CN102360332B - Software reliability accelerated test and evaluation method and computer-aided tool used in same - Google Patents

Abstract

The invention provides a software reliability accelerated test and evaluation method based on a deformed operational profile as well as a computer-aided tool used in the same, and belongs to the field of software reliability engineering. In the method, the deformed operational profile is constructed, and then software reliability growth and verification accelerated test and evaluation are carried out on tested software based on the profile. By using the tool, a traditional operational profile is constructed by a traditional operational profile construction module, the deformed operational profile is further constructed by a deformed operational profile construction module, a reliability verification accelerated test module and a reliability growth accelerated test module are executed according to selection of a user, and cases in test are generated by an automatic test case generation module. By adopting the software reliability accelerated test and evaluation method provided by the invention, unbiased estimation of reliability parameters between the deformed operational profile and the traditional Musa operational profile is realized and the test efficiency is better; and compared with the traditional test method, the method has the advantages of short test time and accurate test result, thus facilitating understanding and practical operation of engineering personnel.

Description

A kind of software reliability accelerated test and appraisal procedure and computer-aided tool thereof

Technical field

The invention belongs to the software reliability engineering field, relate in particular to software reliability test and evaluation areas, specifically, refer to a kind of software reliability accelerated test based on abnormalization Operation Profile and appraisal procedure and computer-aided tool thereof.

Background technology

Software reliability test is extremely important for the quality assurance of software, need to test and constantly get rid of the software defect that software reliability is had material impact by software reliability growth on the one hand, realize that the continuous growth of software reliability level is to meet the requirement of user to software reliability; Need on the other hand to judge by the software reliability validation test whether software has met the requirements of reliability level, and then determine that can software receive or issue.

But at present, the application of software reliability test is unsatisfactory, the huge test case of quantity under traditional reliability testing theory and method guidance is a major reason that causes it to hesitate to move forward, for highly reliable software, such problem is particularly outstanding, and Butler once thought and utilizes traditional software reliability validation test method can not complete assessment requirement to highly reliable software.How to improve the efficiency of traditional software reliability testing, accelerate the process of software reliability test, realize identical reliability testing target under the condition of limited time and resource consumption, namely realize software reliability accelerated test, become a focus and the difficulties of pendulum in face of the software reliability area research personnel.

Current traditional reliability testing and appraisal procedure mainly contain two kinds, a kind ofly are based on the statistical test method that markov uses a model, and another kind is based on the method for testing reliability of Musa Operation Profile.The basic thought of above-mentioned two kinds of methods all is based on statistical ultimate principle, namely according to the description of the statistical law of the actual service condition of software, software is carried out random test, utilizes fail data to carry out software reliability evaluation.These characteristics has determined that also the software reliability test process time is long, expense is high, resource cost is large, causes the inefficiency of software reliability test.

H.Hecht respectively at 1993 in its paper " Rare conditions-an important cause of failures " and a large amount of data by collecting in its paper " Rare Conditions and Their Effect on Software Failures " in 1994, sum up the great reason that highly reliable software lost efficacy, namely in highly reliable software, the inefficacy of small probability event is the main cause that causes thrashing, the issue of these data and the analysis of causes provide foundation for follow-up a series of propositions that are sampled to the method for acceleration mechanism with importance degree.B.Cukic, H.Alam, D.Tang etc. are per capita on the basis of H.Hecht data, thought based on the importance degree sampling has been proposed, probability of happening to small probability event promotes, the importance degree Sampling is described, and tests on the software based on Markov chain building Operation Profile.Below all be based on the importance degree principle software reliability accelerated method has been carried out extraordinary annotation, but the Markov of all using uses a model as setting forth object, and be all to occur that at hypothesis software thereby software defect assurance Software failure probability is not rejected in inefficacy or crash rate is constant basically, namely take the software reliability validation test as research object, and accelerated test is not set forth for software reliability growth.Therefore not tight with two types of combinations of present software reliability test, and, due to the complicacy of Markov chain structure itself, affected its application in practice.Therefore, how in the situation that take the Musa Operation Profile, as basis, study software reliability accelerated test, and software reliability checking accelerated test and software reliability growth accelerated test are comprehensively studied and considered, have important Research Significance and using value.

Summary of the invention

The present invention, in order to solve based on the inefficient problem of the software reliability test of traditional Musa Operation Profile, provides a kind of software reliability accelerated test based on abnormalization Operation Profile and appraisal procedure and computer-aided tool thereof.

The construction method of abnormalization Operation Profile provided by the invention, divide following two kinds:

Method 1: the probability of happening of normal operations is set to 0, and the probability of happening sum of key operation is as speedup factor.Described key operation need to refer to the operation of stress test, and described normal operations do not need to refer to the operation of stress test.

Method 2: utilize operation probability of happening value to carry out the size upset to the probability of happening value of all operations.

A kind of accelerated test of software reliability growth based on abnormalization Operation Profile of the present invention and appraisal procedure, concrete steps are:

Step 1, abnormalization of structure Operation Profile, determine the speedup factor of each operation and the speedup factor of whole abnormalization Operation Profile;

Step 2, according to the abnormalization Operation Profile of constructing, carry out the random sampling generating test use case;

Step 3, judgement tested software are discrete type software or continuous type software, if discrete type software performs step four, if continuous type software performs step five;

Step 4, structure STE, the test cases that execution step two generates, collect inefficacy interval number of cases, then determines to cause the corresponding operation of input of inefficacy, and utilize the speedup factor of determined operation to lose efficacy with the conversion processing of number of cases; This step rear execution step six that is finished;

Step 5, structure STE, the test case that execution step two generates, collect the time between failures data, then determine to cause the corresponding operation of input of inefficacy, and utilize the speedup factor of determined operation to carry out the conversion processing of time between failures; This step rear execution step six that is finished;

Step 6, the suitable software reliability growth model of selection, carry out the estimation of dependability parameter;

Whether step 7, judgement tested software this moment meet the reliability requirement of regulation, if, stop test, if not, turn step 2 and carry out.

A kind of checking accelerated test of software reliability based on abnormalization Operation Profile of the present invention and appraisal procedure, concrete steps are:

Step 1, abnormalization of structure Operation Profile, and the speedup factor of definite the entire profile;

Step 2, determine the test volume of tested software:

If tested software is discrete type software, can determine in advance for fixed number ending testing scheme and no-failure validation test scheme the test case amount that accelerated test is required, for example for no-failure validation test scheme, can determine the test case amount that accelerated test is required Wherein, [.] expression is right. round F _UThe higher limit of expression failure probability, α represents level of significance;

If tested software is continuous type software, can determine in advance for regularly end up testing scheme and no-failure validation test scheme the continuous execution time that accelerated test is required, for example for no-failure validation test scheme, can determine the required continuous execution time t ' of accelerated test=t Λ, wherein, the t representative needs the test duration of carrying out under the traditional operation section, Λ represents the speedup factor of whole abnormalization Operation Profile;

Step 3, the test volume of determining according to step 2, and the abnormalization Operation Profile that generates based on step 1, generating test use case, carry out the test case that generates, and collects fail message;

Step 4, in conjunction with validation test scheme and experimental result, received or rejected conclusion;

Step 5, according to the experimental result of reality, provide the assessment result of current software reliability parameter.

A kind of computer-aided tool of realizing said method of the present invention, comprise traditional operation profile construction module, abnormalization Operation Profile constructing module, test case automatically-generating module, reliability demonstration accelerated test module and reliability growth accelerated test module.

Traditional operation profile construction module receives user's input operation pattern information and operation information, set up the traditional operation section, and the Operation Profile information data of setting up is comprised evolutionary operator probability, types of variables and span, send abnormalization Operation Profile constructing module to; Abnormalization Operation Profile constructing module is according to the selection of user to the building method of abnormalization Operation Profile, generate abnormalization of the second Operation Profile that does not comprise abnormalization of the first Operation Profile of normal operations or comprise all operations, and preserve the speedup factor of each operation and the speedup factor of whole abnormalization Operation Profile.The information data of the abnormalization Operation Profile that abnormalization of test case automatically-generating module reception Operation Profile constructing module is come comprises evolutionary operator probability, types of variables and the span on abnormalization Operation Profile, test volume and validation test scheme generating test use case in conjunction with the reliability demonstration accelerated test module input that receives, the growth testing scheme generating test use case of perhaps in conjunction with the reliability growth accelerated test module that receives, inputting, and test case is exported to software under testing.reliability demonstration accelerated test module, the information data that receives the abnormalization Operation Profile that abnormalization Operation Profile constructing module comes comprises the evolutionary operator probability on abnormalization Operation Profile, types of variables, span and speedup factor, be achieved as follows concrete function: according to the validation test index of user's input, be at first tested discrete type software or continuous type Software Create validation test scheme, and definite test volume, the test volume and the validation test scheme that generate are exported to the test case automatically-generating module, then receive the user and move the collected fail data of testing software implementation of test cases in constructed STE, in conjunction with the validation test scheme, determine to receive or reject this tested software, and obtain the assessment result of current software reliability parameter according to experimental result.reliability growth accelerated test module, the information data that receives the abnormalization Operation Profile that abnormalization Operation Profile constructing module comes comprises the evolutionary operator probability on abnormalization Operation Profile, types of variables, span and speedup factor, be achieved as follows concrete function: according to the growth test index of user's input, be at first that tested discrete type software or continuous type Software Create increases testing scheme, to increase testing scheme and export to the test case automatically-generating module, receive the user and move the collected fail data of testing software implementation of test cases in constructed STE, determine to cause the corresponding operation of input of inefficacy, utilize the speedup factor of this operation to carry out the conversion processing of fail data, and the data conversion results of test result is exported.

The present invention compared with prior art, has obvious advantage and beneficial effect:

(1) software reliability accelerated test method and the appraisal procedure of the present invention's proposition, be based on that the abnormalization Operation Profile that generates on traditional operation section basis launches, solved the reliability accelerated test problem with Operation Profile generating test use case aspect, and from principle, scheme, implementation, outcome evaluation overall process, accelerated test method is covered, and around reliability demonstration and two kinds of reliability testing types of growth, launch, realized the reliability testing to software, cover the software reliability test process fully, had important directive significance.

(2) the present invention can utilize theory and the achievement of original reliability testing aspect, software reliability growth test model for example, transform by fail data, realize docking and the improvement on traditional validation test scheme basis of accelerated test and conventional model, these all make method and the existing achievement in research that the present invention proposes realize good continuity, rather than abandon the accent research of original achievement in research fully.

(3) by exemplary application, the present invention has testing efficiency preferably, compares the traditional test technology, and the test duration is short, and test result is also accurate, is convenient to engineering staff's understanding and practical operation.

Description of drawings

Fig. 1 is the schematic flow sheet of software reliability growth accelerated test of the present invention and appraisal procedure;

Fig. 2 is the Operation Profile of software and the expression schematic diagram in defect space;

Fig. 3 is the schematic flow sheet of software reliability checking accelerated test of the present invention and appraisal procedure;

Fig. 4 is the modular structure figure of computer-aided tool of the present invention.

Wherein,

1-traditional operation profile construction module, abnormalization of 2-Operation Profile constructing module, 3-test case automatically-generating module, 4-reliability demonstration accelerated test module, 5-reliability growth accelerated test module.

Embodiment

The present invention is described in detail below in conjunction with drawings and Examples.

The construction method of abnormalization Operation Profile of the present invention has two kinds, below two kinds of methods is illustrated.

Method 1: the probability of happening of normal operations is set to 0, with the probability of happening sum of key operation as speedup factor.Be specially: will need the operation of stress test as key operation, and not need the operation of stress test as normal operations, establish undiscoveredly has the defect of significant impact to be present in key operation to reliability more.Division for key operation and normal operations, a kind of method is to start with from small probability event, reason is as follows: 1) to operate in traditional test lower due to self probability of happening in the stage for small probability, the test case of being distributed is less, and therefore mistake exposes relatively late or also do not expose complete; 2) much software continuous is moved the operation that plays significant role,, as operations such as failure recovery, abnormality processing, all belong to the small probability operation; 3), although key operation mainly is present in small probability operation, and not all small probability operations all belong to key operation, although some operation probability of happening is not high, the realization of whole software function, without direct impact, is not listed among key operation yet.In method 1, make in abnormalization Operation Profile and only comprise key operation, namely establish undiscovered defect and mainly be present in key operation, therefore can ignore the impact of normal operations, then realize redistributing of evolutionary operator probability according to equal proportion raising and normalized method, have:

${\mathrm{pc}}_i^{\′}=\frac{{\mathrm{pc}}_i}{{\mathrm{pc}}_1+{\mathrm{pc}}_2+...+{\mathrm{pc}}_k}---\left(1\right)$

Wherein, k represents the number of key operation,

pc _iRepresent the probability of happening of i key operation on the traditional operation section, pc ' _iRepresent the probability of happening of i key operation on abnormalization Operation Profile, pc ₁+ pc ₂+ ... + pc _kFor the key operation sum on the traditional operation section.

Therefore, the speedup factor Λ of section is:

Λ＝pc ₁+pc ₂+…+pc _k (2)

The speedup factor Λ of i key operation _iFor:

Λ _i＝pc ₁+pc ₂+…+pc _k (3)

It is the speedup factor that the speedup factor of any key operation all is same as the entire profile.

Method 2: utilize operation probability of happening value to carry out the size upset to the probability of happening value of all operations.Be specially:

$p_i^{\′}=\frac{p_{\max }-p_i}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(p_{\max }-p_i)}---\left(4\right)$

In formula, p _maxRepresent the maximum probability of happening among all operations in the traditional operation section, p _iRepresent the probability of happening of i operation on the traditional operation section, p ' _iRepresent the probability of happening of i operation on abnormalization Operation Profile, n represents the number of the operation on the traditional operation section, and the operation number on abnormalization Operation Profile is lacked 1 than the operation number on the traditional operation section.

At this moment, the speedup factor Λ of the entire profile is:

$\mathrm{\Λ}=\frac{1}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{p_i^{\′2}}{p_i}}---\left(5\right)$

The speedup factor Λ of i key operation _iFor:

${\mathrm{\Λ}}_i=p_i\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(p_{\max }-p_i)}{p_{\max }-p_i}---\left(6\right)$

Obviously, the utilization of method 1 need to possess more rich posterior infromation or have certain prior imformation, and have higher confidence, think that defect remaining, that reliability is had a considerable influence only is present in the key operation of small probability and appointment, method itself has certain risk.When not possessing such prior imformation, preferably adopt method 2.Since normal operations is compared sufficient test in the software test in early stage, the operation of therefore large probability has obtained test more fully, utilize the size upset of maximum operation probability of happening value realization to the probability of happening value of all operations this moment, not only comprise key operation in abnormalization Operation Profile like this, also comprise normal operations, can make up in method 1 subjectivity while determining key operation, realize the compensatory test to all operations.

Abnormalization Operation Profile based on the present invention proposes, the invention provides a kind of software reliability accelerated test based on abnormalization Operation Profile and appraisal procedure and computer-aided tool thereof.Software reliability accelerated test and appraisal procedure are respectively for software reliability growth test and two types of expansion of validation test, each, again respectively for discrete type software and continuous type software, namely specifically comprises following 4 kinds of methods: 1) based on the accelerated test of discrete type software reliability growth and the appraisal procedure of abnormalization Operation Profile; 2) based on the accelerated test of continuous type software reliability growth and the appraisal procedure of abnormalization Operation Profile; 3) verify accelerated test and appraisal procedure based on the discrete type software reliability of abnormalization Operation Profile; 4) verify accelerated test and appraisal procedure based on the continuous type software reliability of abnormalization Operation Profile.Specifically being divided into (1) software reliability growth accelerated test and appraisal procedure and the checking accelerated test of (2) software reliability and appraisal procedure is specifically described respectively.

As shown in Figure 1, the accelerated test of the software reliability growth based on abnormalization Operation Profile provided by the invention and appraisal procedure, concrete steps are as follows:

Step 1, abnormalization of structure Operation Profile, determine the speedup factor of each operation and the speedup factor of whole abnormalization Operation Profile.Be specially: at first according to the structure of Operation Profile in traditional reliability testing and appraisal procedure, for example adopt the make of Musa Operation Profile, structure traditional operation section, then according to method 1 or the method 2 of above-mentioned abnormalization of structure Operation Profile, construct abnormalization Operation Profile, and obtain the speedup factor of each operation and the speedup factor of whole abnormalization Operation Profile.

Step 2, according to the abnormalization Operation Profile of step 1 structure, carry out the random sampling generating test use case, detailed process is: the occurrence of each input variable in the operation of at first definite tested software, by the sampling of extraction operation with these two steps of value that extract each input variable, generating test use case.

Step 3, judgement tested software are discrete type software or continuous type software, if discrete type software performs step four, if continuous type software performs step five.

Step 4, according to tested software characteristics and reliability testing, require to build the reliability testing environment, operating software in the reliability testing environment, the test case that generates in execution step two, and collect and record fail data.For discrete type software, the fail data of collection just is inefficacy interval number of cases.

Then determine to cause the corresponding operation of input of inefficacy, and the speedup factor of utilizing determined operation lost efficacy with the conversion processing of number of cases, for discrete type software, just with the inefficacy interval of collecting, multiply by the speedup factor of definite respective operations with number of cases, the data that obtain are exactly the data after conversion processing.This step rear execution step six that is finished.

For discrete type software, the principle of the accelerated test of the reliability growth based on abnormalization Operation Profile provided by the invention is as follows:

Suppose that α is operation O _iThe middle a certain defect that can cause software failure, such defect also have β, γ ..., μ etc., as shown in Figure 2, wherein, the input domain S that can cause defect α to be triggered _αRepresent, use | S _α| expression input domain S _αIn the number of contained input data, be called S _αMould.Because in the reliability growth test process, lost efficacy and all by some defects, caused each time, and in increasing test process, in a single day defect is found just to be excluded immediately, therefore, the inefficacy that each defect causes only can occur once, and suppose between defect independent of one another, therefore, under the traditional operation section, for discrete type software, when having carried out n test case, while having occurred once to lose efficacy, establish this inefficacy herein and caused by defect α, the probability P of the software failure that is caused by defect α _fαCan be expressed as:

$P_{\mathrm{f\α}}=p_i\frac{\left|S_{\mathrm{\α}}\right|}{\left|S_{D_i}\right|}---\left(7\right)$

In formula, p _iOperation O _iProbability of happening in the traditional operation section, | S _α| the input domain S that expression can cause defect α to be triggered _αIn the number of contained input data,

Expression operation O _iCorresponding input domain

In the number of contained input data.

In actual test, when having tested n use-case, the inefficacy that once due to defect α, causes having occurred, has used this estimated value as formula (7), has had:

$P_{f_{\mathrm{\α}}}=\frac{1}{n}---\left(8\right)$

Two formulas above simultaneous have:

$p_i\frac{\left|S_{\mathrm{\α}}\right|}{\left|S_{D_i}\right|}=\frac{1}{n}---\left(9\right)$

In like manner, under abnormalization Operation Profile, also have following formula to exist:

$p_i^{\′}\frac{\left|S_{\mathrm{\α}}\right|}{\left|S_{D_i^{\′}}\right|}=\frac{1}{n^{\′}}---\left(10\right)$

Although operation _QVariation has occurred in the probability of happening of i under traditional operation section and abnormalization Operation Profile of the present invention, but the input domain of this operational correspondence is constant, namely

Therefore have:

p _i/p′ _i＝n′/n (11)

Namely have:

n＝n′p′ _i/p _i＝n′/Λ _i (12)

That is to say, test the individual use-case of n ' under abnormalization Operation Profile and find once to lose efficacy, be equivalent to test under the traditional operation section

Individual use-case is found once to lose efficacy.

Step 5, according to tested software characteristics and reliability testing, require to build the reliability testing environment, operating software in the reliability testing environment, the test case that generates in execution step two, and collect and record fail data.For continuous type software, the fail data of collection is time between failures.

Then determine to cause the corresponding operation of input of inefficacy, and the speedup factor of utilizing determined operation is carried out the conversion processing of time between failures, for continuous type software, multiply by the speedup factor of definite respective operations with the time between failures of collecting, the data that obtain are exactly the data after conversion processing.This step rear execution step six that is finished.

For continuous type software, the principle of the accelerated test of the reliability growth based on abnormalization Operation Profile provided by the invention is as follows:

If it is P that i defect makes the out of order probability of software _i, establishing this defect is α, without loss of generality, P is arranged _i=P _fα, P _fαFor the probability of the software failure that caused by defect α, for continuous type software, formula (7) is applicable equally.If software execution within the unit interval (unit interval can be minute, time, sky, week etc.)

Subtask, namely be k task time, unit interval i defect probability λ that software is broken down _iFor:

λ _i＝P _i/k＝P _fα/k (13)

Software puts into operation under random initial conditions, the time of to appearance, losing efficacy is called time to failure (time to failure), and this time to failure is a stochastic variable, and its expectation value is mean time to failure (mean time to failure, be called for short MTTF), be denoted as θ.Utilize under the traditional operation section and tested the t time inefficacy that once due to defect α, causes has occurred, be used as the estimated value of formula (13), have:

${\mathrm{\λ}}_i=\frac{1}{t}---\left(14\right)$

Have: P _fα/ k=1/t (15)

Namely have: $p_i\frac{\left|S_{\mathrm{\α}}\right|}{\left|S_{D_i}\right|}\frac{1}{k}=\frac{1}{t}---\left(16\right)$

In like manner, under abnormalization Operation Profile, the inefficacy that once due to defect α, causes has occurred in the t ' time of having tested, and has following formula to exist:

$p_i^{\′}\frac{\left|S_{\mathrm{\α}}\right|}{\left|S_{D_i^{\′}}\right|}\frac{1}{k}=\frac{1}{t^{\′}}---\left(17\right)$

Although operate Q under traditional operation section and abnormalization Operation Profile _iProbability of happening variation has occurred, but the input domain of this operational correspondence is constant, namely

Therefore have:

p _i/p′ _i＝t′/t (18)

Namely have: t=p ' _iT '/p _i=t '/Λ _i(19)

That is to say, test the t ' time under abnormalization Operation Profile and find once to lose efficacy, be equivalent to test under the traditional operation section

Time is found once to lose efficacy.

The software reliability growth model that step 6, utilization are fit to carries out the estimation of dependability parameter.Generally, for discrete type software, the corresponding discrete type NHPP of reliability growth model recommendations for selection model carries out the estimation of dependability parameter; For continuous type software, reliability growth model can be selected from the continuous type NHPP models such as GO, MO, carries out the estimation of dependability parameter.

Whether step 7, judgement tested software this moment meet the reliability requirement of regulation,, if meet the reliability requirement of regulation, stop test; Carry out otherwise turn step 2, according to the abnormalization Operation Profile of constructing in step 1, append generating test use case, repeating step two is to step 7, until the estimated value of dependability parameter stops test while meeting the reliability requirement of regulation.

The principle of the checking of the software reliability based on abnormalization Operation Profile accelerated test provided by the invention is:

The characteristics of software reliability validation test are, in the process of software test, when occurring to lose efficacy, to causing the software defect that lost efficacy, do not reject, in like manner, the software reliability checking accelerated test under abnormalization Operation Profile can not carried out the rejecting of software defect yet, therefore, wish by accelerated test, can obtain partially estimating with the consistent nothing to Software failure probability/crash rate of traditional Operation Profile under abnormalization Operation Profile.According to the definition of traditional operation section, the input space D of software can be expressed as:

D＝{(x ₁，x ₂，…，x _n)：Ф(x _i)＝Oor1，i＝1，2，…，n，x _i∈D _i，P(D _i)＝p _i} (20)

In formula, x _iInput domain D is taken from expression _iThe input data, n represents the number of the operation in the traditional operation section, D _iOperation O in expression and traditional operation section _iCorresponding input domain, so P (D _i) be equal to the operation O in the traditional operation section _iProbability of happening, i.e. P (D _i)=p _iФ (x) is the function that a 0-1 distributes, and has following feature: when software occurred to lose efficacy on input x, Ф (x) value was 1; When software did not lose efficacy on x, Ф (x) value was 0, so the indicator function that Ф (x) is otherwise known as and lost efficacy.Make Y=Ф (x), the failure probability of software can be expressed as the mathematical expectation of the function Y of stochastic variable, and namely failure probability F is:

$F=E\left(Y\right)=\underset{x\∈D}{\mathrm{\Σ}}\mathrm{\Φ}\left(x\right)P\left(x\right)---\left(21\right)$

P in formula (x)=p _i, as x ∈ D _iThe time, and

If Q (x) is any one is defined in discrete distribution on D, and if P (x) ≠ 0, Q (x) ≠ 0,

$F=\underset{x\∈D}{\mathrm{\Σ}}\mathrm{\Φ}\left(x\right)\frac{P\left(x\right)}{Q\left(x\right)}Q\left(x\right)=\underset{x\∈D}{\mathrm{\Σ}}\mathrm{\Φ}\left(x\right)\mathrm{\Λ}\left(x\right)Q\left(x\right)---\left(22\right)$

Order

Y '=Ф (x) Λ (x), Q (x)=p ' _i, as x ∈ D _iThe time, and

F＝E(Y)＝E(Y′) (23)

It is the mathematical expectation that the failure probability of software is expressed as the function of the stochastic variable on abnormalization Operation Profile, the mathematical expectation of the function Y ' that failure probability of software can be by stochastic variable=Ф (x) Λ (x) obtains, namely under abnormalization Operation Profile, the probability of happening of each operation is by p _iBecome p ' _i, in order to make up at p ' _iThe difference of allocation for test use-case under probability distribution, introduce speedup factor

Carry out zero deflection, under same test case condition, by the change of probability distribution Q (x), importance sampling is carried out in the operation that need are paid close attention to, complete zero deflection between Y ' and Y by speedup factor Λ (x), and then complete the nothing of Software failure probability is estimated partially.

As shown in Figure 3, the present invention is based on software reliability checking accelerated test and the appraisal procedure of abnormalization Operation Profile, concrete steps are as follows:

Step 1, abnormalization of structure Operation Profile, and the speedup factor of definite the entire profile.Adopt the building method of the abnormalization Operation Profile of the present invention's proposition to realize in this step.

Step 2, determine the test volume of tested software, distinguishing tested software is discrete type software or two kinds of situations of continuous type software., for continuous type software, utilize based on the continuous type software reliability checking accelerated test method of abnormalization Operation Profile and determine the continuous execution time that accelerated test is required; , for discrete type software, utilize based on the discrete type software reliability checking accelerated test method of abnormalization Operation Profile and determine the test case amount that accelerated test is required.

Step 3, the test volume of determining according to step 2, and the abnormalization Operation Profile that generates based on step 1, generating test use case, implementation of test cases, and collect fail message.

Step 4, in conjunction with the result data of collecting in the criterion of the reception that provides in the validation test scheme, rejection and actual test, can access and receive or reject conclusion.

Step 5, according to the actual result data of collecting, provide the assessment result of current software reliability parameter.Adopt no reliability estimation method to carry out the software reliability parameter evaluation for discrete type software and continuous type software.

The software reliability of the discrete type based on abnormalization Operation Profile checking accelerated test method described in step 2 is:

Common discrete type software reliability validation test scheme has: Based on Censored Data test, sequential test and no-failure operation test etc.Because no-failure operation testing scheme is applicable to highly reliable software, so the present invention only describes as an example of no-failure operation test example, and Method And Principle may extend to the testing scheme of other type.

It is generally acknowledged, the discrete type software failure is to obey binomial distribution,, because the conjugation of binomial distribution is distributed as Beta (a, b) function, use for reference the thought that Bayes distributes, use Beta (a in the present invention, b) function, as the prior distribution of discrete type software, adopts the form of B (a, b) to represent below in formula, requirement according to degree of confidence obtains:

P(F＜F _U)≥1-α (24)

In formula: F _UThe higher limit of expression failure probability.1-α is the degree of confidence requirement that the user provides, and usually α gets one less than 1 number, such as 0.1,0.2 etc.

The probability density function of failure probability F is

$f(f/r,N,\mathrm{1,1})=\frac{f^r{(1-f)}^{N-r}}{B(1+r,1+N-r)}---\left(25\right)$

In formula: r table non ageing number, N represents the test case number, therefore has:

${\∫}_0^{F_U}\frac{{(1-f)}^N\mathrm{df}}{B(\mathrm{1,1}+N)}\≥1-\mathrm{\α}---\left(26\right)$

As can be known, the minimum test case number of the needs under the traditional operation section is:

$N=\left[\frac{\ln \mathrm{\α}}{\ln (1-F_U)}\right]---\left(27\right)$

In formula: [.] expression is right. round.

While calculating test case under acceleration environment, the test case number N under the traditional operation section is equivalent to the test case number under abnormalization Operation Profile

Therefore N is replaced with

Substitution, that is:

$f(r,\frac{N^{\′}}{\mathrm{\Λ}},\mathrm{1,1})=\frac{f^r{(1-f)}^{N^{\′}/\mathrm{\Λ}-r}}{B(1+r,1+N^{\′}/\mathrm{\Λ}-r)}---\left(28\right)$

Minimum test case number that can be required under abnormalization Operation Profile is:

$N^{\′}=\left[\mathrm{\Λ}\frac{\ln \mathrm{\α}}{\ln (1-F_U)}\right]---\left(29\right)$

Due to the speedup factor Λ of abnormalization Operation Profile＜1, by formula (27) and formula (29) as can be known, needed test case amount under abnormalization Operation Profile condition less than the normal operations section under needed test case amount.

The software reliability of the continuous type based on abnormalization Operation Profile checking accelerated test method described in step 2 is:

Common continuous type software verification testing scheme has: Censoring test, sequential test and no-failure operation test etc.Because no-failure operation test is applicable to highly reliable software, the present invention only describes as an example of no-failure operation test example.

, for continuous type software, generally by calculating the continuous execution time, realize validation test, because the Failure count of continuous type software is to obey Poisson distribution.Simultaneously, in order to shorten the test duration, the present invention uses for reference the thought that Bayes distributes and realizes, the conjugate prior of Poisson distribution is distributed as Gamma and distributes, and according to Gamma, distributes and the condition such as degree of confidence, and following equation can be arranged:

${\∫}_0^{{\mathrm{\λ}}_0}\mathrm{Gamma}(a_0,b_0+t)\mathrm{d\λ}=1-\mathrm{\β}---\left(30\right)$

In formula: a ₀, b ₀Gamma distribution parameter for intrinsic, can obtain by prior imformation, and the t representative needs the test duration of carrying out under the traditional operation section, and β is user's risk, λ ₀The higher limit of expression crash rate.

During test duration of calculate to need carrying out under acceleration environment, because the test duration t under the traditional operation section is equivalent to test duration under abnormalization Operation Profile

Therefore, t is replaced with

Substitution, that is:

t′＝tΛ (31)

The t ' that calculates is the needed test duration under abnormalization Operation Profile, due to the speedup factor Λ of abnormalization Operation Profile＜1, t '＜t as can be known, namely the needed continuous execution time under abnormalization Operation Profile condition less than the traditional operation section under the needed continuous execution time.

The reliability estimation method for discrete type software reliability checking accelerated test described in step 5 of the present invention is:

The reliability assessment of discrete type software is completed by failure probability (or success ratio), and the failure probability point estimate F ' under abnormalization Operation Profile can be expressed as:

$F^{\′}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{n_i^{\′}}{N_i^{\′}}{\mathrm{\Λ}}_i{p}_i^{\′}---\left(32\right)$

In formula: n ' _iFor falling into operation o _iIn test case in the use-case number that occur to lose efficacy, N ' _iFor falling into operation o _iIn the total number of test case.In addition, if establish the use number of cases of total of n ' for occurring in the test case under abnormalization Operation Profile to lose efficacy, N ' is total test case number, and when utilizing method 1 of the present invention to generate abnormalization Operation Profile, the point estimation that is obtained the tested software failure probability by (32) is:

$F^{\′}={\mathrm{\Λ}}_i\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{n_i^{\′}}{N_i^{\′}}{p}_i^{\′}=\mathrm{\Λ}\frac{n^{\′}}{N^{\′}}---\left(33\right)$

Therefore, the success ratio of software is:

p＝1-F′ (34)

Namely do not comprising under the abnormalization Operation Profile of normal operations, p is the point estimate of the success ratio of whole software.

If employing method 2 is produced abnormalization Operation Profile, the point estimation of the failure probability of tested software is arranged:

$F^{\′}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{n_i^{\′}}{N_i^{\′}}{p}_i---\left(35\right)$

In formula: p _iFor operating o in the traditional operation section _iProbability of happening.

But comprising under the abnormalization Operation Profile of normal operations, due to the speedup factor Λ of each operation _iNot identical, so failure probability meets formula (35).

Below provide the interval estimation of the failure probability when utilizing method 1 of the present invention to generate abnormalization Operation Profile.

The confidence level of F ' is the confidence lower limit F ' of 1-α _LWith confidence upper limit F ' _UBe respectively:

$F_L^{\′}=\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_1,v_2)}---\left(36\right)$

In formula: v ₁=2 (N '-n '+1), v ₂=2n '.

$F_U^{\′}=\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_2,v_1)}---\left(37\right)$

In formula: v ₁=2 (N '-n '), v ₂=2 (n '+1).

The confidence level that can be obtained success ratio p by formula (36), (37) is the confidence lower limit p of 1-α _LWith confidence upper limit p _UBe respectively:

$p_L=1-\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_2,v_1)}---\left(38\right)$

In formula: v ₁=2 (N '-n '), v ₂=2 (n '+1), F _1-α(v ₂, v ₁) expression F Distribution Value.

$p_U=1-\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_1,v_2)}---\left(39\right)$

In formula: v ₁=2 (N '-n '+1), v ₂=2n '.

When n '=0, the one-sided confidence lower limit of success ratio p is:

$p_L=1-\mathrm{\Λ}(1-\sqrt[N^{\′}]{\mathrm{\α}})---\left(40\right)$

The reliability estimation method for continuous type software reliability checking accelerated test described in step 5 of the present invention is:

The reliability assessment of continuous type software is completed by crash rate, and the formula below the present invention has used for reference is assessed the higher limit λ ' of crash rate _upp:

${\mathrm{\λ}}_{\mathrm{upp}}^{\′}=\frac{{\mathrm{\χ}}_{1-\mathrm{\α};{2n}^{\′}+2}^2}{2t^{\′}}---\left(41\right)$

In formula: α represents level of significance (1-α represents degree of confidence),

That card side distributes,

Can be used for assessing the upper limit of crash rate, t ' is the test duration on abnormalization Operation Profile.The total number that lost efficacy occurs in n ' in test case under abnormalization Operation Profile.

When n '=0, formula (41) can be reduced to:

${\mathrm{\λ}}_{\mathrm{upp}}^{\′}=\frac{-\ln \left(\mathrm{\α}\right)}{t^{\′}}---\left(42\right)$

So crash rate λ of tested software _uppFor:

λ _upp＝Λλ′ _upp (43)

Do not comprising under the abnormalization Operation Profile of normal operations the λ of (43) formula _uppIt is the crash rate assessment result of whole software.

Comprising under the abnormalization Operation Profile of normal operations, because normal operations is more abundant in test before, thus can suppose that the crash rate upper limit residual in normal operations can be higher than the crash rate of key operation, and then can use 2 λ _uppBe used as comprising the crash rate assessed value under the abnormalization Operation Profile of normal operations.

the present invention also provides a kind of computer-aided tool, it is the device of realizing said method of the present invention, it can support the foundation of abnormalization Operation Profile, the selection of key operation, two kinds of methods that build abnormalization Operation Profile according to the present invention build two kinds of abnormalization Operation Profile, determining of the speedup factor of operation and the speedup factor of whole abnormalization Operation Profile, and accelerate flow process and the method for validation test according to software reliability growth accelerated test in the present invention and reliability, provide respectively for the reliability of discrete type software and continuous type software and accelerate determining of validation test scheme, the importing of fail result, receive or reject the judgement of result and point estimation and the interval estimation of dependability parameter, provide simultaneously the reliability growth accelerated test fail data for discrete type software and continuous type software to import, fail data transforms and conversion results derives module.As shown in Figure 4, computer-aided tool provided by the invention specifically comprises: traditional operation profile construction module 1, abnormalization Operation Profile constructing module 2, test case automatically-generating module 3, reliability demonstration accelerated test module 4, reliability growth accelerated test module 5.

Traditional operation profile construction module 1 receives user's input operation pattern information and operation information, set up the traditional operation section, and, with the Operation Profile information data of setting up, comprise evolutionary operator probability, types of variables and span etc., send abnormalization Operation Profile constructing module 2 to.Abnormalization Operation Profile constructing module 2, according to the selection of user to the building method of abnormalization Operation Profile, generates the first abnormalization Operation Profile or abnormalization of the second Operation Profile, and preserves the speedup factor of each operation and the speedup factor of whole abnormalization Operation Profile.4 pairs of discrete type softwares to be tested of reliability demonstration accelerated test module or continuous type software carry out software reliability checking accelerated test and assessment, and 5 pairs of discrete type softwares to be tested of reliability growth accelerated test module or continuous type software carry out software reliability growth accelerated test and assessment.Test case automatically-generating module 3 is used for generating test use case, and exports to reliability demonstration accelerated test module 4 or reliability growth accelerated test module 5.

Described traditional operation profile construction module 1 is used for receiving operation information under user's input operation pattern and associative mode to set up the conventional mode of operation section, and receive user's input operation information, comprise types of variables and span, set up the traditional operation section in conjunction with the conventional mode of operation section of setting up, and the integrate information in the profile information in the conventional mode of operation section and traditional operation section is stored.General traditional operation profile construction module 1 adopts the building method of Musa Operation Profile, constructs the traditional operation section.

Described abnormalization Operation Profile constructing module 2 generates abnormalization of the first Operation Profile or abnormalization of the second Operation Profile according to the user automatically to the selection of the building method of abnormalization Operation Profile, when generating abnormalization of the first Operation Profile, because of a taking into account critical operation, therefore also need the user to select for key operation, after generating abnormalization Operation Profile, speedup factor of each definite operation and the speedup factor of whole abnormalization Operation Profile are stored.Described abnormalization of the first Operation Profile adopts the construction method of abnormalization of the first Operation Profile of the present invention, and abnormalization of the second Operation Profile adopts the construction method of abnormalization of the second Operation Profile of the present invention.

The testing scheme generating test use case that described test case automatically-generating module 3 is used for according to input, carry out the bug check before data generate, generation and the preservation of test case, with the test case output software under testing 6 that generates.The information data of the abnormalization Operation Profile that test case automatically-generating module 3 abnormalization of reception Operation Profile constructing modules 2 are come, be included in evolutionary operator probability, types of variables and span etc. on abnormalization Operation Profile, according to the validation test scheme generating test use case of reliability demonstration accelerated test module 4 inputs, the growth testing scheme generating test use case of perhaps according to reliability growth accelerated test module 5, inputting.

The information data that described reliability demonstration accelerated test module 4 receives the abnormalization Operation Profile that abnormalization Operation Profile constructing modules 2 send over comprises evolutionary operator probability, types of variables, span and the speedup factor on abnormalization Operation Profile, the function that realize that the input of checking index, the abnormalization Operation Profile that receive discrete type software and continuous type software are selected, checking accelerated test scheme output and test result imports, receives or reject decision model and dependability parameter assessment.Specifically, the function that reliability demonstration accelerated test module 4 realizes is: according to the validation test index of user's input, be at first tested discrete type software or continuous type Software Create validation test scheme, and definite test volume, the test volume and the validation test scheme that generate are exported to test case automatically-generating module 3, software under testing 6 is preserved and exported to test case automatically-generating module 3, according to validation test scheme generating test use case.The user moves software under testing 6 in constructed STE, implementation of test cases, collect fail data, and fail data is inputed to reliability demonstration accelerated test module 4, reliability demonstration accelerated test module 4 is in conjunction with the validation test scheme, determine receive or reject this tested software, and obtain the assessment result of current software reliability parameter according to experimental result.For discrete type software, test volume refers to the test case amount, and for continuous type software, test volume refers to the continuous execution time.

The information data of the abnormalization Operation Profile that described reliability growth accelerated test module 5 abnormalization of reception Operation Profile constructing modules 2 send over comprises evolutionary operator probability, types of variables, span and the speedup factor on abnormalization Operation Profile, and realization is for functions such as the growth test index input that receives discrete type software and continuous type software, the selection of abnormalization Operation Profile and test result importing, test result data conversion and outputs.specifically, the function that reliability growth accelerated test module 5 realizes is: according to the growth test index of user's input, be that tested discrete type software or continuous type Software Create increases testing scheme, to increase testing scheme and export to test case automatically-generating module 3, test case automatically-generating module 3 is according to increasing the testing scheme generating test use case, preserve and export to software under testing 6, the user moves software under testing 6 in constructed STE, implementation of test cases, collect fail data, and fail data is input to reliability growth accelerated test module 5, reliability growth accelerated test module 5 determines to cause the corresponding operation of input of inefficacy, utilize the speedup factor of this operation to carry out the conversion processing of fail data, and the data conversion results of test result is exported.

, for the assessment of software reliability growth accelerated test, select suitable software reliability growth model in prior art to carry out the estimation of dependability parameter, and judge whether tested software meets the reliability requirement of regulation.

The embodiment of the present invention, only take continuous type software as representative, is carried out respectively the method explanation of software reliability growth accelerated test and checking accelerated test.

Embodiment 1: the test experiments below by certain software S describes software reliability growth accelerated test provided by the invention and appraisal procedure.Table 1 is the expression signal of traditional operation section in the prior art of software S foundation.

The traditional operation section of table 1 software S

Operation

o 1

o 2

…

o i

…

o m

Input domain

D 1

D 2

…

D i

…

D m

Probability of happening P (x)

p 1

p 2

…

p i

…

p m

Table 2 is illustrated for the expression of abnormalization of the present invention Operation Profile that software S sets up.

The abnormalization Operation Profile of table 2 software S

This experiment adopts certain merchandise sales management system software to build tested software, at first utilizes the building method of Musa Operation Profile to build traditional Operation Profile, as shown in the 1st row and the 2nd row of table 3:

Table 3 builds the Operation Profile of merchandise sales management system

Method 1 according to abnormalization of structure Operation Profile provided by the invention builds abnormalization of the first Operation Profile, method 2 according to abnormalization of structure Operation Profile provided by the invention builds abnormalization of the second Operation Profile, as shown in the row of the 3rd in table 3 and the 5th row, wherein key operation is indicated with adding the surplus body.As can be seen from Table 3, all operations probability sum in abnormalization Operation Profile is all 1, but only comprises key operation in abnormalization of the first Operation Profile, and the normal operations probability is all 0; Comprised normal operations and both comprised key operation in abnormalization of the second Operation Profile.

Respectively based on abnormalization of the first Operation Profile, abnormalization of the second Operation Profile generating test use case and test, collect fail data, and according to the speedup factor of the operation of the input correspondence of fail data, carry out the conversion of fail data, obtain the fail data shown in table 5 and table 6.

Table 5 fail data that test obtains based on abnormalization of the first Operation Profile and the data after conversion

The defect sequence number	Time between failures (s)	Speedup factor	Time between failures after conversion (s)
				1	702	0.036601	19179.8
2	1164	0.036601	31802.41
				3	1080	0.036601	29507.39
4	3168	0.036601	86555.01
				5	882	0.036601	24097.7
6	10512	0.036601	287205.3
				7	13536	0.036601	369826
8	5400	0.036601	147537

Table 6 fail data that test obtains based on abnormalization of the second Operation Profile and the data after conversion

The defect sequence number	Time between failures (s)	Speedup factor	Time between failures after conversion (s)
				1	666	0.096383	6909.946
2	180	0.096383	1867.553
				3	2718	0.096383	28200.05
4	900	0.079889	11265.56
				5	31482	0.096383	326635
6	68616	0.079889	858886.4
				7	5508	0.096383	57147.12
8	32562	0.096383	337840.4

By Software Reliability Estimation Tools, select suitable model to carry out the assessment of dependability parameter, can obtain result as shown in table 7.

Software reliability growth result contrast under the different abnormalization Operation Profile of table 7

	Selected model	The crash rate assessed value	Reject number of defects	Total testing time (h)
					Abnormalization of the first Operation Profile	GO	0.00000281	8	10.12
Abnormalization of the second Operation Profile	GO	0.00000172	8	39.62

As seen, when utilizing abnormalization Operation Profile to make crash rate reach 10 ^-6During the order of magnitude, need to test 10.12h under abnormalization of the first Operation Profile of utilizing method 1 to obtain, and need to test 39.62h under abnormalization of the second Operation Profile of utilizing method 2 to obtain.And the above results is compared with the test result under traditional Operation Profile.Utilize traditional operation section generating test use case, the fail data that the operation test case obtains is as shown in table 8.

The fail data of table 8 normal operations section

The defect sequence number	Time between failures (s)
		1	1231.2
2	9705.6
		3	13960.8
4	3960
		5	128181.6
6	305496
		7	97477.2
8	120434.4

Utilize the fail data of table 8 by the reliability assessment instrument, select suitable model to carry out the assessment of dependability parameter, can obtain result as shown in table 9.

Reliability growth test result under table 9 traditional operation section

	Model	Crash rate	Reject number of defects	Total testing time (h)
					The traditional operation section	GO	0.00000405	8	189.01

As seen, when utilizing the traditional operation section to make crash rate reach 10 ^-6During the order of magnitude, needing test 189.01h, is to utilize 18.68 times of required test duration under abnormalization of the first Operation Profile of the present invention, is to utilize 4.77 times of required test duration under abnormalization of the second Operation Profile of the present invention.As seen, the time that reaches same test effect was far longer than based on the required time of abnormalization Operation Profile reliability accelerated test of the present invention.

Embodiment 2: followingly test software reliability provided by the invention checking accelerated test method and appraisal procedure are illustrated by certain software test again., for simplified illustration, still adopt above mentioned software S.

Utilize equally traditional operation profile construction method, abnormalization Operation Profile construction method 1, method 2 to construct respectively traditional operation section, abnormalization of the first Operation Profile, abnormalization of the second Operation Profile, the Operation Profile of constructing is as shown in table 3.

Utilize respectively no-failure validation test scheme out-of-service time MTTF θ ₁=27.78, consumer's risk β=0.1 and no-failure validation test scheme θ ₁The software reliability validation test is carried out in=2.78 β=0.1.According to the present invention is directed to the accelerated test of the reliability demonstration based on abnormalization Operation Profile and the appraisal procedure that continuous type software provides, for no-failure validation test scheme θ ₁=27.78 β=0.1, the tradition reliability demonstration required time of test is 63.96h, be 0.036601 according to the speedup factor of abnormalization of the first Operation Profile, the speedup factor of abnormalization of the second Operation Profile is 0.091144, utilizing formula (31) to calculate separately the required validation test time is 2.34h and 5.83h, namely follow-on test 2.34h did not lose efficacy and just can obtain receiving the conclusion of this software under abnormalization of the first Operation Profile, otherwise rejected this software; In like manner, follow-on test 5.83h did not lose efficacy and just can obtain receiving the conclusion of this software under abnormalization of the second Operation Profile, otherwise rejected this software.

After the validation test scheme is determined, obtain test case according to separately Operation Profile by random sampling respectively, utilize the test case operating software to obtain fail data, that provide as table 10 is no-failure validation test scheme θ ₁Test result under=27.78 β=0.1, as seen within the validation test time that requires, all finding inefficacy under abnormalization of the first Operation Profile He under abnormalization of the second Operation Profile, therefore conclusion, for rejecting this software, in like manner, can be found out according to table 10, also found inefficacy under the traditional operation section, therefore conclusion is for rejecting this software, and namely the test result under the different operating section is identical, is rejection; According to formula (41), can obtain the dependability parameter estimated value under section separately, the estimated value of the crash rate that herein obtains is as shown in table 10, and the reliability estimated value under visible abnormalization Operation Profile and the error of the dependability parameter estimated value under the traditional operation section are very little.

Table 10 no-failure validation test scheme θ ₁=27.78 β=0.1 conclusion is the test data contrast of rejection situation

In like manner, obtain as shown in table 11 for no-failure validation test scheme θ ₁The test result of=2.78 β=0.1 o'clock, the tradition reliability demonstration required time of test is 6.4h, be 0.036601 according to the speedup factor of abnormalization of the first Operation Profile, the speedup factor of abnormalization of the second Operation Profile is 0.091144, utilizing formula (31) to calculate separately the required validation test time is 0.23h and 0.58h, namely follow-on test 0.23h does not occur to lose efficacy and just can obtain receiving conclusion under abnormalization of the first Operation Profile, otherwise rejection; Follow-on test 0.58h does not occur to lose efficacy and just can obtain receiving conclusion under abnormalization of the second Operation Profile, otherwise rejects.

After the validation test scheme is determined, obtain test case according to separately Operation Profile by random sampling respectively, utilize the test case operating software to obtain fail data, that provide as table 11 is no-failure validation test scheme θ ₁Test result under=2.78 β=0.1, as seen within the validation test time that requires, all do not finding to lose efficacy under abnormalization of the first Operation Profile and under abnormalization of the second Operation Profile, therefore conclusion, for receiving, in like manner, can be found out according to table 11, do not find to lose efficacy under the traditional operation section yet, therefore conclusion is for receiving, and namely the test result under the different operating section is identical, is reception; According to formula (43), can obtain the dependability parameter estimated value under section separately, the estimated value of the crash rate that herein obtains is as shown in table 11, and it is very little that the reliability estimated value under visible abnormalization Operation Profile and the dependability parameter estimated value under the traditional operation section obtain error.

Table 11 no-failure validation test scheme θ ₁=2.78 β=0.1 conclusion is the test data contrast of reception condition

Operation Profile	The test duration (h) that needs	Conclusion	The estimated value of crash rate (/s)	Error
					The traditional operation section	6.4	Receive	0.00009994	0
Abnormalization of the first Operation Profile	0.23	Receive	0.0001018	1.9％
					Abnormalization of the second Operation Profile	0.58	Receive	0.0001005	0.6％

Can find out in the example of the accelerated test of the software reliability growth based on abnormalization Operation Profile that provides from above and software reliability checking accelerated test, reliability accelerated test method based on abnormalization Operation Profile utilizes the test duration still less to reach same software reliability growth or validation test target really, namely realized the raising of testing efficiency, and guaranteed the relative stability of dependability parameter estimated result on confidence degree.

The present invention has realized that the nothing of the dependability parameter between abnormalization Operation Profile and traditional Musa Operation Profile estimates partially, and flow process, manner of execution and software reliability estimation method according to the traditional software reliability testing, analogy has proposed flow process, manner of execution and the software reliability estimation method of reliability accelerated test, make between software reliability accelerated test method and traditional method for testing reliability and set up and contact, convenient control to test process.Wherein the test of the operation in reliability accelerated test is Black-box Testing, therefore accelerated test method provided by the invention is still a kind of Black-box Testing method with traditional method for testing reliability 1 sample, does not need to increase the understanding to needed information of white-box testing such as software inhouse information.

It should be noted that finally: above embodiment is only in order to illustrate the present invention and unrestricted technical scheme described in the invention; Therefore, although this instructions has been described in detail the present invention with reference to each above-mentioned embodiment,, will be understood by those skilled in the art that, still can modify or be equal to replacement the present invention; And all do not break away from technical scheme and the improvement thereof of the spirit and scope of the present invention, and it all should be encompassed in the middle of claim scope of the present invention.

Claims (7)

1. the construction method of an abnormalization Operation Profile, is characterized in that, described abnormalization Operation Profile has two kinds of implementation methods:

(1) method 1: at first determine normal operations and the key operation probability of happening in abnormalization Operation Profile, specifically: the probability of happening of normal operations is set to 0, and the probability of happening of key operation obtains according to following formula:

${\mathrm{pc}}_i^{\′}=\frac{{\mathrm{pc}}_i}{{\mathrm{pc}}_1+{\mathrm{pc}}_2+\·\·\·+{\mathrm{pc}}_k}---\left(1\right)$

Wherein, pc _iRepresent the probability of happening of i key operation on the traditional operation section, k represents the number of key operation, Represent the probability of happening of i key operation on abnormalization Operation Profile, the probability of happening sum pc with each key operation on the traditional operation section ₁+ pc ₂+ ... + pc _kAs whole abnormalization Operation Profile speedup factor, the speedup factor of each key operation is identical with the speedup factor of whole abnormalization Operation Profile; Described key operation need to refer to the operation of stress test, and described normal operations do not need to refer to the operation of stress test;

(2) method 2: utilize operation probability of happening value to carry out the size upset to the probability of happening value of all operations, i probability of happening that operates on abnormalization Operation Profile is:

$p_i^{\′}=\frac{p_{\max }-p_i}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(p_{\max }-p_i)}---\left(2\right)$

The speedup factor Λ of whole abnormalization Operation Profile is:

$\mathrm{\Λ}=\frac{1}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{p_i^{\′2}}{p_i}}---\left(3\right)$

The speedup factor Λ of i key operation _iFor:

${\mathrm{\Λ}}_i=p_i\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(p_{\max }-p_i)}{p_{\max }-p_i}---\left(4\right)$

Wherein, n represents the total number that operates, p _maxRepresent in the traditional operation section probability of happening maximum among all operations, p _iRepresent i probability of happening that operates on the traditional operation section,

Represent i probability of happening that operates on abnormalization Operation Profile.

2. software reliability growth accelerated test and the appraisal procedure of the abnormalization Operation Profile that builds based on method claimed in claim 1, is characterized in that, specifically by following steps, realizes:

Step 1, abnormalization of structure Operation Profile, determine the speedup factor of each operation and the speedup factor of whole abnormalization Operation Profile;

Step 2, according to the abnormalization Operation Profile of constructing, carry out the random sampling generating test use case;

Step 3, judgement tested software are discrete type software or continuous type software, if discrete type software performs step four, if continuous type software performs step five;

Step 4, structure STE, the test case that execution step two generates, collect inefficacy interval number of cases, then determine to cause the corresponding operation of input of inefficacy, and the speedup factor of utilizing this operation lost efficacy with the conversion processing of number of cases, described conversion processing is specifically: with the speedup factor of number of cases divided by respective operations, the data that obtain are exactly the data after conversion processing with the inefficacy interval of collecting; This step rear execution step six that is finished;

Step 5, structure STE, the test case that execution step two generates, collect time between failures, then determine to cause the corresponding operation of input of inefficacy, and the speedup factor of utilizing this operation is carried out the conversion processing of time between failures, described conversion processing is specifically: with the speedup factor of the time between failures of collecting divided by respective operations, the data that obtain are exactly the data after conversion processing; This step rear execution step six that is finished;

Step 6, the suitable software reliability growth model of selection, carry out the estimation of dependability parameter;

Whether step 7, judgement tested software this moment meet the reliability requirement of regulation, if, stop test and assessment, finish the method, if not, turn step 2 and carry out.

3. software reliability checking accelerated test and the appraisal procedure of the abnormalization Operation Profile that builds based on method claimed in claim 1, is characterized in that, specifically comprises the steps:

Step 1, abnormalization of structure Operation Profile, and the speedup factor of definite the entire profile;

Step 2, determine the test volume of tested software;

Step 3, the test volume of determining according to step 2, and, based on the abnormalization Operation Profile that step 1 generates, generate the reliability testing use-case, carry out the test case that generates, collect fail message;

Step 4, in conjunction with the fail message of validation test scheme and collection, obtain receiving software or reject the conclusion of software;

Step 5, according to the experimental result of reality, provide the assessment result of current software reliability parameter.

4. software reliability checking accelerated test and the appraisal procedure of the abnormalization Operation Profile based on method structure claimed in claim 1 according to claim 3, is characterized in that, the test volume described in step 2, specifically:

(1) tested software is discrete type software, for fixed number ending testing scheme and no-failure validation test scheme, determines in advance the required test case amount of checking accelerated test,, for no-failure validation test scheme, determines required test case amount

Wherein, [.] expression is right. round F _UThe higher limit of expression failure probability, α represents level of significance;

(2) tested software is continuous type software, testing scheme and no-failure validation test scheme are determined the checking required continuous execution time of accelerated test in advance for regularly ending up, for no-failure validation test scheme, determine required continuous execution time t '=t Λ, wherein, the t representative needs the test duration of carrying out under the traditional operation section, Λ represents the speedup factor of whole abnormalization Operation Profile.

5. software reliability checking accelerated test and the appraisal procedure of the abnormalization Operation Profile based on method structure claimed in claim 1 according to claim 3, is characterized in that, the assessment result of the software reliability parameter described in step 5, specifically:

(1) for discrete type software, assessment result refers to failure probability or success ratio, and when employing method 1 did not comprise the constructed abnormalization Operation Profile of normal operations, the failure probability point estimate F ' of tested software was:

$F^{\text{'}}=\mathrm{\Λ}\frac{n^{\′}}{N^{\′}}---\left(5\right)$

Wherein, Λ represents the speedup factor of whole abnormalization Operation Profile, and the total number that lost efficacy occurs in test case under abnormalization Operation Profile n ', and N ' is test case number total under abnormalization Operation Profile; The success ratio of tested software is p=1-F ';

It is as follows that correspondence obtains the interval estimation of failure probability:

The confidence level of F ' is the confidence lower limit of 1-α

And confidence upper limit

Be respectively:

$F_L^{\′}=\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_1,v_2)}---\left(6\right)$

In formula (6): v ₁=2 (N '-n '+1), v ₂=2n ';

$F_U^{\′}=\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_2,v_1)}---\left(7\right)$

In formula (7): v ₁=2 (N '-n '), v ₂=2 (n '+1);

Be the confidence lower limit p of 1-α by the succeed confidence level of rate p of formula (6) and formula (7) _LWith confidence upper limit p _UBe respectively:

$p_L=1-\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_2,v_1)}---\left(8\right)$

In formula (8): v ₁=2 (N '-n '), v ₂=2 (n '+1),

$p_U=1-\mathrm{\Λ}\frac{v_2}{v_2+v_1{F}_{1-\mathrm{\α}}(v_1,v_2)}---\left(9\right)$

In formula (9): v ₁=2 (N '-n '+1), v ₂=2n ';

F in formula (6)-Shi (9) _1-α(v ₂, v ₁) expression F Distribution Value;

When employing method 2 comprised the constructed abnormalization Operation Profile of all operations, the failure probability point estimate F ' of tested software was:

$F^{\′}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{n_i^{\′}}{N_i^{\′}}{p}_i---\left(10\right)$

Wherein, p _iFor operating o in the traditional operation section _iProbability of happening, n represents the total number that operates,

For falling into operation o _iIn test case in the use-case number that occur to lose efficacy,

For falling into operation o _iIn the total number of test case;

(2) for continuous type software, assessment result refers to crash rate, when employing method 1 does not comprise the constructed abnormalization Operation Profile of normal operations, and the crash rate λ of tested software _uppFor:

${\mathrm{\λ}}_{\mathrm{upp}}=\mathrm{\Λ}{\mathrm{\λ}}_{\mathrm{upp}}^{\′}---\left(11\right)$

Wherein, ${\mathrm{\λ}}_{\mathrm{upp}}^{\′}=\frac{{\mathrm{\χ}}_{1-\mathrm{\α};2n^{\′}+2}^2}{2t^{\′}}$ α represents level of significance,

That card side distributes, $P({\mathrm{\&chi;}}_{2n^{\&prime;}+2}<{\mathrm{\&chi;}}_{1-\mathrm{\&alpha;};2n^{\&prime;}+2}^2)=1-\mathrm{\&alpha;}$ Be used for assessing the upper limit of crash rate, t ' is the test duration on abnormalization Operation Profile, and the total number that lost efficacy occurs in test case under abnormalization Operation Profile n '; When employing method 2 comprised the constructed abnormalization Operation Profile of all operations, the crash rate of tested software was 2 λ _upp

6. a kind of based on the software reliability accelerated test of abnormalization Operation Profile and the computer-aided tool of assessment according to above-mentioned arbitrary described claim, it is characterized in that, comprise traditional operation profile construction module, abnormalization Operation Profile constructing module, test case automatically-generating module, reliability demonstration accelerated test module and reliability growth accelerated test module;

Described traditional operation profile construction module receives user's input operation pattern information and operation information, set up the traditional operation section, and the Operation Profile information data of setting up is comprised evolutionary operator probability, types of variables and span, send abnormalization Operation Profile constructing module to;

Described abnormalization Operation Profile constructing module is according to the selection of user to the building method of abnormalization Operation Profile, generate abnormalization of the second Operation Profile that does not comprise abnormalization of the first Operation Profile of normal operations or comprise all operations, and preserve the speedup factor of each operation and the speedup factor of whole abnormalization Operation Profile;

The information data of the abnormalization Operation Profile that described abnormalization of test case automatically-generating module reception Operation Profile constructing module is come comprises evolutionary operator probability, types of variables and the span on abnormalization Operation Profile, test volume and validation test scheme generating test use case in conjunction with the reliability demonstration accelerated test module input that receives, the growth testing scheme generating test use case of perhaps in conjunction with the reliability growth accelerated test module that receives, inputting, and test case is exported to software under testing;

described reliability demonstration accelerated test module, the information data that receives the abnormalization Operation Profile that abnormalization Operation Profile constructing module comes comprises the evolutionary operator probability on abnormalization Operation Profile, types of variables, span and speedup factor, be achieved as follows concrete function: according to the validation test index of user's input, be at first tested discrete type software or continuous type Software Create validation test scheme, and definite test volume, the test volume and the validation test scheme that generate are exported to the test case automatically-generating module, then receive the user and move the collected fail data of testing software implementation of test cases in constructed STE, in conjunction with the validation test scheme, determine to receive or reject this tested software, and obtain the assessment result of current software reliability parameter according to experimental result,

described reliability growth accelerated test module, the information data that receives the abnormalization Operation Profile that abnormalization Operation Profile constructing module comes comprises the evolutionary operator probability on abnormalization Operation Profile, types of variables, span and speedup factor, be achieved as follows concrete function: according to the growth test index of user's input, be at first that tested discrete type software or continuous type Software Create increases testing scheme, to increase testing scheme and export to the test case automatically-generating module, receive the user and move the collected fail data of testing software implementation of test cases in constructed STE, determine to cause the corresponding operation of input of inefficacy, utilize the speedup factor of this operation to carry out the conversion processing of fail data, and the data conversion results of test result is exported.

7. according to claim 6 a kind of based on the software reliability accelerated test of abnormalization Operation Profile and the computer-aided tool of assessment, it is characterized in that, described traditional operation profile construction module adopts the building method of Musa Operation Profile to construct the traditional operation section.

Images