CN104598361B - A kind of method for monitoring performance and device - Google Patents
A kind of method for monitoring performance and device Download PDFInfo
- Publication number
- CN104598361B CN104598361B CN201310535299.6A CN201310535299A CN104598361B CN 104598361 B CN104598361 B CN 104598361B CN 201310535299 A CN201310535299 A CN 201310535299A CN 104598361 B CN104598361 B CN 104598361B
- Authority
- CN
- China
- Prior art keywords
- alarm
- sequence
- statistics amount
- threshold range
- observation
- 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.)
- Active
Links
Landscapes
- Alarm Systems (AREA)
- Measurement Of Radiation (AREA)
Abstract
The embodiment of the present invention provides a kind of method for monitoring performance and device, is related to computer field, can be improved the accuracy of abnormality detection, comprising: detection device records the observation of Current observation point into the sequence of observations;The First ray of the top n observation composition of the observation of the Current observation point is obtained in the sequence of observations;Threshold parameter is calculated further according to the First ray;The corresponding first alarm threshold range of the first alarm statistics amount is obtained according to the threshold parameter K.The embodiment of the present invention provides a kind of method for monitoring performance and device for performance detection.
Description
Technical field
The present invention relates to computer field more particularly to a kind of method for monitoring performance and device.
Background technique
With becoming increasingly popular for cloud computing virtualization technology, management and the monitoring of virtualized infrastructure become increasingly to weigh
It wants.Abnormality detection is exactly to monitor in real time to the various performance indicators of virtualization facility, finds the abnormal feelings in environment in time
Condition, and issue alarm notification, to remind operation maintenance personnel to take measures to keep the normal operation of virtualization facility.
In the prior art, commonly used approach is to establish the autoregression model of the observation of observation point, autoregression model
It is that following situation is predicted according to the rule of independent variable itself.Positive statistic is calculated by calculated top n observation point
Average and standard deviation, negative statistic average and standard deviation and by empirical value estimate threshold parameter K, obtain statistic
The first alarm threshold range, calculate current statistic amount further according to the observation of Current observation point, judge that current statistic amount is
No the first alarm threshold range in statistic, so that whether the observation for detecting Current observation point is abnormal.
In the prior art, the selection of threshold parameter K be rule of thumb come the empirical value being arranged, but monitor control index it is thousands of on
Ten thousand, the sequence of observations difference of various indexs is very big, it is impossible to which each index by virtue of experience gives a suitable K value, leads
Causing the accuracy of abnormality detection reduces.
Summary of the invention
The embodiment of the present invention provides a kind of method and apparatus of performance monitoring, can be improved the accuracy of abnormality detection.
In order to achieve the above objectives, the embodiment of the present invention adopts the following technical scheme that
In a first aspect, providing a kind of method for monitoring performance, comprising:
The observation of Current observation point is recorded into the sequence of observations;
The first sequence of the top n observation composition of the observation of the Current observation point is obtained in the sequence of observations
Column, the N are the integer more than or equal to 2;
Threshold parameter K is calculated according to the First ray;
The corresponding first alarm threshold range H of the first alarm statistics amount λ is obtained according to the threshold parameter K1。
With reference to first aspect, the first can be described that threshold value ginseng is calculated according to the First ray in realization mode
Counting K includes:
Obtain the standard deviation sigma of the First rayo;
The direct proportion function a, a for obtaining the period of waves of the First ray are greater than 1;
According to the standard deviation sigmaoThe threshold parameter K is calculated with the direct proportion function a, so that the threshold value is joined
Number K meets:
Mode can be realized with the first with reference to first aspect, in second of achievable mode,
Described obtained according to the threshold parameter K accuses the corresponding first alarm threshold range H of the first police statistic λ1Include:
Obtain statistic sequence { λi};
According to the statistic sequence { λiAnd threshold parameter K acquisition the first alarm threshold range H1, so that institute
State the first alarm threshold range H1Are as follows:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation.
In conjunction with second of achievable mode, can be obtained described according to the threshold parameter K in realization mode at the third
The corresponding first alarm threshold range H of first alarm statistics amount λ1Later, the method also includes:
Obtain the first alarm statistics amount λ of the Current observation point;
Judge the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, issue alarm signal.
With reference to first aspect, the first to second achievable mode, in the 4th kind of achievable mode, described according to institute
It states threshold parameter K and obtains the corresponding first alarm threshold range H of the first alarm statistics amount λ1Later, the method also includes:
Obtain the first alarm statistics amount λ of the Current observation point;
Judge the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, in the sequence of observations
Delete the observation of the Current observation point;
Calculate the second alarm statistics amount φt;
Calculate the second alarm statistics amount φtCorresponding second alarm threshold range H2;
Judge the second alarm statistics amount φtWhether in the second alarm threshold range H2It is interior;
As the second alarm statistics amount φtNot in the second alarm threshold range H2It is interior, issue alarm signal.
In conjunction with the 4th kind of achievable mode, in the 5th kind of achievable mode,
The second alarm statistics amount φ of the calculatingtInclude:
Obtain the first alarm statistics amount λ of the Current observation point;
Obtain statistic sequence { λi};
According to the first alarm statistics amount λ and the statistic sequence { λiObtain the abnormal amount of the Current observation point
δt, the δtMeet:
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation, b is default sampling time interval function.
In conjunction with the 5th various achievable modes, in the 6th kind of achievable mode, the second alarm system is calculated described
Measure φtCorresponding second alarm threshold range H2Before, the method also includes:
Initialization exception observation point parameter d=0;
If there is the first alarm statistics amount λ not in the first alarm threshold range H1Interior observation point, will be current
D add 1 to obtain new d;
If current d is not equal to 0, and the first alarm statistics amount λ occurs in the first alarm threshold range H1Interior
Current d is subtracted 1 and obtains new d by observation point.
It is described to calculate the second alarm statistics amount in the 7th kind of achievable mode in conjunction with the 6th kind of achievable mode
φtCorresponding second alarm threshold range H2Include:
Obtain current Outliers point parameter d;
According to the threshold parameter K, all Outliers point parameter d, calculated by alarm statistics amount threshold formula
Second alarm statistics amount φtCorresponding second alarm threshold range H2, the alarm statistics amount threshold formula are as follows:
In second aspect, a kind of performance monitoring apparatus is provided, comprising:
Recording unit, for recording the observation of Current observation point into the sequence of observations;
First acquisition unit, the top n of the observation for obtaining the Current observation point in the sequence of observations
The First ray of observation composition, the N are the integer more than or equal to 2;
First computing unit, for threshold parameter K to be calculated according to the First ray;
Second acquisition unit, for obtaining the corresponding first alarm threshold of the first alarm statistics amount λ according to the threshold parameter K
It is worth range H1。
In conjunction with second aspect, the first can be described that threshold value ginseng is calculated according to the First ray in realization mode
Counting K includes:
Obtain the standard deviation sigma of the First rayo;
The direct proportion function a, a for obtaining the period of waves of the First ray are greater than 1;
According to the standard deviation sigmaoThe threshold parameter K is calculated with the direct proportion function a, so that the threshold value is joined
Number K meets:
In conjunction with second aspect, the first can realize mode, in second of achievable mode,
The second acquisition unit is specifically used for:
Obtain statistic sequence { λi};
According to the statistic sequence { λiAnd threshold parameter K acquisition the first alarm threshold range H1, so that institute
State the first alarm threshold range H1Are as follows:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation.
In conjunction with second of achievable mode, the third can be in realization mode, the performance monitoring apparatus further include:
Third acquiring unit, for obtaining the first alarm statistics amount λ of the Current observation point;
First judging unit, for judging the first alarm statistics amount λ whether in the first alarm threshold range H1
It is interior;
First Alarm Unit is used in the first alarm statistics amount λ not in the first alarm threshold range H1When interior,
Issue alarm signal.
In conjunction with second aspect, the first to second achievable mode, in the 4th kind of achievable mode, the performance monitoring
Device further include:
4th acquiring unit, for obtaining the first alarm statistics amount λ of the Current observation point;
Second judgment unit, for judging the first alarm statistics amount λ whether in the first alarm threshold range H1
It is interior;
Unit is deleted, if for the first alarm statistics amount λ not in the first alarm threshold range H1It is interior, described
The observation of the Current observation point is deleted in the sequence of observations;
Second computing unit, for calculating the second alarm statistics amount φt;
Third computing unit, for calculating the second alarm statistics amount φtCorresponding second alarm threshold range H2;
Third judging unit, for judging the second alarm statistics amount φtWhether in the second alarm threshold range H2
It is interior;
Second Alarm Unit, in the second alarm statistics amount φtNot in the second alarm threshold range H2It is interior
When, issue alarm signal.
In conjunction with the 4th kind of achievable mode, in the 5th kind of achievable mode,
Second computing unit is specifically used for:
Obtain the first alarm statistics amount λ of the Current observation point;
Obtain statistic sequence { λi};
According to the first alarm statistics amount λ and the statistic sequence { λiObtain the abnormal amount of the Current observation point
δt, the δtMeet:
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation, b is default sampling time interval function.
In conjunction with the 5th various achievable modes, in the 6th kind of achievable mode, the performance monitoring apparatus further include:
Initialization unit is used for initialization exception observation point parameter d=0;
Processing unit is used for: the first alarm statistics amount λ is occurring not in the first alarm threshold range H1Interior
When observation point, 1 is added to obtain new d current d;
It is not equal to 0 in current d, and the first alarm statistics amount λ occurs in the first alarm threshold range H1Interior
When observation point, current d is subtracted 1 and obtains new d.
In conjunction with the 6th kind of achievable mode, in the 7th kind of achievable mode, the third computing unit is specifically used for:
Obtain current Outliers point parameter d;
According to the threshold parameter K, all Outliers point parameter d, calculated by alarm statistics amount threshold formula
Second alarm statistics amount φtCorresponding second alarm threshold range H2, the alarm statistics amount threshold formula are as follows:
The embodiment of the present invention provides a kind of method for monitoring performance and device, comprising: detection device records Current observation point
Observation into the sequence of observations;The top n that the observation of the Current observation point is obtained in the sequence of observations is seen
The First ray of measured value composition;Threshold parameter is calculated further according to the First ray;The is obtained according to the threshold parameter
The corresponding first alarm threshold range of one alarm statistics amount.So, threshold parameter changes according to the variation of First ray,
To keep calculated first alarm threshold range more accurate, the accuracy of abnormality detection is improved.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is a kind of flow chart of method for monitoring performance provided in an embodiment of the present invention;
Fig. 2 is the flow chart of another method for monitoring performance provided in an embodiment of the present invention;
Fig. 3 be in the prior art the first alarm threshold range come the whether abnormal method of observation that judges Current observation point
Schematic diagram;
Fig. 4 is the schematic diagram of the first alarm statistics amount judgment method provided in an embodiment of the present invention;
Fig. 5 is the schematic diagram of the second alarm statistics amount judgment method provided in an embodiment of the present invention
Fig. 6 is a kind of structural schematic diagram of performance monitoring apparatus provided in an embodiment of the present invention;
Fig. 7 is the structural schematic diagram of another performance monitoring apparatus provided in an embodiment of the present invention;
Fig. 8 is the structural schematic diagram of another performance monitoring apparatus provided in an embodiment of the present invention;
Fig. 9 is the structural schematic diagram of another performance monitoring apparatus provided in an embodiment of the present invention;
Figure 10 is structural schematic diagram provided in an embodiment of the present invention and another performance monitoring apparatus.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The embodiment of the invention provides a kind of method for monitoring performance, are applied to detection device, such as cloud computing infrastructure
Deng as shown in Figure 1:
Step 101 records the observation of Current observation point into the sequence of observations.
The first sequence that step 102, the top n observation for the observation that Current observation point is obtained in the sequence of observations form
Column, the N are the integer more than or equal to 2.
Threshold parameter K is calculated according to First ray for step 103.
Firstly, detection device obtains the standard deviation sigma of First rayo, secondly, detection device obtains the fluctuation week of First ray
The direct proportion function a, a of phase is greater than 1;Finally, according to the standard deviation sigma of the First rayoWith the wave of the First ray
The direct proportion function a in dynamic period, is calculated the threshold parameter K, so that the threshold parameter K meets:
Particularly, the value range of K be [2,3), observation fluctuation it is more violent, standard deviation is bigger, and the period is smaller, K value
More level off to 3, observation fluctuation it is gentler, standard deviation is smaller, and the period is bigger, and K value more levels off to 2.It is got in observation
Violent place, threshold range become wider;When observation is gentler, threshold range becomes narrower.
Step 104 obtains the corresponding first alarm threshold range H of the first alarm statistics amount λ according to threshold parameter K1。
Firstly, detection device obtains statistic sequence { λi}。
Secondly, detection device is according to statistic sequence { λiAnd First ray threshold parameter K obtain the first alarm threshold
Range H1, so that the first alarm threshold range H1Are as follows:
Wherein,For statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence { λi +}
Standard deviation,For statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence { λi -Mark
It is quasi- poor.
Again, detection device obtains the first alarm statistics amount λ of Current observation point.
Finally, whether detection device judges the first alarm statistics amount λ in the first alarm threshold range H1It is interior.
So, detection device record calculates threshold parameter K according to the First ray of Current observation value, to obtain
First alarm statistics amount λ and corresponding first alarm threshold range H1, so that whether the observation for detecting Current observation point is abnormal,
Therefore, threshold parameter K changes according to the variation of First ray, to make calculated first alarm threshold range H1It is more quasi-
Really, the accuracy of abnormality detection is improved.
The method of abnormality detection has very much, includes the first alarm statistics amount judgment method and the second alarm system in the present invention
Measure judgment method.
Exemplary, the first alarm statistics amount judgment method may include:
Detection device obtains the measured value of First ray, obtains the first alarm threshold range H1, obtain Current observation point
Whether observation obtains the first alarm statistics amount λ, according to the first alarm statistics amount λ in the first alarm threshold range H1It is interior,
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, it is deleted in the sequence of observations described current
Observation point issues alarm signal;If the first alarm statistics amountAnd the first alarm statistics amountThen
Next observation point of the point of Current observation described in the sequence of observations becomes Current observation point, detects the current sight again
The observation of measuring point.
Exemplary, the second alarm statistics amount judgment method may include:
Firstly, obtaining the first alarm statistics amount λ of Current observation point;Judge whether the first alarm statistics amount λ accuses first
Alert threshold range H1It is interior, if the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, in the observation
The observation of the Current observation point is deleted in sequence.
Secondly, calculating the second alarm statistics amount φt。
Detection device can be according to the first alarm statistics amount λ and statistic sequence { λiObtain Current observation point observation
Abnormal amount δt, the δtMeet:
When the first alarm statistics amount λ is not in the first alarm threshold range H1When interior, the exception of the observation of Current observation point
Measure δtIt is the δtMore than the first alarm threshold range H1Absolute value, when λ is in the first alarm threshold range H1When interior, δtIt is 0.
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein, b is time gap function,For statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For
{λiMiddle on the occasion of sequence { λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For
{λiIn negative value sequence { λi -Standard deviation.
Again, the second alarm statistics amount φ is calculatedtCorresponding second alarm threshold range H2。
Detection device obtains Outliers point parameter d
Initialization exception observation point parameter d=0;If there is the first alarm statistics amount λ not in the first alarm threshold range H1It is interior
Observation point, add 1 to obtain new d current d;If current d is not equal to 0, and the first alarm statistics amount λ occurs described
First alarm threshold range H1Current d is subtracted 1 and obtains new d by interior observation point;If current d=0, and there is the first alarm system
λ is measured in the first alarm threshold range H1Interior observation point keeps d constant.
Specifically, before detecting observation, initialization exception observation point parameter d=0;During detecting observation, sentence
Whether disconnected d is equal to 0, if d is not equal to 0, two kinds of situations occurs in the value of new d: when there is the first alarm statistics amount λ not the
One alarm threshold range H1When interior observation point, 1 is added to obtain new d current d, when there is the first alarm statistics amount λ the
One alarm threshold range H1When interior observation point, current d is subtracted 1 and obtains new d;During detecting observation, if d is equal to
0, there are two kinds of situations in the value of new d: if there is the first alarm statistics amount λ not H within the scope of the first alarm threshold1Sight
Current d is added 1 to obtain new d by measuring point, and the Current observation point is deleted in First ray;If there are the first alarm statistics
λ is measured in the first alarm threshold range H1Interior observation point, d is constant, d=0.
It is exemplary, it is assumed that the corresponding observation of the observation point for needing to detect is y1,y2,......,y7,y8, wherein y2,y5,
y6The first alarm statistics amount not within the scope of corresponding first alarm threshold, y1,y3,y4,y7,y8The first alarm statistics amount exist
Within the scope of corresponding first alarm threshold.
Detection device detection before, initialization exception observation point parameter d=0 starts to detect, using first observation point as
Current observation point detects the observation y of the observation point1, y1The first alarm statistics amount in corresponding first alarm threshold range
Interior, d is constant, d=0;Continue to detect the observation y of the observation point using second observation point as Current observation point2, y2First
For alarm statistics amount not within the scope of corresponding first alarm threshold, current d adds 1 to obtain new d, d=1;By third observation point
As Current observation point, the observation y of the observation point is detected3, current d=1, y3The first alarm statistics amount it is corresponding first accuse
In alert threshold range, current d subtracts 1 and obtains new d, d=0;Using the 4th observation point as Current observation point, the observation is detected
The observation y of point4, current d=0, y4The first alarm statistics amount within the scope of corresponding first alarm threshold, d is constant, d=0;It will
5th observation point detects the observation y of the observation point as Current observation point5, y5The first alarm statistics amount not corresponding
Within the scope of first alarm threshold, current d adds 1 to obtain new d, d=1;Using the 6th observation point as Current observation point, detection
The observation y of the observation point6, y6The first alarm statistics amount not within the scope of corresponding first alarm threshold, current d adds 1
To new d, d=2;Using the 7th observation point as Current observation point, the observation y of the observation point is detected7, current d=2, y7?
Within the scope of corresponding first alarm threshold, current d subtracts 1 and obtains new d, d=1 one alarm statistics amount;By the 8th observation point
As Current observation point, the observation y of the observation point is detected8, current d=1, y8The first alarm statistics amount it is corresponding first accuse
In alert threshold range, current d subtracts 1 and obtains new d, d=0.
According to threshold parameter K, Outliers point parameter d, the second alarm statistics amount φ is calculatedtCorresponding second alarm threshold
Range H2, the alarm statistics amount threshold formula are as follows:
Finally, judging the second alarm statistics amount φtWhether in the second alarm threshold range H2It is interior.
When not alerting currently, if the second alarm statistics amount φtNot in the second alarm threshold range H2It is interior, issue alarm letter
Number;Currently alarm is had occurred and that, if the second alarm statistics amount φtNot in the second alarm threshold range H2It is interior, continue to alert;If the
Two alarm statistics amount φtIn the second alarm threshold range H2It is interior, terminate alarm.
In this way, the second alarm statistics amount, will soon rapidly with exponential increase when continuous multiple test points are abnormal
More than the second alarm threshold range of linear increase;When only a single point is abnormal, i.e., scene of instantaneously leaping high, as long as subsequent do not have
There is the generation of abnormal point, the second alarm statistics amount decays to 0 soon with exponential form, so as to avoid the generation of false-alarm.
The embodiment of the invention provides a kind of method for monitoring performance, are applied to detection device, such as cloud computing infrastructure
Deng as shown in Figure 2, comprising:
Step 201, the observation for recording Current observation point execute step 202 into the sequence of observations.
Assuming that y1,y2,.....,yi-1,yiIt is since the abnormality detection of the index to the sight of current time all observation points
Measured value, the observation of Current observation point are yi, original sequence of observations is all in abnormality detection before being Current observation point
The sequence of the observation composition of the observation point of no exceptions, is (yp,......,yq), detection device records Current observation point
Observation yiTo original sequence of observations, the current sequence of observations (y is constitutedp,......,yq,yi), between p to q
Number can be discontinuous.
The first sequence that step 202, the top n observation for the observation that Current observation point is obtained in the sequence of observations form
Column, the N are the integer more than or equal to 2, execute step 203.
In the sequence of observations (yp,......,yq,yi) in obtain the observation y of the Current observation pointiTop n observation
It is worth the First ray of composition, wherein 2≤N≤q-p.
It is exemplary, it is assumed that the sequence of observations is (y2,y5,y6,y7,y8), it is made of in First ray, observes 3 observations
The observation y of 5th observation point in value sequence8It is the observation of Current observation point.Detection device obtains y8Preceding 3 observations
y5,y6,y7First ray (the y of composition5,y6,y7)。
Threshold parameter K is calculated according to First ray for step 203, executes step 204.
Assuming that N number of measured value of First ray is y1,y2,......,yN-1,yN, the N+1 observation point in the sequence of observations
Observation yN+1Observation as Current observation point.
Firstly, detection device obtains the standard deviation sigma of First rayo
For observation y in the First ray1,y2,......,yN-1,yNAverage value, Mean Value Formulas are as follows:
Wherein, t is 1,2 ..., N-1, N.
The standard deviation of First ray are as follows:
Secondly, direct proportion function a, a that detection device obtains the period of waves of the First ray are greater than 1, first
The formula of the direct proportion function of the period of waves of sequence are as follows:
A=bT,
Wherein, T is the period of waves of the First ray, and b is greater than 0 constant.
According to the standard deviation sigma of the First rayoIt is calculated with the direct proportion function a of the period of waves of the First ray
To threshold parameter K, so that the threshold parameter K meets:
It is worth noting that the value range of K can be [2,3), observation fluctuation it is more violent, standard deviation is bigger, week
Phase is smaller, and K value more levels off to 3, observation fluctuation it is gentler, standard deviation is smaller, and the period is bigger, and K value more levels off to 2.
In the place that observation is more violent, threshold range becomes wider;When observation is gentler, threshold range becomes narrower.
Step 204 obtains the corresponding first alarm threshold range H of the first alarm statistics amount λ1, execute step 205.
Assuming that N number of measured value of First ray is y1,y2,...,yN, the observation of the N+1 observation point in the sequence of observations
Value yN+1Observation as Current observation point.
Firstly, detection device obtains statistic sequence { λi}。
Specifically, calculating the equal difference χ of observationt。
Wherein, t=1,2 ..., N+1
Enable (χ1,χ2,...,χN,χN+1) be the equal difference of observation sequence.
Particularly, observation ytThe range of middle t is 1,2 ..., N+1, the observation ytIt that is include first
The observation y of sequence1,y2,...,yN-1,yNWith the observation y of Current observation pointN+1, obtain the equal difference of observation sequence be for
Acquisition statistic sequence { λiAnd the first alarm statistics amount λ.
Detection device calculates statistic sequence { λ according to autoregression model AR (2)i, comprising:
Autoregression model AR (2) are as follows:
Wherein, t=1,2 ... .., N+1,It is the coefficient of (2) AR, etIt is white noise, the etAverage value be
0, the etMean square deviation be σ2。
By least square method, the estimated value of the coefficient of the AR (2) can be calculated:
According to calculatedIt calculates again and sentences different amount ej, the judgement amount ejFormula are as follows:
Wherein, j=3,4 ... .., N+1
It calculates and sentences different amount e3,e4,......,eN+1, according to the e3,e4,......,eN+1Find out the square of judgement amount
Root σe 2:
Different amount e is sentenced further according to preceding N-23,e4,......,eNWith the root mean square σ of judgement amounte 2, calculate corresponding statistics
Measure λi, the statistic formula λiAre as follows:
Calculate λ1,λ2,......,λn-3As statistic sequence { λiStatistic.
Wherein, i=k-2, k=3,4 ..., N.
Assuming that having n statistic in negative value sequence on the occasion of there is m statistic in sequence.Detection device is by First ray
Statistic sequence { λiBy positive and negative values it is divided into two sequences, it is positive value sequence { λ respectivelyi +And negative value sequence { λi -,For statistics
Measure sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence { λi +Standard deviation,For the statistics
Measure sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence { λi -Standard deviation.And σ+、And σ-
Formula are as follows:
Obtain the first alarm amount threshold range H1, the first alarm amount threshold range H1Formula are as follows:
Again, detection device obtains the first alarm statistics amount λ of the observation of Current observation point.
Detection device obtains eN+1And σe, calculate the first alarm statistics amount λ, the first alarm statistics amount λ formula are as follows:
Wherein, the first alarm statistics amount λ be detection observation whether Yi Chang quantization means.
Whether step 205 judges the first alarm statistics amount λ in the first alarm threshold range H1It is interior.
If the first alarm statistics amount λ is in the first alarm threshold range H1It is interior, execute step 207;
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, Current observation point is deleted in the sequence of observations
yi, execute step 206.
Step 206, the observation y that the Current observation point is deleted in the sequence of observationsi, execute step 207.
It is exemplary, it is assumed that the sequence of observations is (y2,y5,y6,y7,y8), the observation of the 5th observation point in the sequence of observations
y8It is the observation of Current observation point, and the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior.The sequence of observations by
(y originally2,y5,y6,y7,y8) become (y2,y5,y6,y7)。
Step 207 calculates the second alarm statistics amount φt, execute step 208.
Detection device is according to the first alarm statistics amount λ and statistic sequence { λiObtain Current observation point abnormal amount δt, institute
State δtMeet:
When the λ is not in the first alarm threshold range H1When interior, the δtIt is more than the first alarm threshold range H1's
Absolute value, when the λ is in the first alarm threshold range H1When interior, the δtIt is 0.
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein, b is time gap function,For statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For
{λiMiddle on the occasion of sequence { λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For
{λiIn negative value sequence { λi -Standard deviation.
Step 208 calculates the second alarm statistics amount φtCorresponding second alarm threshold range H2, execute step 209.
Firstly, detection device obtains Outliers point parameter d, comprising:
Initialization exception observation point parameter d=0;If there is the first alarm statistics amount λ not in the first alarm threshold range H1It is interior
Observation point, add 1 to obtain new d current d;If current d is not equal to 0, and the first alarm statistics amount λ occurs in the first alarm
Threshold range H1Current d is subtracted 1 and obtains new d by interior observation point;If current d=0, and the first alarm statistics amount λ occur and exist
First alarm threshold range H1Interior observation point keeps d constant.
Specifically, before detecting observation, initialization exception observation point parameter d=0;During detecting observation, sentence
Whether disconnected d is equal to 0, if d is not equal to 0, two kinds of situations occurs in the value of new d: when there is the first alarm statistics amount λ not the
One alarm threshold range H1When interior observation point, 1 is added to obtain new d current d, when there is the first alarm statistics amount λ the
One alarm threshold range H1When interior observation point, current d is subtracted 1 and obtains new d;During detecting observation, if d is equal to
0, there are two kinds of situations in the value of new d: if there is the first alarm statistics amount λ not H within the scope of the first alarm threshold1Sight
Current d is added 1 to obtain new d by measuring point, and the Current observation point is deleted in First ray;If there are the first alarm statistics
λ is measured in the first alarm threshold range H1Interior observation point keeps d constant, i.e. d=0.
It is exemplary, it is assumed that the corresponding observation of the observation point for needing to detect is y1,y2,......,y7,y8, wherein y2,y5,
y6The first alarm statistics amount not within the scope of corresponding first alarm threshold, y1,y3,y4,y7,y8The first alarm statistics amount exist
Within the scope of corresponding first alarm threshold.
Detection device detection before, initialization exception observation point parameter d=0 starts to detect, using first observation point as
Current observation point detects the observation y of the observation point1, y1The first alarm statistics amount in corresponding first alarm threshold range
Interior, d is constant, d=0;Continue to detect the observation y of the observation point using second observation point as Current observation point2, y2First
For alarm statistics amount not within the scope of corresponding first alarm threshold, current d adds 1 to obtain new d, d=1;By third observation point
As Current observation point, the observation y of the observation point is detected3, current d=1, y3The first alarm statistics amount it is corresponding first accuse
In alert threshold range, current d subtracts 1 and obtains new d, d=0;Using the 4th observation point as Current observation point, the observation is detected
The observation y of point4, current d=0, y4The first alarm statistics amount within the scope of corresponding first alarm threshold, d is constant, d=0;It will
5th observation point detects the observation y of the observation point as Current observation point5, y5The first alarm statistics amount not corresponding
Within the scope of first alarm threshold, current d adds 1 to obtain new d, d=1;Using the 6th observation point as Current observation point, detection
The observation y of the observation point6, y6The first alarm statistics amount not within the scope of corresponding first alarm threshold, current d adds 1
To new d, d=2;Using the 7th observation point as Current observation point, the observation y of the observation point is detected7, current d=2, y7?
Within the scope of corresponding first alarm threshold, current d subtracts 1 and obtains new d, d=1 one alarm statistics amount;By the 8th observation point
As Current observation point, the observation y of the observation point is detected8, current d=1, y8The first alarm statistics amount it is corresponding first accuse
In alert threshold range, current d subtracts 1 and obtains new d, d=0.
Secondly, calculating the second alarm statistics amount φ according to threshold parameter K, Outliers point parameter dtCorresponding second accuses
Alert threshold range H2, the second alarm statistics amount threshold formula are as follows:
Wherein,For statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence { λi +}
Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence { λi -}
Standard deviation.
Step 209 judges the second alarm statistics amount φtWhether in the second alarm threshold range H2In range.
Second alarm statistics amount φtIn the second alarm threshold range H2It is interior, execute step 210;
Second alarm statistics amount φtNot in the second alarm threshold range H2It is interior, execute step 211.
Whether step 210, judgement currently alert.
If not alerting currently, the observation y of next observation point is detectedi+1, execute step 201;
If current alarm, step 213 is executed.
Whether step 211, judgement currently alert.
If current alarm, the observation y of next observation point is detectedi+1, execute step 201;
If not alerting currently, step 212 is executed.
Step 212, starting alarm, start timing, save the alarm time started to detection device, detect next current sight
The observation y of measuring pointi+1, execute step 201.
Step 213 closes alarm, terminates timing, saves the alarm end time to detection device, detects next current sight
The observation y of measuring pointi+1, execute step 201.
Detection device saves alarm time started and alarm shut-in time, makes the abnormality detection of the observation of observation point from list
Point alarm becomes continuous time period alarm, and the abnormality detection of the observation of observation point is more accurate, avoids scene of instantaneously leaping high
The single-point false-alarm of generation.
Fig. 3 indicate in the prior art the first alarm threshold range come judge Current observation point observation whether exception side
Method, Fig. 4 are expressed as the first alarm statistics amount judgment method provided in an embodiment of the present invention, and Fig. 5 is expressed as the embodiment of the present invention and mentions
The the second alarm statistics amount judgment method supplied.In Fig. 3 and Fig. 4, x-axis indicates the time, and y-axis indicates the first alarm statistics amount, Fig. 5
In, x-axis indicates the time, and y-axis indicates the second alarm statistics amount.
It is compared by Fig. 3 with the prior art it can be seen from Fig. 4 with method of the invention, the threshold parameter of the prior art is
Selected definite value, threshold parameter of the invention change with the fluctuating change of the detected value of First ray by rule of thumb, and first
Alarm threshold range can also change with the variation of threshold parameter, and big place is fluctuated in observation, and threshold parameter becomes larger, the
One alarm threshold range broadens, and small place is fluctuated in observation, and threshold parameter becomes smaller, and the first alarm threshold range narrows, this
The abnormality detection of sample, the observation of observation point is more accurate;As can be seen from Figure 4 and Figure 5, in the first alarm statistics of the invention
Judgment method situation is measured, when only a single point is abnormal, i.e., scene of instantaneously leaping high, the first alarm statistics amount has been more than the first announcement
Alert threshold range will determine that the observation of Current observation point is abnormal, but in practice, scene of instantaneously leaping high belongs to normal phenomenon,
This creates the terminal false-alarms;In the second alarm statistics amount judgment method situation of the invention, if the first alarm statistics amount is more than
First alarm threshold range, as long as the generation of subsequent not abnormal point, the second alarm statistics amount is decayed soon with exponential form
To 0, so as to avoid the generation of false-alarm, when continuous multiple test points are abnormal, the second alarm statistics amount is rapidly with index
Increase, will soon be more than the second alarm threshold range of linear increase.
The embodiment of the present invention provides a kind of performance monitoring apparatus 30, such as cloud computing infrastructure etc., as shown in fig. 6, packet
It includes:
Recording unit 301, for recording the observation of Current observation point into the sequence of observations;
First acquisition unit 302, the preceding N of the observation for obtaining the Current observation point in the sequence of observations
The First ray of a observation composition, the N are the integer more than or equal to 2;
First computing unit 303, for threshold parameter K to be calculated according to the First ray;
Second acquisition unit 304 is accused for obtaining the first alarm statistics amount λ corresponding first according to the threshold parameter K
Alert threshold range H1。
So, detection device record calculates threshold parameter K according to the First ray of Current observation value, to obtain
First alarm statistics amount λ and corresponding first alarm threshold range H1, thus whether extremely to detect Current observation point, and therefore, threshold
Value parameter K changes according to the variation of First ray, to make calculated first alarm threshold range H1It is more accurate, it improves
The accuracy of abnormality detection.
First computing unit 303 is specifically used for:
Obtain the standard deviation sigma of the First rayo;
The direct proportion function a, a for obtaining the period of waves of the First ray are greater than 1;
According to the standard deviation sigmaoThe threshold parameter K is calculated with the direct proportion function a, so that the threshold value is joined
Number K meets:
The second acquisition unit 304 is specifically used for:
Obtain statistic sequence { λi};
According to the statistic sequence { λiAnd threshold parameter K acquisition the first alarm threshold range H1, so that institute
State the first alarm threshold range H1Are as follows:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation.
The performance monitoring apparatus 30, as shown in Figure 7, further includes:
Third acquiring unit 305, for obtaining the first alarm statistics amount λ of the Current observation point;
First judging unit 306, for judging the first alarm statistics amount λ whether in the first alarm threshold range
H1It is interior;
First Alarm Unit 307 is used in the first alarm statistics amount λ not in the first alarm threshold range H1It is interior
When, issue alarm signal.
The performance monitoring apparatus 30, as shown in Figure 8, further includes:
4th acquiring unit 308, for obtaining the first alarm statistics amount λ of the Current observation point;
Second judgment unit 309, for judging the first alarm statistics amount λ whether in the first alarm threshold range
H1It is interior;
Unit 310 is deleted, if for the first alarm statistics amount λ not in the first alarm threshold range H1It is interior,
The observation of the Current observation point is deleted in the sequence of observations;
Second computing unit 311, for calculating the second alarm statistics amount φt;
Second computing unit 311 is specifically used for:
Obtain the first alarm statistics amount λ of the Current observation point;
Obtain statistic sequence { λi};
According to the first alarm statistics amount λ and the statistic sequence { λiObtain the abnormal amount of the Current observation point
δt, the δtMeet:
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation, b is default sampling time interval function.
Third computing unit 312, for calculating the second alarm statistics amount φtCorresponding second alarm threshold range H2;
The third computing unit 312 is specifically used for:
Obtain current Outliers point parameter d;
According to the threshold parameter K, all Outliers point parameter d, calculated by alarm statistics amount threshold formula
Second alarm statistics amount φtCorresponding second alarm threshold range H2, the alarm statistics amount threshold formula are as follows:
Third judging unit 313, for judging the second alarm statistics amount φtWhether in the second alarm threshold model
Enclose H2It is interior;
Second Alarm Unit 314, in the second alarm statistics amount φtNot in the second alarm threshold range H2
When interior, alarm signal is issued.
The performance monitoring apparatus 30, as shown in Figure 9, further includes:
Initialization unit 315 is used for initialization exception observation point parameter d=0;
Processing unit 316, is used for: the first alarm statistics amount λ is occurring not in the first alarm threshold range H1
When interior observation point, 1 is added to obtain new d current d;
It is not equal to 0 in current d, and the first alarm statistics amount λ occurs in the first alarm threshold range H1Interior
When observation point, current d is subtracted 1 and obtains new d.
In this way, the second alarm statistics amount, will soon rapidly with exponential increase when continuous multiple test points are abnormal
More than the second alarm threshold range of linear increase;When only a single point is abnormal, i.e., scene of instantaneously leaping high, as long as subsequent do not have
There is the generation of abnormal point, the second alarm statistics amount decays to 0 soon with exponential form, so as to avoid the generation of false-alarm.
The embodiment of the present invention provides a kind of performance monitoring apparatus 40, such as cloud computing infrastructure etc., as shown in Figure 10, packet
It includes:
Processor 401, is used for:
The observation of Current observation point is recorded into the sequence of observations;
The first sequence of the top n observation composition of the observation of the Current observation point is obtained in the sequence of observations
Column, the N are the integer more than or equal to 2;
Threshold parameter K is calculated according to the First ray;
The corresponding first alarm threshold range H of the first alarm statistics amount λ is obtained according to the threshold parameter K1。
So, detection device record calculates threshold parameter K according to the First ray of Current observation value, to obtain
First alarm statistics amount λ and corresponding first alarm threshold range H1, thus whether extremely to detect Current observation point, and therefore, threshold
Value parameter K changes according to the variation of First ray, to make calculated first alarm threshold range H1It is more accurate, it improves
The accuracy of abnormality detection.
The processor 401 is specifically used for:
Obtain the standard deviation sigma of the First rayo;
The direct proportion function a, a for obtaining the period of waves of the First ray are greater than 1;
According to the standard deviation sigmaoThe threshold parameter K is calculated with the direct proportion function a, so that the threshold value is joined
Number K meets:
The processor 401 is specifically used for:
Obtain statistic sequence { λi};
According to the statistic sequence { λiAnd threshold parameter K acquisition the first alarm threshold range H1, so that institute
State the first alarm threshold range H1Are as follows:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation.
The processor 401 is specifically used for:
Obtain the first alarm statistics amount λ of the Current observation point;
Judge the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, issue alarm signal.
The processor 401 is also used to:
Obtain the first alarm statistics amount λ of the Current observation point;
Judge the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, in the sequence of observations
Delete the observation of the Current observation point;
Calculate the second alarm statistics amount φt;
Calculate the second alarm statistics amount φtCorresponding second alarm threshold range H2;
Judge the second alarm statistics amount φtWhether in the second alarm threshold range H2It is interior;
As the second alarm statistics amount φtNot in the second alarm threshold range H2It is interior, issue alarm signal.
The processor 401 is specifically used for:
Obtain the first alarm statistics amount λ of the Current observation point;
Obtain statistic sequence { λi};
According to the first alarm statistics amount λ and the statistic sequence { λiObtain the abnormal amount of the Current observation point
δt, the δtMeet:
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence
{λi +Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence
{λi -Standard deviation, b is default sampling time interval function.
The processor 401 is specifically used for:
Initialization exception observation point parameter d=0;
If there is the first alarm statistics amount λ not in the first alarm threshold range H1Interior observation point, will be current
D add 1 to obtain new d;
If current d is not equal to 0, and the first alarm statistics amount λ occurs in the first alarm threshold range H1Interior
Current d is subtracted 1 and obtains new d by observation point.
Locating processor 401 is specifically used for:
Obtain current Outliers point parameter d;
According to the threshold parameter K, all Outliers point parameter d, calculated by alarm statistics amount threshold formula
Second alarm statistics amount φtCorresponding second alarm threshold range H2, the alarm statistics amount threshold formula are as follows:
In this way, the second alarm statistics amount, will soon rapidly with exponential increase when continuous multiple test points are abnormal
More than the second alarm threshold range of linear increase;When only a single point is abnormal, i.e., scene of instantaneously leaping high, as long as subsequent do not have
There is the generation of abnormal point, the second alarm statistics amount decays to 0 soon with exponential form, so as to avoid the generation of false-alarm.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can be stored in a computer readable storage medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed;And storage medium above-mentioned includes: ROM, RAM, magnetic disk or light
The various media that can store program code such as disk.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, protection scope of the present invention should be based on the protection scope of the described claims.
Claims (14)
1. a kind of method for monitoring performance characterized by comprising
The observation of Current observation point is recorded into the sequence of observations;
The First ray of the top n observation composition of the observation of the Current observation point is obtained in the sequence of observations,
The N is the integer more than or equal to 2;
Threshold parameter K is calculated according to the First ray;
Wherein, described threshold parameter K is calculated according to the First ray to include:
Obtain the standard deviation sigma of the First rayo;
The direct proportion function a, a for obtaining the period of waves of the First ray are greater than 1;
According to the standard deviation sigmaoThe threshold parameter is calculated with the direct proportion function a, so that the threshold parameter K is full
Foot:
The corresponding first alarm threshold range H of the first alarm statistics amount λ is obtained according to the threshold parameter K1。
2. the method according to claim 1, wherein
It is described that the corresponding first alarm threshold range H of first alarm statistics amount λ is obtained according to the threshold parameter K1Include:
Obtain statistic sequence { λi};
According to the statistic sequence { λiAnd threshold parameter K acquisition the first alarm threshold range H1, so that described
One alarm threshold range H1Are as follows:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence { λi +?
Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence { λi -?
Standard deviation.
3. according to the method described in claim 2, it is characterized in that, obtaining the first alarm according to the threshold parameter K described
The corresponding first alarm threshold range H of statistic λ1Later, the method also includes:
Obtain the first alarm statistics amount λ of the Current observation point;
Judge the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, issue alarm signal.
4. the method according to claim 1, wherein obtaining the first alarm according to the threshold parameter K described
The corresponding first alarm threshold range H of statistic λ1Later, the method also includes:
Obtain the first alarm statistics amount λ of the Current observation point;
Judge the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
If the first alarm statistics amount λ is not in the first alarm threshold range H1It is interior, institute is deleted in the sequence of observations
State the observation of Current observation point;
Calculate the second alarm statistics amount φt;
Calculate the second alarm statistics amount φtCorresponding second alarm threshold range H2;
Judge the second alarm statistics amount φtWhether in the second alarm threshold range H2It is interior;
As the second alarm statistics amount φtNot in the second alarm threshold range H2It is interior, issue alarm signal.
5. according to the method described in claim 4, it is characterized in that,
The second alarm statistics amount φ of the calculatingtInclude:
Obtain the first alarm statistics amount λ of the Current observation point;
Obtain statistic sequence { λi};
According to the first alarm statistics amount λ and the statistic sequence { λiObtain the abnormal amount δ of the Current observation pointt, institute
State δtMeet:
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence { λi +}
Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence { λi -}
Standard deviation, b is default sampling time interval function.
6. according to the method described in claim 5, it is characterized in that, calculating the second alarm statistics amount φ describedtIt is corresponding
Second alarm threshold range H2Before, the method also includes:
Initialization exception observation point parameter d=0;
If there is the first alarm statistics amount λ not in the first alarm threshold range H1Current d is added 1 by interior observation point
Obtain new d;
If current d is not equal to 0, and the first alarm statistics amount λ occurs in the first alarm threshold range H1Interior observation
Current d is subtracted 1 and obtains new d by point.
7. according to the method described in claim 6, it is characterized in that,
It is described to calculate the second alarm statistics amount φtCorresponding second alarm threshold range H2Include:
Obtain current Outliers point parameter d;
According to the threshold parameter K, Outliers point parameter d, the second alarm is calculated by alarm statistics amount threshold formula
Statistic φtCorresponding second alarm threshold range H2, the alarm statistics amount threshold formula are as follows:
8. a kind of performance monitoring apparatus characterized by comprising
Recording unit, for recording the observation of Current observation point into the sequence of observations;
First acquisition unit, the top n observation of the observation for obtaining the Current observation point in the sequence of observations
It is worth the First ray of composition, the N is the integer more than or equal to 2;
First computing unit, for threshold parameter K to be calculated according to the First ray;
Wherein, first computing unit is specifically used for:
Obtain the standard deviation sigma of the First rayo;
The direct proportion function a, a for obtaining the period of waves of the First ray are greater than 1;
According to the standard deviation sigmaoThe threshold parameter K is calculated with the direct proportion function a, so that the threshold parameter K is full
Foot:
Second acquisition unit, for obtaining the corresponding first alarm threshold model of the first alarm statistics amount λ according to the threshold parameter K
Enclose H1。
9. performance monitoring apparatus according to claim 8, which is characterized in that
The second acquisition unit is specifically used for:
Obtain statistic sequence { λi};
According to the statistic sequence { λiAnd threshold parameter K acquisition the first alarm threshold range H1, so that described
One alarm threshold range H1Are as follows:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence { λi +?
Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence { λi -?
Standard deviation.
10. performance monitoring apparatus according to claim 9, which is characterized in that the performance monitoring apparatus further include:
Third acquiring unit, for obtaining the first alarm statistics amount λ of the Current observation point;
First judging unit, for judging the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
First Alarm Unit is used in the first alarm statistics amount λ not in the first alarm threshold range H1When interior, issue
Alarm signal.
11. performance monitoring apparatus according to claim 8, which is characterized in that the performance monitoring apparatus further include:
4th acquiring unit, for obtaining the first alarm statistics amount λ of the Current observation point;
Second judgment unit, for judging the first alarm statistics amount λ whether in the first alarm threshold range H1It is interior;
Unit is deleted, if for the first alarm statistics amount λ not in the first alarm threshold range H1It is interior, in the observation
The observation of the Current observation point is deleted in value sequence;
Second computing unit, for calculating the second alarm statistics amount φt;
Third computing unit, for calculating the second alarm statistics amount φtCorresponding second alarm threshold range H2;
Third judging unit, for judging the second alarm statistics amount φtWhether in the second alarm threshold range H2It is interior;
Second Alarm Unit, in the second alarm statistics amount φtNot in the second alarm threshold range H2When interior, hair
Alarm signal out.
12. performance monitoring apparatus according to claim 11, which is characterized in that second computing unit is specifically used for:
Obtain the first alarm statistics amount λ of the Current observation point;
Obtain statistic sequence { λi};
According to the first alarm statistics amount λ and the statistic sequence { λiObtain the abnormal amount δ of the Current observation pointt, institute
State δtMeet:
Obtain the second alarm statistics amount φt, the φtMeet:
Wherein,For the statistic sequence { λiMiddle on the occasion of sequence { λi +Average value, σ+For { λiMiddle on the occasion of sequence { λi +}
Standard deviation,For the statistic sequence { λiIn negative value sequence { λi -Average value, σ-For { λiIn negative value sequence { λi -}
Standard deviation, b is default sampling time interval function.
13. performance monitoring apparatus according to claim 12, which is characterized in that the performance monitoring apparatus further include:
Initialization unit is used for initialization exception observation point parameter d=0;
Processing unit is used for: the first alarm statistics amount λ is occurring not in the first alarm threshold range H1Interior observation
When point, 1 is added to obtain new d current d;
It is not equal to 0 in current d, and the first alarm statistics amount λ occurs in the first alarm threshold range H1Interior observation point
When, current d is subtracted 1 and obtains new d.
14. performance monitoring apparatus according to claim 13, which is characterized in that
The third computing unit is specifically used for:
Obtain current Outliers point parameter d;
According to the threshold parameter K, Outliers point parameter d, the second alarm is calculated by alarm statistics amount threshold formula
Statistic φtCorresponding second alarm threshold range H2, the alarm statistics amount threshold formula are as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310535299.6A CN104598361B (en) | 2013-10-31 | 2013-10-31 | A kind of method for monitoring performance and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310535299.6A CN104598361B (en) | 2013-10-31 | 2013-10-31 | A kind of method for monitoring performance and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104598361A CN104598361A (en) | 2015-05-06 |
CN104598361B true CN104598361B (en) | 2018-12-14 |
Family
ID=53124174
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310535299.6A Active CN104598361B (en) | 2013-10-31 | 2013-10-31 | A kind of method for monitoring performance and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104598361B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106991145B (en) * | 2017-03-23 | 2021-03-23 | ***股份有限公司 | Data monitoring method and device |
CN113535486A (en) * | 2020-04-14 | 2021-10-22 | 北京沃东天骏信息技术有限公司 | Method and device for testing performance of machine |
CN113468021B (en) * | 2021-06-28 | 2024-02-09 | 阿波罗智联(北京)科技有限公司 | Method, device, equipment and storage medium for monitoring performance data |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101651568A (en) * | 2009-07-01 | 2010-02-17 | 青岛农业大学 | Method for predicting network flow and detecting abnormality |
CN102130798A (en) * | 2011-03-25 | 2011-07-20 | 中国电子科技集团公司第三十研究所 | Method and device for detecting multidimensional flow anomalies of distributed network |
-
2013
- 2013-10-31 CN CN201310535299.6A patent/CN104598361B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101651568A (en) * | 2009-07-01 | 2010-02-17 | 青岛农业大学 | Method for predicting network flow and detecting abnormality |
CN102130798A (en) * | 2011-03-25 | 2011-07-20 | 中国电子科技集团公司第三十研究所 | Method and device for detecting multidimensional flow anomalies of distributed network |
Non-Patent Citations (2)
Title |
---|
一种网络异常实时监测方法;邹柏贤;《计算机学报》;20030831;第26卷(第8期);第940-947页 * |
基于动态阈值的网络流量异常检测方法研究与实现;李中魁;《中国优秀硕士学位论文全文数据库信息科技辑》;20110315;第2011年卷(第3期);正文第16-27页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104598361A (en) | 2015-05-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104598361B (en) | A kind of method for monitoring performance and device | |
JP2015527586A (en) | Monitoring system and method for detecting changes in a mechanical system and adapting limits associated with the mechanical system to reflect the current conditions of the mechanical system | |
CN108759991B (en) | Measurement error diagnosis method and device for sensor in air conditioning system and air conditioning system | |
JP5425027B2 (en) | Random noise signal detection and filtering method | |
JP2000029513A (en) | Device and method for monitoring process data | |
CN103185015A (en) | Fan abnormity alarm system and method | |
JP5933386B2 (en) | Data management apparatus and program | |
SE1251203A1 (en) | Inspection system and method for correlating data from sensors and display means | |
CN105450454A (en) | Service monitoring and warning method and device | |
CN108226560A (en) | A kind of method and device for obtaining motor slow-speed of revolution angular speed | |
CN107637091A (en) | The method and apparatus synchronous for sensor | |
CN109471154B (en) | Small-size GM count pipe wide range monitoring instrument | |
JP2015152567A (en) | Chromatograph operation state monitoring device | |
KR102059112B1 (en) | IoT STREAM DATA QUALITY MEASUREMENT INDICATORS AND PROFILING METHOD FOR INTERNET OF THINGS AND SYSTEM THEREFORE | |
JP5373711B2 (en) | Radiation monitoring device | |
CN110962930A (en) | Steering wheel corner analysis method and device | |
CN102792167B (en) | Speed detection device | |
CN108957174B (en) | Voltage sag detection device and method | |
KR101969936B1 (en) | System for managing gas sensor and operating mehtod thereof | |
CN115866154A (en) | Time delay measuring method, device and system of vehicle-mounted multi-camera system and automobile | |
US8909968B2 (en) | Server and method of recording enviroment parameters of server | |
CN108999902A (en) | A kind of measurement method and device improving crane brake downslide accuracy of measurement | |
JPWO2017164368A1 (en) | Monitoring device, monitoring method, program | |
JP7266376B2 (en) | Waveform display device and waveform display program | |
CN108169544B (en) | Voltage sag display method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220224 Address after: 550025 Huawei cloud data center, jiaoxinggong Road, Qianzhong Avenue, Gui'an New District, Guiyang City, Guizhou Province Patentee after: Huawei Cloud Computing Technology Co.,Ltd. Address before: 518129 Bantian HUAWEI headquarters office building, Longgang District, Guangdong, Shenzhen Patentee before: HUAWEI TECHNOLOGIES Co.,Ltd. |
|
TR01 | Transfer of patent right |