in the embodiment will be briefly described as follows:
FIG. 1 is a flow chart of a method for determining abnormality of computer performance indicators.
Detailed Description
The embodiment provides a method for judging abnormality of computer performance indexes, wherein the common performance indexes of a computer include: CPU utilization, throughput, response time, average number of threads in a run queue within a particular time interval, disk space usage, memory usage, network traffic, and the like. For those skilled in the art, the above listed performance indexes are not limited, and the present application does not require any limiting requirement for the performance indexes. The execution main body of the computer performance index abnormality judgment method can be a computer (such as a notebook computer and a desktop computer) or a background server (such as an independent server and a cluster server). In this embodiment, an execution subject of the method for determining the abnormality of the performance index of the computer is exemplified by a computer.
As shown in fig. 1, the method for determining the abnormality of the computer performance index includes the following steps:
acquiring a data value of the performance index at the current moment:
and the current moment is the current data acquisition moment, and the data value of the performance index of the computer at the current moment is acquired.
Denoising initial time sequence data of performance indexes in a past preset time period through a multilayer self-encoder, and outputting corresponding target time sequence data; the past preset time period comprises a plurality of historical moments corresponding to the current moment; the initial time sequence data is a sequence of data values of performance indexes arranged according to a time sequence at each historical time within the past preset time period:
the past preset time period is set by actual needs, such as: the past preset time period is the past 3 weeks. The past preset time period includes a plurality of (i.e. at least two) historical times corresponding to the current time, and the correspondence refers to dividing the past preset time period into sub-time periods equal to the number of the historical times, where the historical time and the current time in each sub-time period are the same time, for example: the current time is 10 am of the current day, the sub-period is one day, and each historical time is 10 am of each day in the past preset period.
The initial time series data is a sequence of data values of the performance index arranged in time sequence at each historical time within a past preset time period, such as: the initial time series data is a data value sequence obtained by arranging data values of performance indexes of 10 am every day in the last 3 weeks according to a time sequence, and specifically comprises the following steps: data values of the performance index at 10 am on the first day, data values of the performance index at 10 am on the next day, … …, and data values of the performance index at 10 am on the last day.
In the present embodiment, the multi-layer self-encoder is a self-encoder including an encoder and a decoder and having a symmetric structure of convolutional layer-self-encoding algorithm-deconvolution layer. Further, the multi-layer self-encoder is trained according to time sequence data containing the performance index of the preset noise. In this embodiment, in order that the multi-layer self-encoder can well simulate the noise of the initial timing sequence data in the noise reduction process, the timing sequence data added with the preset noise is adopted as a training sample during the training of the multi-layer self-encoder. Wherein the predetermined noise may be Gaussian noise conforming to Gaussian distribution (normal distribution), that is
Where x is gaussian noise, which includes but is not limited to 0-valued noise and high-valued noise.
The method comprises the steps of inputting initial time sequence data into a multilayer self-encoder, conducting multilayer hidden layer encoding on the initial time sequence data through an encoder in the multilayer self-encoder to obtain a low-dimensional characteristic vector, conducting multilayer hidden layer decoding on the low-dimensional characteristic vector through a decoder in the multilayer self-encoder, and outputting target time sequence data, wherein the obtained target time sequence data are time sequence data after noise in the initial time sequence data is removed.
Decomposing the target time sequence data to obtain periodic components respectively corresponding to the historical moments:
based on "decomposing the target time series data to obtain periodic components corresponding to the historical moments respectively", a specific implementation process is given as follows:
and generating a periodic subsequence corresponding to each historical moment according to the target time sequence data. The period subsequence is a subsequence formed by sample points at the same position in each period in the target time sequence data. Such as: the time length of the target time sequence data is 3 weeks, the period is 1 day, and the index data corresponding to the same time every day is the sample point at the same position in the same period. For example, the target time series data is data of the previous 3 weeks, and data at the time of 10 o ' clock each day is grouped into a periodic subsequence (A, B, …, N, where data a is data at 10 o ' clock on the first day, data B is data at 10 o ' clock on the second day, and so on to data N on the last day). In this embodiment, the target time series data is decomposed into periodic components by an STL (secure-future decomposition procedure based loss) algorithm. The STL algorithm decomposes the time series data Yv into a trend component (tend component), a periodic component (periodic component), and a remainder (remainderiscomponent) based on the loses: yv = Tv + Sv + Rv, v =1~ N. The present embodiment employs an inner loop to perform trend fitting and calculation of periodic components.
And performing smooth regression on each periodic subsequence to obtain a smooth result corresponding to each periodic subsequence. Presence in setting target timing data
A periodic subsequence of LOESS: (
) Performing smooth regression on each period subsequence, namely, extending each period subsequence forward and backward for one period respectively to obtain a smooth result
，
Wherein
For the smoothing parameter of the LOESS smoothing regression, k represents the kth pass in the inner loop.
And removing the low flux in each smoothing result to obtain periodic components corresponding to each historical moment. Extracting smoothing results
Medium low flux
，
: for the smooth result
Do in turn
、
And 3, obtaining an average result by moving average (movingaverage), and adopting LOESS (A), (B), (C), (
) Performing smooth regression on the average result; removing smoothing results
Medium low flux
Obtaining a periodic component
。
Calculating the average value of the periodic components corresponding to the historical moments respectively to obtain the periodic component average value:
and obtaining the periodic components corresponding to the historical moments, and calculating the average value of the periodic components corresponding to the historical moments, wherein the average value is the average value of the periodic components.
Calculating the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic components according to the average value of the periodic components and the data value of the performance index at the current moment:
and calculating the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic components according to the obtained average value of the periodic components and the data value of the performance index at the current moment, namely calculating the difference between the data value of the performance index at the current moment and the average value of the periodic components, and then taking the absolute value.
Comparing the absolute value of the difference with an error threshold value:
comparing the obtained absolute value of the difference with an error threshold, wherein the calculation process of the error threshold comprises the following steps: calculating a difference absolute value between every two data values in the data values of the performance index at each historical moment and the data values of the performance index at the current moment, wherein the obtained difference absolute value is an initial difference absolute value, for example: assuming that the number of the data values of the performance index at each historical time is N, then the total of the data values of the performance index at each historical time and the data values of the performance index at the current time is N +1 data values, and the difference absolute value of every two data values in the N +1 data values is calculated, so that an initial difference absolute value can be obtained; then, the average value of the absolute values of the initial differences is calculated, and the obtained average value is an error threshold value. Then, when the current time is updated, the number of data values of the performance index at the historical time changes, the data value of the performance index at the current time also changes, and the finally calculated error threshold value also changes.
If the absolute value of the difference is larger than or equal to the error threshold, judging that the performance index at the current moment is abnormal; if the absolute value of the difference is smaller than the error threshold, judging that the performance index at the current moment is not abnormal:
if the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic component is greater than or equal to the error threshold, the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic component is larger, and the difference between the data value of the performance index at the current moment and the average value of the periodic component is larger, the performance index at the current moment is judged to be abnormal; correspondingly, if the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic component is smaller than the error threshold, the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic component is smaller, and the difference between the data value of the performance index at the current moment and the average value of the periodic component is smaller, it is determined that the performance index at the current moment is not abnormal.
And displaying the judgment result of whether the performance index at the current moment is abnormal or not by a corresponding display screen in a text form:
in order to facilitate the operation and maintenance personnel to intuitively know whether the performance index of the computer at the current moment is abnormal, the judgment result of whether the performance index of the computer at the current moment is abnormal is displayed by a corresponding display screen (namely the display screen of the computer) in a text form, for example: a text box appears on the display screen, and the text box has a judgment result of whether the performance index at the current moment is abnormal, such as: "the performance index at the current time is abnormal" or "the performance index at the current time is not abnormal".
The above-mentioned embodiments are merely illustrative of the technical solutions of the present invention in a specific embodiment, and any equivalent substitutions and modifications or partial substitutions of the present invention without departing from the spirit and scope of the present invention should be covered by the claims of the present invention.
Claims (3)
1. A computer performance index abnormity judgment method is characterized by comprising the following steps:
acquiring a data value of a performance index at the current moment;
denoising initial time sequence data of performance indexes in a past preset time period through a multilayer self-encoder, and outputting corresponding target time sequence data; the past preset time period comprises a plurality of historical moments corresponding to the current moment; the initial time sequence data is a sequence of data values of the performance indexes arranged according to a time sequence at each historical moment in the past preset time period;
decomposing the target time sequence data to obtain periodic components corresponding to the historical moments respectively;
calculating the average value of the periodic components corresponding to the historical moments respectively to obtain the average value of the periodic components;
calculating the absolute value of the difference between the data value of the performance index at the current moment and the average value of the periodic components according to the average value of the periodic components and the data value of the performance index at the current moment;
comparing the absolute value of the difference value with an error threshold value;
if the absolute value of the difference is larger than or equal to the error threshold, judging that the performance index at the current moment is abnormal; if the absolute value of the difference is smaller than the error threshold, judging that the performance index at the current moment is not abnormal;
displaying the judgment result of whether the performance index at the current moment is abnormal or not by a corresponding display screen in a text form;
wherein, the calculation process of the error threshold value comprises the following steps:
calculating the difference absolute value of every two data values in the data values of the performance indexes at the historical moments and the data values of the performance indexes at the current moments to obtain corresponding initial difference absolute values;
and calculating the average value of the initial difference absolute values, wherein the obtained average value is the error threshold.
2. The method for determining abnormality in computer performance index according to claim 1, wherein the denoising, by a multi-layer self-encoder, of the initial time series data of the performance index in a past preset time period and outputting corresponding target time series data includes:
inputting the initial timing data into the multi-layer self-encoder;
carrying out multilayer hidden layer coding on the initial time sequence data through a coder in the multilayer self-coder to obtain a low-dimensional feature vector;
and performing multilayer hidden layer decoding on the low-dimensional feature vector through a decoder in the multilayer self-encoder, and outputting the target time sequence data.
3. The method of claim 1, wherein decomposing the target time series data to obtain periodic components corresponding to the historical moments comprises:
generating a periodic subsequence corresponding to each historical moment according to the target time sequence data;
performing smooth regression on each periodic subsequence to obtain a smooth result corresponding to each periodic subsequence;
and removing the low-pass quantity in each smoothing result to obtain the periodic component corresponding to each historical moment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010211662.9A CN111367779A (en) | 2020-03-24 | 2020-03-24 | Computer performance index abnormity judgment method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010211662.9A CN111367779A (en) | 2020-03-24 | 2020-03-24 | Computer performance index abnormity judgment method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111367779A true CN111367779A (en) | 2020-07-03 |
Family
ID=71210596
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010211662.9A Withdrawn CN111367779A (en) | 2020-03-24 | 2020-03-24 | Computer performance index abnormity judgment method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111367779A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112346939A (en) * | 2020-11-02 | 2021-02-09 | 长沙市到家悠享网络科技有限公司 | Alarm method, device, equipment and storage medium |
| CN114910810A (en) * | 2021-02-07 | 2022-08-16 | 广州汽车集团股份有限公司 | Fuel cell detection apparatus, method and system |
| CN115001853A (en) * | 2022-07-18 | 2022-09-02 | 山东云天安全技术有限公司 | Abnormal data identification method and device, storage medium and computer equipment |
| CN117961975A (en) * | 2024-03-28 | 2024-05-03 | 法奥意威(苏州)机器人系统有限公司 | Collision detection method and device, storage medium and electronic equipment |
-
2020
- 2020-03-24 CN CN202010211662.9A patent/CN111367779A/en not_active Withdrawn
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112346939A (en) * | 2020-11-02 | 2021-02-09 | 长沙市到家悠享网络科技有限公司 | Alarm method, device, equipment and storage medium |
| CN114910810A (en) * | 2021-02-07 | 2022-08-16 | 广州汽车集团股份有限公司 | Fuel cell detection apparatus, method and system |
| CN115001853A (en) * | 2022-07-18 | 2022-09-02 | 山东云天安全技术有限公司 | Abnormal data identification method and device, storage medium and computer equipment |
| CN115001853B (en) * | 2022-07-18 | 2022-11-04 | 山东云天安全技术有限公司 | Abnormal data identification method and device, storage medium and computer equipment |
| CN117961975A (en) * | 2024-03-28 | 2024-05-03 | 法奥意威(苏州)机器人系统有限公司 | Collision detection method and device, storage medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111367779A (en) | Computer performance index abnormity judgment method | |
| CN111338878A (en) | Anomaly detection method and device, terminal device and storage medium | |
| CN109978379B (en) | Time series data abnormity detection method and device, computer equipment and storage medium | |
| CN108038040B (en) | Computer cluster performance index detection method, electronic equipment and storage medium | |
| WO2021189844A1 (en) | Detection method and apparatus for multivariate kpi time series, and device and storage medium | |
| CN116070163B (en) | Indoor harmful gas concentration anomaly monitoring data processing method | |
| US11736363B2 (en) | Techniques for analyzing a network and increasing network availability | |
| CN110991504A (en) | MOTCN-AE-based multi-dimensional time sequence abnormal value detection method and system | |
| CN113360359B (en) | Index abnormal data tracing method, device, equipment and storage medium | |
| CN110689141A (en) | Fault diagnosis method and equipment for wind generating set | |
| CN112257263A (en) | Equipment residual life prediction system based on self-attention mechanism | |
| WO2020075019A1 (en) | Prediction model enhancement | |
| CN107273234A (en) | A kind of time series data rejecting outliers and bearing calibration based on EEMD | |
| CN111459746A (en) | Alarm generation method and device, electronic equipment and readable storage medium | |
| CN108491875A (en) | A kind of data exception detection method, device, equipment and medium | |
| CN117786371B (en) | Temperature monitoring data optimization prediction analysis method and system | |
| CN115456107A (en) | Time series abnormity detection system and method | |
| Kumar | Conditional analysis of Phase II exponential chart for monitoring times to an event | |
| CN111666187A (en) | Method and apparatus for detecting abnormal response time | |
| CN116562120A (en) | RVE-based turbine engine system health condition assessment method and RVE-based turbine engine system health condition assessment device | |
| CN110781781A (en) | Time series concept drift detection method, system, medium and equipment | |
| CN112150447B (en) | Abnormal data determination method and device based on location-based service and electronic equipment | |
| US20180039718A1 (en) | Prediction of inhalable particles concentration | |
| CN116226770A (en) | Time sequence data anomaly detection method and device | |
| CN112162528B (en) | Fault diagnosis method, device, equipment and storage medium of numerical control machine tool |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200703 |
|
| WW01 | Invention patent application withdrawn after publication |