CN104572455B - One kind is based on markovian component-based software reliability estimation method - Google Patents

Abstract

It is of the invention a kind of comprehensive and accurate, based on markovian component-based software reliability estimation method, comprise the following steps, (1) obtains application state transition probability matrix, each component and include function body quantity and each component extent of injury by analysis software structure；(2) software is tested using the use-case quantity no less than minimum test case, number is repaired in be performed number of times, failure number and the failure for counting each component, is obtained each component and is performed probability；(3) each component occupancy is obtained according to transition probability between the component obtained in step (1) and according to each component execution probability calculation obtained in step (2)；(4) each component failure rate is calculated；(5) each component complexity is calculated comprising function body quantity according to each component obtained in step (1)；(6) each component importance degree is calculated according to each component extent of injury obtained in step (1)；(7) each Member Reliability Analysis are assessed；(8) software reliability is assessed.

Description

One kind is based on markovian component-based software reliability estimation method

Technical field

The invention belongs to field of software engineering, it is related to software evaluation method, it is specially a kind of based on markovian Component-based software reliability estimation method.

Background technology

Large-scale real time software system is constantly updated with function constantly expanding with software size, and software configuration is also increasingly Complexity, requirement to its reliability also more and more higher, therefore need to set up software reliability assessment model, real-time software is carried out can By property evaluation.Current large-scale real time software system is made up of many software components, and each component frequency of use and pass Stroke degree is all different, it is therefore desirable to go to assess software reliability from the angle of software inhouse framework and software component.Typically Component software appraisal procedure has three kinds：Method based on section, the method based on state and the method based on path.Based on section Method mainly by analyzing different software input data sets, calculate the transfer between the probability and each component that different sections occur Probability assesses software reliability；The path that method based on path is experienced mainly for test case, counts on the path The frequency of use and probability of each component, computational software system reliability；Method based on state assumes the operation of software systems Journey is the state migration procedure between each component, and software systems reliability is assessed using theory of random processes.Wherein main flow is random Process model has two kinds：Markov Process Model (Markov Process Model) and nonhomogeneous Poisson process model (Nonhomogeneous Poission Model).Markov Process Model with J-M models as representative, by Jelinski and Moranda proposed that its main target is software estimation failure moment, residual error number and reliability in 1972.

The reliability estimation method of existing software systems mainly considers the single factor of component occupancy, it is believed that occupancy is high The crash rate of component can produce material impact to the crash rate of software systems.But the complexity of component and importance degree are to software The reliability of system is also critically important, and element structure is more complicated, and software systems reliability is lower；Although certain class important component utilization rate It is not high, but its failure will produce devastating impact to software systems.It is existing that objective comprehensive evaluation cannot be all carried out to it, Actual demand can not be met.

The content of the invention

For problems of the prior art, the present invention provides one kind can consider software occupancy, complexity And importance degree, carry out thoroughly evaluating based on markovian component-based software reliability estimation method.

The present invention is to be achieved through the following technical solutions：

One kind of the invention is based on markovian component-based software reliability estimation method, comprises the following steps,

(1) by analysis software structure, application state transition probability matrix, each component are obtained and includes function body quantity and each The component extent of injury；Wherein, state transition probability matrix P is by the transition probability p between component_ijComposition, p_ijComponent i is referred to arrive The probability of component j, the dimension of matrix is the sum of software component；

(2) software is tested using the use-case quantity no less than minimum test case, counts being performed for each component Number is repaired in number of times, failure number and failure, is obtained each component and is performed probability；All structures in software under the quantity of minimum test case Part performs number and is no less than 2 times；Failure member will be repaired in real time in test process, and obtain repairing the probability of success；

(3) performed according to transition probability between the component obtained in step (1) and according to each component obtained in step (2) general Rate is calculated each component occupancy；

(4) each component failure rate is calculated；The crash rate of component refers to probability of the component from normal condition to abnormality, Comprise the following steps that：

Step 4.1：If certain component there are normal and abnormal two states, component level Markov Chain is formed, set up by g_u,vGroup Into state transition probability matrix G；g_u,vExpression is transferred to the probability of state v by state u, by being performed time in step (2) Number, failure number and failure are repaired number and are calculated；

Step 4.2：Calculate state probability vector B of the component after k cycle^(k)：B^(k)=B⁽⁰⁾G^(k), wherein, initially Probability vector B⁽⁰⁾=[0 1], G^(k)The k powers of representing matrix G；According to steady markovian property, B^(k)Will converge to Fixed value；

Step 4.3：Calculate the crash rate of the component；

The crash rate at current time is equal to state probability vector B⁽⁰⁾Second element, i.e. λ⁽⁰⁾=g₁₂；

Crash rate after k cycle is equal to the state probability vector B after convergence^(k)Second element, i.e.,

(5) each component complexity is calculated comprising function body quantity according to each component obtained in step (1)；

(6) each component importance degree is calculated according to each component extent of injury obtained in step (1)；

(7) each Member Reliability Analysis are assessed；According to following Member Reliability Analysis model obtain each component reliability assessment value and Predicted value；

R(λ_j,γ_j,π_j,μ_j；T)=exp (- γ_jλ_jπ_jμ_jT),

Wherein, λ_jIt is the crash rate of component j, γ_jIt is the occupancy of component j, π_jIt is the complexity of component j, μ_jIt is component j Importance degree, t be running software periodicity.

(8) software reliability is assessed；According to each Member Reliability Analysis obtained in following software reliability model and step (7) The reliability assessment value and predicted value of software are obtained, the reliability to software is estimated；

Wherein, λ_jIt is the crash rate of component j, γ_jIt is the occupancy of component j, π_jIt is the complexity of component j, μ_jIt is component j Importance degree, t be running software periodicity, n be software in number of components.

Preferably, in step (1), function body quantity in Software Detailed Design specification by obtaining；The component extent of injury refers to Be criticality of the component to whole software, be divided into high, medium and low and four grades of non-hazardous；

When component failures will cause software crash, the extent of injury for bringing about great losses is height；

When component failures will cause the major function of software to fail, but the extent of injury of software crash will not be caused to be；

When component failures will cause software disabler, but the extent of injury of the major function of software is not influenceed for low；

When the extent of injury that component failures do not interfere with the normal operation of software is nothing.

Further, in step (6), four grades of the component extent of injury assign the value successively decreased at random, it is high, medium and low and Without lev4~lev1 is corresponded to respectively, then the importance degree of component j is：

Preferably, in step (3), occupancy refers to the ratio that component is performed in task, the occupancy of component jWherein, b_iRepresent the probability for performing component i, p_ijExpression is gone to the probability of component j by component i.

Preferably, in step (5), the complexity of component jWherein, β_jIt is the function body quantity that component j is included, Obtained by step (1).

Preferably, in step (7), when the reliability assessment value that t=1 obtains current time component j is R (λ_j,γ_j,π_j, μ_j；T)=exp (- γ_jλ_jπ_jμ_j)；When the reliability prediction value that t=k obtains following k cycle rear part j isIn formula：K is not equal to 1,

Preferably, in step (8), when the reliability assessment value that t=1 obtains current time software isThe reliability prediction value of software is after t=k obtains the following k cycleIn formula：K is not equal to 1,

Compared with prior art, the present invention has following beneficial technique effect：

The present invention has taken into full account the influence of software architecture and component feature to software reliability, it is proposed that comprehensive to comment Estimate set of factors.Comprehensive assessment set of factors has not been concerned only with the shadow of Member Reliability Analysis that high-frequency calls to software systems reliability Ring, it is also considered that influence of the complexity and importance of component to software systems reliability.It is scalar grain to establish with function body The complexity measure model of degree, has taken into account the complexity of component and structural；Establish the weight that priority and exponential model are combined The property wanted measurement model, it is to avoid the subjective interference of artificial assignment.

Further, reliability assessment can be carried out to software component, software systems can also be entered using the model set up Row reliability assessment；Selected by the difference to the cycle in model and assignment, software current reliability can be estimated, The reliability in software later stage can be predicted, this global reliability to improve software systems wider using scope is controlled soft Part quality is significant.

Brief description of the drawings

Fig. 1 is appraisal procedure flow chart of the present invention.

Fig. 2 is the component level Markov Chain schematic diagram described in present example.

Specific embodiment

With reference to specific embodiment, the present invention is described in further detail, it is described be explanation of the invention and It is not to limit.

Method of the present invention is applied in the middle of the assessment of certain large-scale real-time software reliability, by considering structure The influence of part occupancy, complexity and importance degree to software reliability, establishes complexity model and importance degree model, uses for reference horse Er Kefu chain thoughts have been respectively created the Reliability Evaluation Model of component level and software systems level, software are carried out complete and comprehensive Reliability assessment.

As shown in figure 1, comprising the following steps that.

Step (1), by analyzing the large-scale real-time software structure, obtains application state transition probability matrix, each component bag Quantity containing function body and each component extent of injury.Wherein, function body refers to completing in component being called for specific function Code collection, is the chief component of component, can be by being obtained in Software Detailed Design specification.

State transition probability matrix P is by the transition probability p between component_i,jComposition, p_i,jRefer to component i to the general of component j Rate, the dimension of matrix is 6；The component extent of injury refers to criticality of the component to whole software, is divided into four grades：High, In, low, non-hazardous.Statistics is as shown in table 1-1；And the division of the component extent of injury and corresponding sequence number are as shown in table 1-2.

The table 1-1 software structure analysis tables

Component extent of injury classification chart in the table 1-2 softwares

Step (2), is tested software using the test case of 1932, counts being performed number of times, losing for each component Number is repaired in effect number, failure.Wherein, failure number refers to the number of times that component makes a mistake in testing；Repaired by introducing failure Mechanism so that failure member will be repaired in real time in test process such that it is able to obtain it and repair the probability of success, failure is repaiied Plural number refers to the number of times of success reparation after component failure.Statistics is as shown in table 2.

The use-case test recording table of table 2

Step (3), calculates each component occupancy.Occupancy refers to the ratio that component is performed in task, component j's Occupancy γ_jCan be tried to achieve by following formula：Wherein, b_iRepresent the probability for performing component i, p_ijExpression is gone to by component i The probability of component j.

Therefore current time each component occupancy in systems is γ⁽⁰⁾=(0.11,0.22,0.19,0.15,0.13, 0.2), each component occupancy in systems is after the following k cycle：γ^(k)=γ⁽⁰⁾P^(k)=(0.1061,0.2233, 0.1705,0.1706,0.1423,0.1871).

Step (4), calculates each component failure rate.The crash rate of component refers to component from normal condition to abnormality Probability, specific solution procedure is：

Step 4.1：If component there are normal and abnormal two states, component level Markov Chain is formed, as shown in Figure 2.Build Stand by g_u,vThe state transition probability matrix G, g of composition_u,vExpression is transferred to the probability of state v by state u, can be by step (2) Be performed number of times, failure number, failure repair number be calculated.The state transition probability matrix of 6 components is：

Step 4.2：Calculate state probability vector B of the component after k cycle^(k), its method is：B^(k)=B⁽⁰⁾G^(k), its In, probability vector B⁽⁰⁾=[0 1], G^(k)The k powers of representing matrix G.State probability vector after the k cycle of 6 components For：

Step 4.3：Calculate the crash rate of component.Crash rate is equal to the state probability vector B after convergence^(k) Second element, i.e. λ=g₁₂.Then the crash rate of crash rate composition is vectorial after the k cycle of each componentEach structure The crash rate vector λ=(λ of part current time crash rate composition₁,λ₂,…,λ₆)=(0.1296,0.0701,0.0485, 0.0380,0.0882,0.1189).

Step (5), calculates the complexity of each component.The complexity π of component j_jComputational methods be：Wherein, β_jFor the function body quantity that component j is included, can be obtained by table 1.Therefore the complexity of each component is：π={ π_j；J=1...6 }= (0.0857,0.2000,0.1429,0.2214,0.1929,0.1571).

Step (6), calculates the importance degree of each component.Component importance degree is relevant with the component extent of injury, and component is endangered into journey Four grades of degree assign the value successively decreased at random, and respectively lev4~lev1, assignment and corresponding importance degree are as shown in table 3, then The importance degree computing formula of component j is：

The component importance degree allocation table of table 3

Therefore the vector of the importance degree of each component composition is：μ={ μ_j；J=1...n }=(0.6412,0.5134,0.4111, 0.5134,0.3679,0.4111).

Step (7), assesses each Member Reliability Analysis.Value in model has been tried to achieve in step (3) to (6), and taking t=1 can Obtain the reliability assessment of current time component j：R(λ_j,γ_j,π_j,μ_j；T)=exp (- γ_jλ_jπ_jμ_j), the assessment result such as institute of table 4 Show：

The Member Reliability Analysis of table 4 assess table

Take the reliability prediction that t=k can obtain following k cycle rear part j： In formula：

Step (8), assesses software systems reliability.Value in model has been tried to achieve in step (3) to (6), takes t=1 Can obtain the reliability assessment value of current time software systems：Take t =k can obtain the reliability prediction of software systems after the following k cycle：Formula In：

Claims (8)

1. It is 1. a kind of to be based on markovian component-based software reliability estimation method, it is characterised in that to comprise the following steps,

   (1) by analysis software structure, application state transition probability matrix, each component are obtained and includes function body quantity and each component The extent of injury；Wherein, state transition probability matrix P is by the transition probability p between component_ijComposition, p_ijRefer to component i to component j Probability, the dimension of matrix is the sum of software component；

   (2) software is tested using no less than the use-case quantity of minimum test case, count each component be performed number of times, Number is repaired in failure number and failure, is obtained each component and is performed probability；All components are held in software under the quantity of minimum test case Line number is no less than 2 times；Failure member will be repaired in real time in test process, and obtain repairing the probability of success；

   (3) probability meter is performed according to transition probability between the component obtained in step (1) and according to each component obtained in step (2) Calculation obtains each component occupancy；

   (4) each component failure rate is calculated；The crash rate of component refers to probability of the component from normal condition to abnormality, specifically Step is as follows：

   Step 4.1：If certain component there are normal and abnormal two states, component level Markov Chain is formed, set up by g_u,vComposition State transition probability matrix G；g_u,vExpression is transferred to the probability of state v by state u, and number of times, mistake are performed by step (2) Effect number and failure are repaired number and are calculated；

   Step 4.2：Calculate state probability vector B of the component after k cycle^(k)：B^(k)=B⁽⁰⁾G^(k), wherein, probability Vectorial B⁽⁰⁾=[0 1], G^(k)The k powers of representing matrix G；According to steady markovian property, B^(k)Fixation will be converged to Value；

   Step 4.3：Calculate the crash rate of the component；

   The crash rate at current time is equal to state probability vector B⁽⁰⁾Second element, i.e. λ⁽⁰⁾=g₁₂；

   Crash rate after k cycle is equal to the state probability vector B after convergence^(k)Second element, i.e.,

   (5) each component complexity is calculated comprising function body quantity according to each component obtained in step (1)；

   (6) each component importance degree is calculated according to each component extent of injury obtained in step (1)；

   (7) each Member Reliability Analysis are assessed；Reliability assessment value and the prediction of each component are obtained according to following Member Reliability Analysis model Value；

   R(λ_j,γ_j,π_j,μ_j；T)=exp (- γ_jλ_jπ_jμ_jT),

   Wherein, λ_jIt is the crash rate of component j, γ_jIt is the occupancy of component j, π_jIt is the complexity of component j, μ_jIt is important for component j Degree, t is running software periodicity；

   (8) software reliability is assessed；Each Member Reliability Analysis according to being obtained in following software reliability model and step (7) are obtained The reliability assessment value and predicted value of software, the reliability to software are estimated；

   $R({\mathrm{\λ}}_{1,...,n},{\mathrm{\π}}_{1,...,n},{\mathrm{\μ}}_{1,...,n};t)=\underset{j=1}{\overset{n}{\Π}}\exp (-{\mathrm{\γ}}_j{\mathrm{\λ}}_j{\mathrm{\π}}_j{\mathrm{\μ}}_{j}t),$

   Wherein, λ_jIt is the crash rate of component j, γ_jIt is the occupancy of component j, π_jIt is the complexity of component j, μ_jIt is important for component j Degree, t is running software periodicity, and n is the number of components in software.
2. 2. according to claim 1 a kind of based on markovian component-based software reliability estimation method, its feature It is that in step (1), function body quantity in Software Detailed Design specification by obtaining.
3. 3. according to claim 1 a kind of based on markovian component-based software reliability estimation method, its feature It is that in step (1), the component extent of injury refers to criticality of the component to whole software, is divided into high, medium and low and without danger Four grades of evil；

   When component failures will cause software crash, the extent of injury for bringing about great losses is height；

   When component failures will cause the major function of software to fail, but the extent of injury of software crash will not be caused to be；

   When component failures will cause software disabler, but the extent of injury of the major function of software is not influenceed for low；

   When the extent of injury that component failures do not interfere with the normal operation of software is nothing.
4. 4. according to claim 3 a kind of based on markovian component-based software reliability estimation method, its feature It is in step (6), four grades of the component extent of injury to assign the value successively decreased at random, it is high, medium and low and without corresponding to respectively It is lev4~lev1, then the importance degree of component j is：
5. 5. according to claim 1 a kind of based on markovian component-based software reliability estimation method, its feature It is that in step (3), occupancy refers to the ratio that component is performed in task, the occupancy of component jIts In, b_iRepresent the probability for performing component i, p_ijExpression is gone to the probability of component j by component i.
6. 6. according to claim 1 a kind of based on markovian component-based software reliability estimation method, its feature It is, in step (5), the complexity of component jWherein, β_jIt is the function body quantity that component j is included, by step (1) Obtain.
7. 7. according to claim 1 a kind of based on markovian component-based software reliability estimation method, its feature It is, in step (7), when the reliability assessment value that t=1 obtains current time component j is R (λ_j,γ_j,π_j,μ_j；T)=exp (- γ_jλ_jπ_jμ_j)；When the reliability prediction value that t=k obtains following k cycle rear part j is In formula：K is not equal to 1, the k occupancy of cycle rear part jWherein, b_iRepresent and perform component i Probability.
8. 8. according to claim 1 a kind of based on markovian component-based software reliability estimation method, its feature It is, in step (8), when the reliability assessment value that t=1 obtains current time software isWhen The reliability prediction value that t=k obtains software after the following k cycle isFormula In：K is not equal to 1, the k occupancy of cycle rear part jWherein, b_iRepresent and perform component i's Probability.