CN101442762B - Method and apparatus for analyzing network performance and locating network fault - Google Patents
Method and apparatus for analyzing network performance and locating network fault Download PDFInfo
- Publication number
- CN101442762B CN101442762B CN2008102247770A CN200810224777A CN101442762B CN 101442762 B CN101442762 B CN 101442762B CN 2008102247770 A CN2008102247770 A CN 2008102247770A CN 200810224777 A CN200810224777 A CN 200810224777A CN 101442762 B CN101442762 B CN 101442762B
- Authority
- CN
- China
- Prior art keywords
- network
- attribute
- matrix model
- coordinate
- event
- 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
Images
Abstract
The invention discloses a network performance analysis method and device for automatic analysis for network performance correlation. The invention also discloses a network failure positioning method and device based on network performance analysis for reducing analysis quantity for positioning network failure, promoting accuracy of positioning result. The network performance analysis method comprises: setting time range, distance range and network attribute according to network performance analysis demand; determining correlated attribute of network attribute according to network fluctuation information in distance range received in time range; establishing specimen matrix model of every network affair according to association of every network affair with related network affair to which network attribute corresponds in time range and distance range; generating network attribute matrix model according to specimen matrix model of every network affair, the matrix model is used for describing network attribute in time range and distance range and probability of occurrence of related attribute.
Description
Technical field
The present invention relates to moving communicating field, relate in particular to a kind of network performance analysis and network failure location technology.
Background technology
The network quality of mobile communication system has become one of core competitiveness of operator.In order to consolidate the network quality of mobile communication system, promote user satisfaction, operator not only needs to solve the most basic network hardware fault or alarm in network optimization maintenance process, also need the fluctuation of every network performance index is monitored and analyzed, thereby in time find problem potential in the network, and the problem that these are potential is got rid of before influencing user's perception.
In the present network optimization maintenance process, available network performance index reaches thousands of, relation between every network performance index is intricate, the unusual fluctuations of every network performance index may cause owing to different reasons, and the ripple effect of network causes the appearance of a problem can cause the chain generation of peripheral a series of problems.Therefore for network optimization attendant, the situation when network performance analysis not only needs the problem that is conceived to take place also needs to analyze the variation that the front and back network performance takes place this problem; When locating the basic reason of this problem, can not be confined to the current problem point, also need comprehensively to judge in conjunction with network perimeter situation and relevant informations such as Internet resources, network parameter.
The peripheral network performance that causes with disconnected station, base station is the ripple effect of example explanation network unusually, as shown in Figure 1.Suppose that base station A resolves the station, the traffic of subscriber equipment will be absorbed by peripheral base station (base station B, C, D, H) in its overlay area, but because the subscriber equipment in the A overlay area, base station is far away apart from the distance of base station distance B, C, D, H, with problems such as the quality of handoff decline of initiation base station B, C, D, H, cutting off rate risings; (for example 1 hour) afterwards through after a while, suppose that the telephone traffic in the D overlay area, base station rises, simultaneously because base station D has absorbed the part traffic of base station A, the telephone traffic excess load that will cause base station D, it is relatively more difficult when base station D switches to cause its peripheral base station (base station C, E, F, G, H) thus, and it is congested to cause base station D that TCH takes place; The congested handover success rate that will cause its peripheral base station (base station C, E, F, G, H) of the TCH of base station D descends, and can not carry out causing the cutting off rate of base station C, E, F, G, H to rise under the optimum situation about switching; (for example 2 hours) because the cutting off rate of base station F raises, repeatedly but under the situation of aborted, may cause the complaint of user at the call drop problem at user's communication afterwards through after a while again.This shows, because the disconnected station of base station A, to cause the chain generation of a series of problems such as handover success rate decline, cutting off rate rising, telephone traffic excess load, the TCH of peripheral base station (base station B, C, D, E, F, G, H) be congested, and the problems referred to above are not to occur simultaneously.
Because the complexity and the uncertainty of network performance analysis, mainly adopting when carrying out network performance analysis at present manually is that master, webmaster supporting tool are the mode of assisting.Is example explanation network performance analysis method with the user in the F overlay area, above-mentioned base station at the complaint of call drop problem.Network optimization attendant must begin from the base station F in customer complaint place to carry out network performance analysis after receiving the complaint of user at the call drop problem, generally comprises following handling process:
The base station that the call drop problem is caused in step 1, location is the fault base station, and the fault base station is base station F herein;
The relevant network performance index of problem that step 2, retrieval and current base station occur;
The network performance index that step 3, analysis retrieve;
Step 4, judge whether to locate the basic reason of this call drop problem, if then execution in step 7, if not, then execution in step 5;
Step 5, judge whether to retrieve other the relevant network performance index of problem that occurs with current base station, if then execution in step 6, if not, then execution in step 7;
Other network performance index that step 6, analysis retrieve, and return execution in step 4;
Step 7, to judge whether problem that current base station occurs is adjacent the base station relevant, if then execution in step 8, if not, then execution in step 9;
The relevant adjacent base station of problem that step 8, screening and current base station occur, and return execution in step 2;
The basic reason that step 9, location call drop problem produce.
Based on above-mentioned handling process, the minimum amount of analysis that location call drop problem needs as shown in Figure 2.As can be seen, basic reason from the call drop problem of customer complaint to this call drop problem of location is the disconnected station of base station A, the relevant base station number that needs to analyze reaches 8 (if calculate according to the scale of each 20 adjacent base station in base station of existing network standard, the dependency basis standing-meeting reaches tens), the network performance index that needs to analyze reaches 15, if the network of relation performance index of the required reference of each network performance index of binding analysis, need to analyze the variation of tens network performance indexes before and after this call drop problem takes place altogether, and wherein surpass the possibility that 50% amount of analysis all is used to get rid of other reason that produces this call drop problem, make analysis efficiency very low.In addition, the analysis conclusion in each step and next step analysis content all by network optimization O﹠M personnel control, very likely increase more invalid amount of analysis, even get the wrong sow by the ear.
In the prior art, also can decision tree be set, judge default various possible causes by network management system automatically according to the criterion of setting then at problem common in the network.This scheme requires very high to the accuracy of the criterion set and decision tree and level of detail, be difficult to generally speaking arrive, and makes positional accuracy lower.
Mainly there is following problem in the prior art:
The basic reason of the problem that single network performance index analysis is existed in can't fixer network.For example, be that cutting off rate raises equally, may cause owing to many-sided reasons such as base station carrier fault, frequency interferences, switching problem, covering problems, only can't determine the basic reason that cutting off rate raises from network performance index of cutting off rate.
If every network performance index is carried out comprehensive analysis, owing to before analyzing, can't confirm the network performance index relevant with current problem, can only analyze investigation one by one, will cause a large amount of amount of analysis to be used to get rid of irrelevant network performance index, make analysis efficiency very low.For example, owing to the call drop problem, in order to determine the basic reason of this call drop problem, analytic process must travel through the network of relation performance index of 8 base stations of base station F periphery at above-mentioned.
The ripple effect of network and complicated network configuration make every basic reason to problem advance one deck, and the amount of analysis of being paid all increases by geometric progression, and therefore is difficult to the basic reason of accurate orientation problem.For example at above-mentioned call drop problem, suppose that each base station has only under the scene of 3～6 adjacent base stations, because the disconnected station of base station A has caused 8 base stations of periphery that the unusual fluctuations of 15 network performance indexes have taken place, for the basic reason of the call drop problem of locating base station F, nearly tens of the amount of analysis of network performance index.And in existing network, the adjacent base station of each base station is generally more than 20, and the disconnected station of a key position base station may cause every network performance index of tens even up to a hundred peripheral base stations to be affected.
The method of the existing basic reason that produces by the orientation problem of default decision tree and criterion, the experience that depends on network optimization O﹠M personnel make its accuracy and level of detail limited, are only applicable to simple application scenarios.Use decision tree and make that between the analysis conclusion in each step must be mutual exclusion, analyze conclusion B, cause when analyzing concurrent problem, can't using decision tree if promptly obtain analyzing just essential eliminating of conclusion A.For example, cutting off rate raises, may be owing to multiple reasons such as base station carrier fault, frequency interferences, switching problem, covering problem cause, if carrier frequency fault and covering problem have taken place simultaneously in certain base station, use decision tree and can only obtain wherein a kind of in carrier frequency fault and two kinds of reasons of covering problem.
During basic reason that existing fixer network produces, can't break away from network optimization O﹠M personnel's judgement and decision-making in the network performance analysis process, the analysis conclusion in each step and next step analysis content are all decided by network optimization O﹠M personnel.In addition, when the relevance of analyzing between the basic reason of multinomial network performance index and problem,, need the people for carrying out association owing to all there is ambiguity in the time and space relevance between the heterogeneous networks performance index.For example, at above-mentioned call drop problem, the time of customer complaint call drop problem is not the time that disconnected station takes place base station A, the time of telephone traffic excess load that neither base station D, therefore when these several network performance indexes are analyzed, have time ambiguity, need artificial property association in time of determining these several network performance indexes.Same this situation that exists on spatial correlation, for example, the generation area of customer complaint call drop problem is an approximate location, for the fault location base station, needs the people for carrying out space correlation.
Summary of the invention
The invention provides a kind of network performance analysis method and apparatus, in order to realize automatic analysis to the network performance relevance.
Accordingly, the invention provides a kind of network failure locating method and device of performance evaluation Network Based, the amount of analysis during in order to the minimizing locating network fault, the accuracy of lifting positioning result.
The invention provides a kind of network performance analysis method, comprising:
According to the network performance analysis demand time range, distance range and network attribute are set;
According to the network fluctuation information in the described distance range that receives in the described time range, determine the association attributes of described network attribute;
According to each network event of described network attribute correspondence in described time range and the distance range and the incidence relation of its network of relation incident, set up the sample matrix model of each network event, the network of relation incident of each network event comprises the network event of the association attributes correspondence of other network event of network attribute correspondence of current network incident and described network attribute;
Generate the matrix model of described network attribute according to the sample matrix model of each network event, the matrix model of described network attribute is used to describe the probability of happening of described network attribute in described time range and the distance range and association attributes thereof.
The invention provides a kind of network performance analysis device, comprising:
The unit is set, is used for time range, distance range and network attribute being set according to the network performance analysis demand;
Determining unit is used for determining the association attributes of described network attribute according to the network fluctuation information in the described distance range that receives in the described time range;
Analytic unit, be used for according to each network event of described network attribute correspondence in described time range and the distance range and the incidence relation of its network of relation incident, set up the sample matrix model of each network event, the network of relation incident of each network event comprises the network event of the association attributes correspondence of other network event of network attribute correspondence of current network incident and described network attribute;
Generation unit, be used for generating according to the sample matrix model of each network event the matrix model of described network attribute, the matrix model of described network attribute is used to describe the probability of happening of described time range and interior described network attribute of distance range and association attributes thereof.
The invention provides a kind of network failure locating method of performance evaluation Network Based, comprising:
According to the network attribute selection matrix model of network failure correspondence, the matrix model of described network attribute is used to describe the probability of happening of setting-up time scope and interior described network attribute of distance range and association attributes thereof;
Determine that according to described matrix model probability of happening is not less than setting threshold and the network of relation incident of generation before the time of origin point of described network failure;
Locate the basic reason of described network failure according to the network of relation incident of determining.
The invention provides a kind of network failure positioner of performance evaluation Network Based, comprising:
Memory cell is used to store the matrix model of each network attribute, and the matrix model of each network attribute is used to describe the probability of happening of setting-up time scope and interior described network attribute of distance range and association attributes thereof;
Selected cell is used for the network attribute selection matrix model according to the network failure correspondence;
Determining unit determines that according to described matrix model probability of happening is not less than setting threshold and the network of relation incident of generation before the time of origin point of described network failure;
Positioning unit is used for locating according to the network of relation incident determined the basic reason of described network failure.
Network performance analysis method and apparatus provided by the invention, according to each network event of network attribute correspondence and the incidence relation of its network of relation incident, set up the sample matrix model of each network event, generate the matrix model of network attribute according to the sample matrix model of each network event, matrix model is used to describe the probability of happening of this network attribute in setting-up time scope and the distance range and association attributes thereof.This programme has been realized the automatic analysis to the network performance relevance, has effectively promoted the analysis efficiency to network performance simultaneously.
The network failure locating method of performance evaluation Network Based provided by the invention and device, network attribute selection matrix model according to the network failure correspondence, determine that according to matrix model probability of happening is not less than setting threshold and the network of relation incident of generation before the time of origin point of network failure, thus the basic reason of locating network fault.This programme can be chosen the optimum analysis path automatically when locating network fault, thereby has reduced amount of analysis, has improved analysis efficiency, has effectively promoted the accuracy of positioning result simultaneously.
Description of drawings
Fig. 1 is the ripple effect schematic diagram of network in the prior art;
Fig. 2 is the minimum amount of analysis schematic diagram that location call drop problem needs in the prior art;
Fig. 3 is a nodal distance schematic diagram in the embodiment of the invention;
Fig. 4 is the expression mode schematic diagram of ripple effect in the sample matrix model of network in the embodiment of the invention;
Fig. 5 is a network performance analysis method flow diagram in the embodiment of the invention;
Fig. 6 is the sample matrix model schematic diagram of B high cutting off rate incident in base station in the embodiment of the invention;
Fig. 7 is the sample matrix model schematic diagram of F high cutting off rate incident in base station in the embodiment of the invention;
Fig. 8 is the matrix model schematic diagram that generates according to Fig. 6 and sample matrix model shown in Figure 7 in the embodiment of the invention;
Fig. 9 is a network performance analysis device block diagram in the embodiment of the invention;
Figure 10 is the network failure locating method flow chart of performance evaluation Network Based in the embodiment of the invention;
Figure 11 is the network failure positioner block diagram of performance evaluation Network Based in the embodiment of the invention.
Embodiment
At first define the several basic conceptions that relates in the embodiment of the invention.Network event is meant in particular point in time and locality, taken place in the network that change has taken place for specific problem, Internet resources, adjustment or the like has taken place network parameter; Network failure is meant influences user's perception and network quality, needs to solve as early as possible the network event of handling.
The embodiment of the invention at first provides the matrix notation mode of network performance relevance.The available data relevance has two kinds of expression modes, and a kind of mode is to utilize relevance function, calculates two groups of coefficient R between the data, and relevance function can only be represented the relevance between two metavariables, can't represent for the relevance between the polytomy variable.Another kind of mode is to utilize multiple regression analysis, by one group of polynary N rank equation, relevance between the expression polytomy variable, if be used for the phase-split network performance, because that network performance index reaches is thousands of, then regression analysis need be set up the equation group with network performance index corresponding element number, if require the accuracy of analysis, then the order of equation number will reach more than 10 rank at least, and the operand during each regression analysis is very big, has limited the application of multiple regression analysis in network performance analysis.
In the embodiment of the invention in conjunction with the demand of mobile communication network structure and network performance analysis, for each network event is set up the sample matrix model, this sample matrix model has three dimensions, being respectively time dimension (T), Spatial Dimension (S) and attribute dimensions (A), is reference axis with above-mentioned three dimensions, can set up three dimensional space coordinate system, the coordinate of matrix dot adopts R (T in the coordinate system, S, A) expression, below to the explanation that makes an explanation of the implication of each dimension.
Time dimension (T), expression be apart from the relative time of this network event time of origin point, span be from-∞ to the integer the+∞, when T=0, expression and this network event take place simultaneously; When T=1, be illustrated in 1 chronomere after the time of origin point of this network event; When T=2, be illustrated in 2 chronomeres after the time of origin point of this network event; If T equals negative value, for example-during X, be illustrated in X chronomere before the time of origin point of this network event.Wherein, chronomere can be hour, minute, second, also can be day, week, month, can set flexibly.
Spatial Dimension (S), expression be apart from the relative distance of this network event scene, and span is the integer from 0 to+∞.When S=0, expression takes place in the overlay area of same place or same network element with this network event; During S=1, expression is apart from 1 parasang of scene of this network event; S=2 represents 2 parasangs of scene apart from this network event.Parasang can be rice, km, also can be the nodal distance of network element topology.After the nodal distance of network element topology is meant network element topology structure is showed with similar tree, the shortest topological tree number of nodes at institute interval between two network elements.As shown in Figure 3, the nodal distance of network element CellA and network element NbRCellA is 1, and the nodal distance of network element CellA and network element NbRCellC is 2.
Attribute dimensions (A), the representation attribute title is meant the title set of every network performance relevant with this network event, that be correlated with network performance variation, fluctuation or network abnormal conditions, network element hardware fault, network event, network parameter.For example the numerical value of A can be the Property Names relevant with network performance analysis such as disconnected station, high cutting off rate, carrier frequency fault, a certain parameter variation.
The coordinate of each matrix dot is R (T in the sample matrix model, S, A), expression is apart from the time of origin point T chronomere of this network event, changes or unusual network of relation incident apart from the A attribute of certain network element of the scene S parasang of this network event.
With the disconnected station of base station A in the background technology is example, and the matrix notation mode of network performance relevance is described, promptly how to set up the sample matrix model at the disconnected station of base station A (network event).
Base station A resolves the station, and the telephone traffic excess load of base station D after a hour causes the cutting off rate of base station F to raise simultaneously, and the ripple effect of network is expression in the following way in the coordinate system of sample matrix model, as shown in Figure 4, and wherein:
X=R (T, S, A)=and R (0,0, disconnected station)=1, expression current network incident is the disconnected station of base station A, value is that 1 this network event of expression takes place;
Y=R (T, S, A)=R (1,1, the telephone traffic excess load)=1, expression is apart from the disconnected station of a base station A chronomere (1 hour), with in the base station A distance network element group of 1 nodal distance (because base station C is the adjacent base station of base station A, so nodal distance is 1) the telephone traffic excess load, value is that 1 this network event of expression takes place.
Z=R (T, S, A)=R (1,2, high cutting off rate)=1, expression is apart from the disconnected station of a base station A chronomere (a hour), with in the base station A distance network element group of 2 nodal distances (because base station F is the adjacent base station of base station C, so nodal distance is 2) high cutting off rate, value is that 1 this network event of expression takes place.
The sample matrix model representation incidence relation of a certain network event and its network of relation incident.Only once or the sample matrix model of a spot of network event network performance relevance randomness of being showed stronger, need generate the matrix model of this network attribute according to the magnanimity sample matrix model of consolidated network attribute correspondence, matrix model has been represented the common trait of the network performance relevance of same class network event, has generality.The matrix model of network attribute is consistent with the sample matrix model of each network event, has three dimensions equally, is respectively time dimension (T), Spatial Dimension (S) and attribute dimensions (A), the coordinate of each matrix dot adopts R (T equally, S, A) expression, wherein:
T represents T chronomere of range coordinate initial point, and S represents S parasang of range coordinate initial point, A representation attribute title, and the implication of each dimension is consistent with the sample matrix model, repeats no more; The coordinate figure R of each matrix dot represents probability of happening, and span is [0,1].
In the embodiment of the invention, adopt matrix-style to represent the network performance relevance, can represent the relevance of polynary network performance fluctuation, and only need suitable span of extending each dimension to get final product, have unlimited extensibility for the relevance of complexity; Information with height concentrates property, only represents to get final product with three dimensions, greatly reduces the amount of analysis of relevance; Can reduce the redundant information that the ripple effect of network causes.From the sample matrix model of each network event as can be seen, the telephone traffic excess load of base station D will cause the high cutting off rate of peripheral base station C, E, F, G, H, originally relate to 5 network of relation incidents, and employing sample matrix model, the high cutting off rate of five base stations is expressed as the high cutting off rate of adjacent base station (or be called nodal distance be 1 network element group), and only represent to get final product with 1 matrix dot, it is original 20% that amount of information is compressed to, and further reduced the amount of analysis of network performance relevance.Matrix model by network attribute can be automatically, the accurate basic reason of locating network fault, can determine the possible cause of network failure less than 0 matrix dot according to T value, and (coordinate figure A) determines its accuracy for T, S by R; Because dimension T and S are discrete variable, and its span is the relative scope of a dynamic change, therefore the time ambiguity that in the past when network performance analysis, existed and the problem of space ambiguity have been solved, can finish network performance analysis automatically, realize accurate, the automatic location of network failure.
The invention provides a kind of network performance analysis method, as shown in Figure 5, comprise following handling process:
S501, time range, distance range and network attribute are set according to the network performance analysis demand;
According to the network performance analysis demand, can determine time range, distance range and the network attribute of network performance analysis, for example: it is within nearest 30 days that time range is set, and distance range is a network element all in the existing network, and network attribute is high cutting off rate.
S502, according to the network fluctuation information in this distance range that receives in this time range, determine the association attributes of network attribute;
What network performance analysis was paid close attention to is the variation of network performance, unusual and fault, and the related influence that network is produced when these phenomenons take place.In order to improve analysis efficiency, need carry out Information Compression and preliminary treatment to thousands of kinds of primitive network performance datas, data that wherein do not change or N/R are peeled off, only keep that those change, information obvious fluctuation and unusual arranged, for example network monitoring warning information, Internet resources transition information, network traffic information, network parameter adjustment record, customer complaint information or the like are referred to as network fluctuation information with above-mentioned information in the embodiment of the invention.For example, the disconnected station of above-mentioned base station A causes in the whole process of ripple effect of network, have with high cutting off rate that potential related network of relation incident comprises that base station B, C, D, E, F, G, G switch that the TCH of unusual, base station D telephone traffic excess load, base station D is congested, the customer complaint of base station F and the disconnected station of base station A etc., then the association attributes of high cutting off rate comprise switch that unusual, telephone traffic excess load, TCH are congested, disconnected station etc.
S503, according to each network event of network attribute correspondence in this time range and the distance range and the incidence relation of its network of relation incident, set up the sample matrix model of each network event, the network of relation incident comprises other network event of network attribute correspondence and the network event of association attributes correspondence;
Each network of relation incident front and back, the network performance variation of periphery, abnormal conditions are set up independently sample matrix model according to incidence relation.
S504, generate the matrix model of network attribute according to the sample matrix model of each network event, wherein, matrix model is used to describe the probability of happening of network attribute in time range and the distance range and association attributes thereof.
The sample matrix model of each network event (similar incident) of this network attribute correspondence is carried out iteration average, because each network presentation when independently on behalf of this network event, the sample matrix model of network event only take place.But with the sample matrix model of all similar incidents carry out iteration average calculate after, the matrix model of the final network attribute that generates will have the statistical probability meaning, can be used for the basic reason location of network event and the ripple effect of network event and predict.Iteration is averaged can directly independently T, S, the A dimension of sample matrix model be carried out computing to all, (T, S, coordinate figure A) get all flat value will to be in the R of same coordinate, this mean value is the coordinate figure of respective coordinates in the matrix model, promptly a certain network change or unusual probability of happening.
Describe with the example in the background technology,, finally need set up 7 independently sample matrix models because there are high cutting off rate problem in base station B, C, D, E, F, G, G.The sample matrix model of the high cutting off rate incident of base station B as shown in Figure 6, wherein:
B1=R (0,0, high cutting off rate)=1, high cutting off rate incident has taken place in expression base station B;
B2=R (0,0, switch unusual)=1 was illustrated in the same time that high cutting off rate incident takes place base station B, and the switching anomalous event has also taken place base station B;
B3=R (0,1, high cutting off rate)=1 was illustrated in the same time that high cutting off rate incident takes place base station B, in the adjacent base station of base station B (being base station C) high cutting off rate incident had taken place also;
B4=R (0,1, switch unusual)=1 was illustrated in the same time that high cutting off rate takes place base station B, in the adjacent base station of base station B (being base station C) the switching anomalous event had taken place also;
B5=R (1,2, switch unusual)=1, be illustrated in the chronomere that high cutting off rate takes place base station B after, the switching anomalous event has taken place in the adjacent base station (base station D, E) of the adjacent base station of base station B (base station C);
B6=R (0,1, disconnected station)=1 was illustrated in the same time that high cutting off rate incident takes place base station B, resolved the station incident in the adjacent base station of base station B (being base station A).
The sample matrix model of the high cutting off rate incident of base station F, as shown in Figure 7, wherein:
F1=R (0,0, high cutting off rate)=1, high cutting off rate incident has taken place in expression base station F;
F2=R (0,0, switch unusual)=1 was illustrated in the same time that high cutting off rate incident takes place base station F, and the switching anomalous event has also taken place base station F;
F3=R (0,1, high cutting off rate)=1 was illustrated in the same time that high cutting off rate incident takes place base station F, in the adjacent base station of base station F (being base station E, G) high cutting off rate incident had taken place also;
F4=R (0,1, switch unusual)=1 was illustrated in the same time that high cutting off rate incident takes place base station F, in the adjacent base station of base station F (being base station E, G) the switching anomalous event had taken place also;
F5=R (1,2, switch unusual)=1 is illustrated in base station F and takes place in the adjacent base station (base station C, H) of the adjacent base station of base station F (base station D) the switching anomalous event to have taken place in the high cutting off rate incident chronomere before;
F6=R (1,2, disconnected station)=1 is illustrated in base station F and takes place to resolve the station incident in the adjacent base station (base station A) of the adjacent base station of base station F (base station D) in the high cutting off rate incident chronomere before.
The sample matrix model class of the high cutting off rate incident in all the other base stations is seemingly given unnecessary details no longer one by one.
As shown in Figure 8, be the matrix model of the high cutting off rate that obtains after averaging according to the sample matrix model iteration of base station B and base station F, wherein:
F1/B1=R (0,0, high cutting off rate)=1 represents that this matrix model is used for analyzing high cutting off rate or we can say that at the high cutting off rate of this matrix model be an inevitable situation;
F2/B2=R (0,0, switch unusual)=1, when being illustrated in high cutting off rate and taking place, same network element has 100% the concurrent switching of probability unusual;
High cutting off rate when being illustrated in high cutting off rate problem and taking place, also will take place in the adjacent base station in F3/B3=R (0,1, high cutting off rate)=1;
F4/B4=R (0,1, switch unusual)=1, when being illustrated in high cutting off rate and taking place, adjacent base station has 100% the concurrent switching of probability unusual;
F5/B5=R (± 1,2, switch unusual)=0.5 is illustrated in chronomere in front and back (1 hour) that high cutting off rate takes place, and has 50% probability to switch unusually in the adjacent base station of adjacent base station;
F6=R (1,2, disconnected station)=0.5, being illustrated in has 50% probability that disconnected station takes place in the adjacent base station that the preceding adjacent base station of a chronomere (1 hour) takes place high cutting off rate;
B6=R (0,1, disconnected station)=0.5 when being illustrated in high cutting off rate and taking place, has 50% probability that disconnected station takes place in the adjacent base station.
Though in the top simple example, only the matrix model of the high cutting off rate that obtains of the sample matrix model by two high cutting off rate incidents has tangible one-sidedness, but if the sample matrix model of a large amount of high cutting off rate incidents is carried out after iteration averages, the probability of some chance phenomena will be substantially equal to zero, and really the association attributes of this network attribute will be retained, and obtain the matrix model of this network attribute thus.
Based on same technical conceive, the embodiment of the invention provides a kind of network performance analysis device, as shown in Figure 9, comprising:
Determining unit 902 is used for determining the association attributes of described network attribute according to the network fluctuation information in the described distance range that receives in the described time range;
Network performance analysis method based on matrix model comprises following advantage:
The matrix model of the attribute of the historical data of phase-split network performance, and automatic generating network automatically is used for the subsequent network failure location, accelerates the analysis efficiency of network failure; Analytical calculation efficient height, filtering change with network performance and unusual irrelevant information, the relevance between the phase-split network attribute emphatically; Can constantly improve the analysis precision of matrix model by learning the new network event of this network attribute correspondence automatically; The analytic process of whole matrix model need not can realize automatical analysis by artificial auxiliary, has broken away from the requirement to attendant's experience and ability; The matrix model that generates can be directly used in the basic reason location (being the matrix dot of T in the searching matrix model＜=0) of the network event of this network attribute correspondence, also can occur the issuable ripple effect in back at the network event of this network attribute correspondence and predict (matrix dot of promptly searching for T＞0); Need not the all-network performance is investigated one by one when utilizing the basic reason of matrix model phase-split network incident, the network performance state of the matrix dot that only needs to check that the R value is bigger in the matrix model gets final product, and can choose the optimum analysis path automatically.
On the matrix model basis that generates each network attribute, the embodiment of the invention provides a kind of network failure locating method of performance evaluation Network Based simultaneously, as shown in figure 10, comprising:
S1001, according to the network attribute selection matrix model of network failure correspondence, wherein, the matrix model of network attribute is used to describe the probability of happening of network attribute in setting-up time scope and the distance range and association attributes thereof.
S1002, the network of relation incident of determining that probability of happening is not less than setting threshold and before the time of origin point of this network failure, taking place according to matrix model;
This step specifically comprises: from described matrix model select time dimension coordinate T less than 0 and coordinate figure be not less than the matrix dot of setting threshold; According to Spatial Dimension coordinate S, time dimension coordinate T and the attribute dimensions coordinate A of the matrix dot of selecting, and the generation case point of described network failure and scene are determined the network of relation incident.
The network of relation incident that S1003, basis are determined is located the basic reason of this network failure.
Preferable, this method also comprises:
Be not less than setting threshold and the network event of deriving of generation after the time of origin point of described network failure according to described matrix model predicted occurrence probability.
Based on same technical conceive, the embodiment of the invention provides a kind of network failure positioner of performance evaluation Network Based, as shown in figure 11, comprising:
Determining unit 1103 determines that according to described matrix model probability of happening is not less than setting threshold and the network of relation incident of generation before the time of origin point of described network failure;
Preferable, this device also comprises predicting unit 1105, is used for the network event of deriving that is not less than setting threshold and takes place after the time of origin point of described network failure according to described matrix model predicted occurrence probability.
The network failure locating method that the embodiment of the invention provides utilizes the matrix model of network attribute to realize that the quantification of relevance between network performance time, space and the many indexs shows, by the basic reason of the relevance orientation problem between a plurality of indexs.Utilize matrix model can in network performance analysis, obtain the optimized analysis path automatically, with the amount of analysis acquisition accurate analytical results of minimum; Can discern the influence with the phase-split network ripple effect automatically, ignore and the irrelevant problem of deriving of problem reason; Need not artificially to set Rule of judgment or decision tree, and can discern the basic reason of the concurrent complex network fault of many problems; Realize the time ambiguity of disparate networks performance, the automatic matching problem of space ambiguity by the discrete feature of matrix model, whole network performance analysis process need not artificial participation can obtain final result.
Obviously, those skilled in the art can carry out various changes and modification to the present invention and not break away from the spirit and scope of the present invention.Like this, if of the present invention these are revised and modification belongs within the scope of claim of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these changes and modification interior.
Claims (11)
1. a network performance analysis method is characterized in that, comprising:
According to the network performance analysis demand time range, distance range and network attribute are set;
According to the network fluctuation information in the described distance range that receives in the described time range, determine the association attributes of described network attribute;
According to each network event of described network attribute correspondence in described time range and the distance range and the incidence relation of its network of relation incident, set up the sample matrix model of each network event, the network of relation incident of each network event comprises the network event of the association attributes correspondence of other network event of network attribute correspondence of current network incident and described network attribute;
Generate the matrix model of described network attribute according to the sample matrix model of each network event, the matrix model of described network attribute is used to describe the probability of happening of described network attribute in described time range and the distance range and association attributes thereof.
2. the method for claim 1 is characterized in that, the sample matrix model of described each network event has time dimension, Spatial Dimension and attribute dimensions, the coordinate employing R of each matrix dot (T, S, A) expression, wherein:
T represents a time of origin point T chronomere of the described network event of distance; S represents a scene S parasang of the described network event of distance; A representation attribute title; The coordinate figure R of the matrix dot of described network of relation incident is 1.
3. method as claimed in claim 2 is characterized in that the matrix model of described network attribute has time dimension, Spatial Dimension and attribute dimensions, the coordinate employing R of each matrix dot (T, S, A) expression, wherein:
T represents T chronomere of range coordinate initial point; S represents S parasang of range coordinate initial point; A representation attribute title; The coordinate figure R of each matrix dot represents probability of happening.
4. method as claimed in claim 3 is characterized in that, describedly generates the matrix model of described network attribute according to the sample matrix model of each network event, specifically comprises:
Coordinate figure to the matrix dot that coordinate is identical in all sample matrix models is averaged; And
With the coordinate figure of each coordinate corresponding average as coordinate described in the matrix model of described network attribute.
5. a network performance analysis device is characterized in that, comprising:
The unit is set, is used for time range, distance range and network attribute being set according to the network performance analysis demand;
Determining unit is used for determining the association attributes of described network attribute according to the network fluctuation information in the described distance range that receives in the described time range;
Analytic unit, be used for according to each network event of described network attribute correspondence in described time range and the distance range and the incidence relation of its network of relation incident, set up the sample matrix model of each network event, the network of relation incident of each network event comprises the network event of the association attributes correspondence of other network event of network attribute correspondence of current network incident and described network attribute;
Generation unit, be used for generating according to the sample matrix model of each network event the matrix model of described network attribute, the matrix model of described network attribute is used to describe the probability of happening of described time range and interior described network attribute of distance range and association attributes thereof.
6. the network failure locating method of a performance evaluation Network Based is characterized in that, comprising:
According to the network attribute selection matrix model of network failure correspondence, the matrix model of described network attribute is used to describe the probability of happening of setting-up time scope and interior described network attribute of distance range and association attributes thereof;
Determine that according to described matrix model probability of happening is not less than setting threshold and the network of relation incident of generation before the time of origin point of described network failure;
Locate the basic reason of described network failure according to the network of relation incident of determining.
7. method as claimed in claim 6 is characterized in that, also comprises:
Be not less than setting threshold and the network event of deriving of generation after the time of origin point of described network failure according to described matrix model predicted occurrence probability.
8. as claim 6 or 7 described methods, it is characterized in that the matrix model of described network attribute has time dimension, Spatial Dimension and attribute dimensions, the coordinate employing R of each matrix dot (T, S, A) expression, wherein:
T represents T chronomere of range coordinate initial point; S represents S parasang of range coordinate initial point; A representation attribute title; The coordinate figure R of each matrix dot represents probability of happening.
9. method as claimed in claim 8 is characterized in that, the described network of relation incident of determining that probability of happening is not less than setting threshold and taking place before the time of origin point of described network failure according to matrix model specifically comprises:
From described matrix model select time dimension coordinate T less than 0 and coordinate figure be not less than the matrix dot of setting threshold;
According to Spatial Dimension coordinate S, time dimension coordinate T and the attribute dimensions coordinate A of the matrix dot of selecting, and the time of origin point of described network failure and scene are determined the network of relation incident.
10. the network failure positioner of a performance evaluation Network Based is characterized in that, comprising:
Memory cell is used to store the matrix model of each network attribute, and the matrix model of each network attribute is used to describe the probability of happening of setting-up time scope and interior described network attribute of distance range and association attributes thereof;
Selected cell is used for the network attribute selection matrix model according to the network failure correspondence;
Determining unit determines that according to described matrix model probability of happening is not less than setting threshold and the network of relation incident of generation before the time of origin point of described network failure;
Positioning unit is used for locating according to the network of relation incident determined the basic reason of described network failure.
11. device as claimed in claim 10 is characterized in that, also comprises:
Predicting unit is used for the network event of deriving that is not less than setting threshold and takes place according to described matrix model predicted occurrence probability after the time of origin point of described network failure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102247770A CN101442762B (en) | 2008-12-29 | 2008-12-29 | Method and apparatus for analyzing network performance and locating network fault |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102247770A CN101442762B (en) | 2008-12-29 | 2008-12-29 | Method and apparatus for analyzing network performance and locating network fault |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101442762A CN101442762A (en) | 2009-05-27 |
| CN101442762B true CN101442762B (en) | 2010-09-22 |
Family
ID=40726970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008102247770A Active CN101442762B (en) | 2008-12-29 | 2008-12-29 | Method and apparatus for analyzing network performance and locating network fault |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101442762B (en) |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102075979B (en) * | 2009-11-23 | 2014-07-02 | 中国移动通信集团北京有限公司 | Method and device for positioning in-band interference |
| CN102083114B (en) * | 2009-11-27 | 2014-03-12 | 中国移动通信集团北京有限公司 | Method and device for checking broken station service |
| CN102572888B (en) * | 2010-12-24 | 2014-10-08 | 中国移动通信集团北京有限公司 | Method and device for determining road dropped call event |
| CN102271348B (en) * | 2011-07-08 | 2014-03-05 | 电子科技大学 | Link quality estimation system and method for cyber physical system |
| CN104734871A (en) * | 2013-12-20 | 2015-06-24 | 中兴通讯股份有限公司 | Method and device for positioning failures |
| CN103973496B (en) * | 2014-05-21 | 2017-10-17 | 华为技术有限公司 | Method for diagnosing faults and device |
| CN105307201A (en) * | 2014-06-30 | 2016-02-03 | 中国移动通信集团浙江有限公司 | Method and device for analyzing quality of CSFB (CS Fall Back) service |
| CN106685674B (en) * | 2015-11-05 | 2020-01-10 | 华为技术有限公司 | Method and device for predicting network event and establishing network event prediction model |
| CN105554786B (en) * | 2015-12-11 | 2019-08-09 | 中国联合网络通信集团有限公司 | A kind of method and device of addressing network problems |
| CN107135117B (en) * | 2016-02-29 | 2020-04-21 | 中国移动通信集团福建有限公司 | Method and device for determining network weak coverage |
| CN108156012B (en) * | 2016-12-06 | 2021-09-10 | 中国移动通信集团设计院有限公司 | Network fault reporting data multi-dimensional classification statistical analysis method and device |
| CN109559583B (en) * | 2017-09-27 | 2022-04-05 | 华为技术有限公司 | Fault simulation method and device |
| CN109995561B (en) * | 2017-12-30 | 2022-03-29 | 中国移动通信集团福建有限公司 | Method, device, equipment and medium for positioning communication network fault |
| CN108683991A (en) * | 2018-08-23 | 2018-10-19 | 珠海市联电科技有限公司 | A kind of mobile network quality diagnostic method and system dividing monitoring system based on the rooms RFID |
| CN110875831B (en) * | 2018-08-31 | 2023-04-07 | 北京京东尚科信息技术有限公司 | Method and device for monitoring network quality |
| CN109814933B (en) * | 2019-01-29 | 2021-08-24 | 腾讯科技（深圳）有限公司 | Service data processing method and device |
| CN110808864A (en) * | 2019-11-12 | 2020-02-18 | 国家电网有限公司 | Communication early warning method, device and system |
-
2008
- 2008-12-29 CN CN2008102247770A patent/CN101442762B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN101442762A (en) | 2009-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101442762B (en) | Method and apparatus for analyzing network performance and locating network fault | |
| CN107466103B (en) | Terminal positioning method and network equipment | |
| CN109151168B (en) | Switching method and device of riding codes, mobile terminal and readable storage medium | |
| CN110662245B (en) | Base station load early warning method and device based on deep learning | |
| CN111352759B (en) | Alarm root cause judging method and device | |
| CN111132190A (en) | Base station load early warning method and device | |
| CN103150696A (en) | Method and device for selecting potential customer of target value-added service | |
| CN111143102A (en) | Abnormal data detection method and device, storage medium and electronic equipment | |
| CN111294819B (en) | Network optimization method and device | |
| CN110348622A (en) | A kind of Time Series Forecasting Methods based on machine learning, system and electronic equipment | |
| CN112200375B (en) | Prediction model generation method, prediction model generation device, and computer-readable medium | |
| CN104113869A (en) | Signaling data-based prediction method and system for potential complaint user | |
| CN116567547A (en) | Population data quality inspection method, system and readable storage medium | |
| CN113448808B (en) | Method, system and storage medium for predicting single task time in batch processing task | |
| CN111885618A (en) | Network performance optimization method and device | |
| CN111798066A (en) | Multi-dimensional prediction method and system for cell flow under urban scale | |
| CN116993227A (en) | Heat supply analysis and evaluation method, system and storage medium based on artificial intelligence | |
| CN106817710B (en) | Method and device for positioning network problem | |
| CN112232662A (en) | Service monitoring system and method | |
| CN111626497A (en) | People flow prediction method, device, equipment and storage medium | |
| CN104219622A (en) | People number measuring method based on LAC and crowd situation monitoring method and system | |
| CN107580329B (en) | Network analysis optimization method and device | |
| CN112836843A (en) | Method and device for predicting base station out-of-service alarm | |
| CN110443451A (en) | Event grading approach, device, computer equipment and storage medium | |
| EP4216073A1 (en) | Data management method, data management apparatus, and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |