Software Error Analysis

This document provides guidance on software error analysis. Software error analysis includes error detection, analysis, and resolution. Error detection techniques considered in the study are those used in software development, software quality assurance, and software verification, validation and testing activities. These techniques are those frequently cited in technical literature and software engineering standards or those representing new approaches to support error detection. The study includes statistical process control techniques and relates them to their use as a software quality assurance technique for both product and process improvement. Finally, the report describes several software reliability models.

Data Collection; Error Detection; Error Removal; High Integrity Software; Metrics; Software Error Analysis; Software Quality Assurance; Software Verification and Validation; Statistical Process Control

The main purpose of this document is to provide the software engineering community with current information regarding error analysis for software, which will assist them to do the following:

• -- Understand how error analysis can aid in improving the software development process;
  
  
• -- Assess the quality of the software, with the aid of error detection techniques;
  
  
• -- Analyze errors for their cause and then fix the errors; and
  
  
• -- Provide guidelines for the evaluation of high-integrity software.

TABLE OF CONTENTS

This document provides guidance on software error analysis. Error analysis includes the activities of detecting errors, of recording errors singly and across projects, and of analyzing single errors and error data collectively. The purpose of error analysis is to provide assurance of the quality of high integrity software.

The software industry is currently still young, without sufficient knowledge and adequate standards to guarantee fault-free software. Although research continues to identify better processes for error prevention, with current practices, errors will likely be entered into the software some time during development and maintenance. Hence, there is the need for error analysis, to aid in detecting, analyzing, and removing the errors.

The main purpose of this study is to provide the software engineering community with current information regarding error analysis, which will assist them to do the following:

• -- Understand how error analysis can aid in improving the software development process;
  
  
• -- Assess the quality of software, with the aid of error detection techniques;
  
  
• -- Analyze errors for their cause and then fix the errors; and
  
  
• -- Provide guidelines for the evaluation of high integrity software.

Section 2 discusses how error detection and analysis techniques can be used to improve the quality of software. Section 3 provides a global description of the principal detection techniques used in each software lifecycle phase and cost benefits for selected categories of these techniques. Section 4 provides guidance on collecting individual error data and removing single errors. Section 5 describes techniques for the collection and analysis of sets of error data, including statistical process control techniques and software reliability models. Section 6 provides a summary and recommendations based on this study of error analysis, and Section 7 provides a list of references. Appendix A contains detailed descriptions of common error detection techniques. Appendix B contains the results of a study of standards for high integrity software to determine the extent of coverage of error analysis techniques.

The error detection techniques and statistical techniques described in this report are a representative sampling of the most widely-used techniques and those most frequently referenced in standards, guidelines and technical literature. This report also describes the more common software reliability estimation models, most which are described in the American Institute of Aeronautics and Astronautics (AIAA) draft handbook for software reliability [AIAA]. Inclusion of any technique in this report does not indicate endorsement by the National Institute of Standards and Technology (NIST).

Definitions of the following key terms used in this report are based on those in [IEEEGLOSS], [JURAN], [FLORAC], [SQE], [SHOOMAN], and [NIST204]. However, this report does not attempt to differentiate between "defect," "error," and "fault," since use of these terms within the software community varies (even among standards addressing these terms). Rather, this report uses those terms in a way which is consistent with the definitions given below, and with other references from which information was extracted.

anomaly. Any condition which departs from the expected. This expectation can come from documentation (e.g., requirements specifications, design documents, user documents) or from perceptions or experiences.
Note: An anomaly is not necessarily a problem in the software, but a deviation from the expected, so that errors, defects, faults, and failures are considered anomalies.

computed measure. A measure that is calculated from primitive measures.

defect. Any state of unfitness for use, or nonconformance to specification.

error. (1) The difference between a computed, observed, or measured value and the true, specified, or theoretically correct value or condition. (2) An incorrect step, process, or data definition. Often called a bug. (3) An incorrect result. (4) A human action that produces an incorrect result.
Note: One distinction assigns definition (1) to error, definition (2) to fault, definition (3) to failure, and definition (4) to mistake.

error analysis. The use of techniques to detect errors, to estimate/predict the number of errors, and to analyze error data both singly and collectively.

fault. An incorrect step, process, or data definition in a computer program. See also: error.

failure. Discrepancy between the external results of a program's operation and the software product requirements. A software failure is evidence of the existence of a fault in the software.

high integrity software. Software that must and can be trusted to work dependably in some critical function, and whose failure to do so may have catastrophic results, such as serious injury, loss of life or property, business failure or breach of security. Examples: nuclear safety systems, medical devices, electronic banking, air traffic control, automated manufacturing, and military systems.

primitive measure. A measure obtained by direct observation, often through a simple count (e.g., number of errors in a module).

primitive metric. A metric whose value is directly measurable or countable.

measure. The numerical value obtained by either direct or indirect measurement; may also be the input, output, or value of a metric.

metric. The definition, algorithm or mathematical function used to make a quantitative assessment of product or process.

problem. Often used interchangeably with anomaly, although problem has a more negative connotation, and implies that an error, fault, failure or defect does exist.

process. Any specific combination of machines, tools, methods, materials and/or people employed to attain specific qualities in a product or service.

reliability (of software). The probability that a given software system operates for some time period, without system failure due to a software fault, on the machine for which it was designed, given that it is used within design limits.

statistical process control. The application of statistical techniques for measuring, analyzing, and controlling the variation in processes.

Software error analysis includes the techniques used to locate, analyze, and estimate errors and data relating to errors. It includes the use of error detection techniques, analysis of single errors, data collection, metrics, statistical process control techniques, error prediction models, and reliability models.

Error detection techniques are techniques of software development, software quality assurance (SQA), software verification, validation and testing used to locate anomalies in software products. Once an anomaly is detected, analysis is performed to determine if the anomaly is an actual error, and if so, to identify precisely the nature and cause of the error so that it can be properly resolved. Often, emphasis is placed only on resolving the single error. However, the single error could be representative of other similar errors which originated from the same incorrect assumptions, or it could indicate the presence of serious problems in the development process. Correcting only the single error and not addressing underlying problems may cause further complications later in the lifecycle.

Thorough error analysis includes the collection of error data, which enables the use of metrics and statistical process control (SPC) techniques. Metrics are used to assess a product or process directly, while SPC techniques are used to locate major problems in the development process and product by observing trends. Error data can be collected over the entire project and stored in an organizational database, for use with the current project or future projects. As an example, SPC techniques may reveal that a large number of errors are related to design, and after further investigation, it is discovered that many designers are making similar errors. It may then be concluded that the design methodology is inappropriate for the particular application, or that designers have not been adequately trained. Proper adjustments can then be made to the development process, which are beneficial not only to the current project, but to future projects.

The collection of error data also supports the use of reliability models to estimate the probability that a system will operate without failures in a specified environment for a given amount of time. A vendor ⁽²⁾ may use software reliability estimation techniques to make changes in the testing process, and a customer may use these techniques in deciding whether to accept a product.

The error data collected by a vendor may be useful to auditors. Auditors could request that vendors submit error data, but with the understanding that confidentiality will be maintained and that recriminations will not be made. Data collected from vendors could be used by the auditors to establish a database, providing a baseline for comparison when performing evaluations of high integrity software. Data from past projects would provide guidance to auditors on what to look for, by identifying common types of errors, or other features related to errors. For example, it could be determined whether the error rates of the project under evaluation are within acceptable bounds, compared with those of past projects.

Ideally, software development processes should be so advanced that no errors will enter a software system during development. Current practices can only help to reduce the number of errors, not prevent all errors. However, even if the best practices were available, it would be risky to assume that no errors enter a system, especially if it is a system requiring high integrity.

The use of error analysis allows for early error detection and correction. When an error made early in the lifecycle goes undetected, problems and costs can accrue rapidly. An incorrectly stated requirement may lead to incorrect assumptions in the design, which in turn cause subsequent errors in the code. It may be difficult to catch all errors during testing, since exhaustive testing, which is testing of the software under all circumstances with all possible input sets, is not possible [MYERS]. Therefore, even a critical error may remain undetected and be delivered along with the final product. This undetected error may subsequently cause a system failure, which results in costs not only to fix the error, but also for the system failure itself (e.g., plant shutdown, loss of life).

Sometimes the cost of fixing an error may affect a decision not to fix an error. This is particularly true if the error is found late in the lifecycle. For example, when an error has caused a failure during system test and the location of the error is found to be in the requirements or design, correcting that error can be expensive. Sometimes the error is allowed to remain and the fix deferred until the next version of the software. Persons responsible for these decisions may justify them simply on the basis of cost or on an analysis which shows that the error, even when exposed, will not cause a critical failure. Decision makers must have confidence in the analyses used to identify the impact of the error, especially for software used in high integrity systems.

A strategy for avoiding the high costs of fixing errors late in the lifecycle is to prevent the situation from occurring altogether, by detecting and correcting errors as early as possible. Studies have shown that it is much more expensive to correct software requirements deficiencies late in the development effort than it is to have correct requirements from the beginning [STSC]. In fact, the cost to correct a defect found late in the lifecycle may be more than one hundred times the cost to detect and correct the problem when the defect was born [DEMMY]. In addition to the lower cost of fixing individual errors, another cost benefit of performing error analysis early in development is that the error propagation rate will be lower, resulting in fewer errors to correct in later phases. Thus, while error analysis at all phases is important, there is no better time, in terms of cost benefit, to conduct error analysis than during the software requirements phase.

Planning for error analysis should be part of the process of planning the software system, along with system hazard analysis ⁽³⁾ and software criticality analysis. System hazard analysis is used to identify potential events and circumstances that might lead to problems of varying degrees of severity, from critical failures resulting in loss of life, to less serious malfunctions in the system. Software hazard analysis focuses on the role of the software relative to the hazards. Software criticality analysis may use the results of system and software hazard analyses to identify the software requirements (or design and code elements) whose erroneous implementation would cause the most severe consequences. Criticality analysis may also be used to identify project requirements that are essential to achieving a working software system. Critical software requirements are traced through the development process, so that developers can identify the software elements which are most error-prone, and whose errors would be catastrophic.

The results of hazard analysis and criticality analysis can be used to build an effective error analysis strategy. They aid in choosing the most appropriate techniques to detect errors during the lifecycle (see sec. 3). They also aid in the planning of the error removal process (i.e., the removal of individual errors, as described in sec. 4). Lastly, they aid in the selection of metrics, statistical process control techniques, and software reliability estimation techniques, which are described in section 5. Error analysis efforts and resources can be concentrated in critical program areas. Error analysis techniques should be chosen according to which type of errors they are best at locating. The selection of techniques should take into account the error profile and the characteristics of the development methodology. No project can afford to apply every technique, and no technique guarantees that every error will be caught. Instead, the most appropriate combination of techniques should be chosen to enable detection of as many errors as possible in the earlier phases.

Software development and maintenance involves many processes resulting in a variety of products collectively essential to the operational software. These products include the statement of the software requirements, software design descriptions, code (source, object), test documentation, user manuals, project plans, documentation of software quality assurance activities, installation manuals, and maintenance manuals. These products will probably contain at least some errors. The techniques described in this section can help to detect these errors. While not all products are necessarily delivered to the customer or provided to a regulatory agency for review, the customer or regulatory agency should have assurance that the products contain no errors, contain no more than an agreed upon level of estimated errors, or contain no errors of a certain type.

This section of the report identifies classes of error detection techniques, provides brief descriptions of these techniques for each phase of the lifecycle, and discusses the benefits for certain categories of these techniques. Detailed descriptions of selected techniques appear in Appendix A. Detailed checklists provided in [NISTIR] identify typical problems that error detection techniques may uncover.

Error detection techniques may be performed by any organization responsible for developing and assuring the quality of the product. In this report, the term "developer" is used to refer to developers, maintainers, software quality assurance personnel, independent software verification and validation personnel, or others who perform error detection techniques.

Error detection techniques generally fall into three main categories of analytic activities: static analysis, dynamic analysis, and formal analysis. Static analysis is "the analysis of requirements, design, code, or other items either manually or automatically, without executing the subject of the analysis to determine its lexical and syntactic properties as opposed to its behavioral properties" [CLARK]. This type of technique is used to examine items at all phases of development. Examples of static analysis techniques include inspections, reviews, code reading, algorithm analysis, and tracing. Other examples include graphical techniques such as control flow analysis, and finite state machines, which are often used with automated tools. Traditionally, static analysis techniques are applied to the software requirements, design, and code, but they may also be applied to test documentation, particularly test cases, to verify traceability to the software requirements and adequacy with respect to test requirements [WALLACE].

Dynamic analysis techniques involve the execution of a product and analysis of its response to sets of input data to determine its validity and to detect errors. The behavioral properties of the program are also observed. The most common type of dynamic analysis technique is testing. Testing of software is usually conducted on individual components (e.g., subroutines, modules) as they are developed, on software subsystems when they are integrated with one another or with other system components, and on the complete system. Another type of testing is acceptance testing, often conducted at the customer's site, but before the product is accepted by the customer. Other examples of dynamic analyses include simulation, sizing and timing analysis, and prototyping, which may be applied throughout the lifecycle.

Formal methods involve rigorous mathematical techniques to specify or analyze the software requirements specification, design, or code. Formal methods can be used as an error detection technique. One method is to write the software requirements in a formal specification language (e.g., VDM, Z), and then verify the requirements using a formal verification (analysis) technique, such as proof of correctness. Another method is to use a formal requirements specification language and then execute the specification with an automated tool. This animation of the specification provides the opportunity to examine the potential behavior of a system without completely developing a system first.

Criteria for selection of techniques for this report include the amount of information available on them, their citation in standards and guidelines, and their recent appearance in research articles and technical conferences. Other techniques exist, but are not included in this report. Table 3-1 provide a mapping of the error detection techniques described in Appendix A to software lifecycle phases. In these tables, the headings R, D, I, T, IC, and OM represent the requirements, design, implementation, test, installation and checkout, and operation and maintenance phases, respectively. The techniques and metrics described in this report are applicable to the products and processes of these phases, regardless of the lifecycle model actually implemented (e.g., waterfall, spiral). Table B-2 in Appendix B lists which high integrity standards cite these error detection techniques.

During the requirements phase, static analysis techniques can be used to check adherence to specification conventions, consistency, completeness, and language syntax. Commonly used static analysis techniques during the requirements phase include control flow analysis, data flow analysis, algorithm analysis, traceability analysis, and interface analysis. Control and data flow analysis are most applicable for real time and data driven systems. These flow analyses employ transformation of text describing logic and data requirements into graphic flows which are easier to examine. Examples of control flow diagrams include state transition and transaction diagrams. Algorithm analysis involves rederivation of equations or the evaluation of the suitability of specific numerical techniques. Traceability analysis involves tracing the requirements in the software requirements specification to system requirements. The identified relationships are then analyzed for correctness, consistency, completeness, and accuracy. Interface analysis in this phase involves evaluating the software requirements specification with the hardware, user, operator, and software interface requirements for correctness, consistency, completeness, accuracy, and readability.

Dynamic analysis techniques can be used to examine information flows, functional interrelationships, and performance requirements. Simulation is used to evaluate the interactions of large, complex systems with many hardware, user, and other interfacing software components. Prototyping helps customers and developers to examine the probable results of implementing software requirements. Examination of a prototype may help to identify incomplete or incorrect requirements statements and may also reveal that the software requirements will not result in system behavior the customer wants. Prototyping is usually worthwhile when the functions of the computer system have not previously been used in automated form by the customer. In this case, the customer can change the requirements before costly implementation. Unless the project is small or an automated method can be used to build a prototype quickly, usually only carefully selected functions are studied by prototyping.

One approach for analyzing individual requirements is requirements parsing. This manual technique involves examination to ensure that each requirement is defined unambiguously by a complete set of attributes (e.g., initiator of an action, source of the action, the action, the object of the action, constraints). Because this technique identifies undefined attributes, it may prevent release of incomplete requirements to the designers. In those cases where the requirements are to be represented by a formal language specification, this analysis aids in clarifying a requirement before its transformation.

Languages based on formal methods, i.e., mathematically based languages, may be used to specify system requirements. The act of specifying the software requirements in a formal language forces reasoning about the requirements and becomes an error detection technique. When requirements have been written in a formal language, the task of simulation may be easier. Then, the behavior of the potential system can be observed through use of the simulation. It may be the combination of formal specifications with other error detection techniques (e.g., control flow analysis and data flow analysis) that provides the biggest payoff for using formal methods.

Evaluation of the design provides assurance that the requirements are not misrepresented, omitted, or incompletely implemented, and that unwanted features are not designed into the product by oversight. Design errors can be introduced by implementation constraints relating to timing, data structures, memory space, and accuracy.

Static analysis techniques help to detect inconsistencies, such as those between the inputs and outputs specified for a high level module and the inputs and outputs of the submodules. The most commonly used static analysis techniques during this phase include algorithm analysis, database analysis, (design) interface analysis, and traceability analysis. As in the requirements phase, algorithm analysis examines the correctness of the equations and numerical techniques, but in addition, it examines truncation and rounding effects, numerical precision of word storage and variables (single vs. extended-precision arithmetic), and data typinginfluences. Database analysis is particularly useful for programs that store program logic in data parameters. Database analysis supports verification of the computer security requirement of confidentiality, by checking carefully the direct and indirect accesses to data. Interface analysis aids in evaluating the software design documentation with hardware, operator, and software interface requirements for correctness, consistency, completeness, and accuracy. Data items should be analyzed at each interface. Traceability analysis involves tracing the software design documentation to the software requirements documentation and vice versa.

Commonly used dynamic analysis techniques for this phase include sizing and timing analysis, prototyping, and simulation. Sizing and timing analysis is useful in analyzing real-time programs with response time requirements and constrained memory and execution space requirements. This type of analysis is especially useful for determining that allocations for hardware and software are made appropriately for the design architecture; it would be quite costly to learn in system test that the performance problems are caused by the basic system design. An automated simulation may be appropriate for larger designs. Prototyping can be used as an aid in examining the design architecture in general or a specific set of functions. For large complicated systems prototyping can prevent inappropriate designs from resulting in costly, wasted implementations.

Formal analysis involves tracing paths through the design specification and formulating a composite function for each, in order to compare these composite functions to that of the previous level. This process ensures that the design continues to specify the same functional solution as is hierarchically elaborated. This process can be applied manually, if the specification is sufficiently formal and exact, but is most feasible only for high level design specifications. However, with automated tools, the functional effects of all levels of the design can be determined, due to the speed and capacity of the tools for manipulating detailed specifications.

Use of static analysis techniques helps to ensure that the implementation phase products (e.g., code and related documentation) are of the proper form. Static analysis involves checking that the products adhere to coding and documentation standards or conventions, and that interfaces and data types are correct. This analysis can be performed either manually or with automated tools.

Frequently used static analysis techniques during this phase include code reading, inspections, walkthroughs, reviews, control flow analysis, database analysis, interface analysis, and traceability analysis. Code reading involves the examination by an individual, usually an expert other than the author, for obvious errors. Inspections, walkthroughs, and reviews, which are all used to detect logic and syntax errors, are effective forerunners to testing. As in previous phases, control flow diagrams are used to show the hierarchy of main routines and their subfunctions. Database analysis is performed on programs with significant data storage to ensure that common data and variable regions are used consistently between all calling routines; that data integrity is enforced and no data or variable can be accidentally overwritten by overflowing data tables; and that data typing and use are consistent throughout the program. With interface analysis, source code is evaluated with the hardware, operator, and software interface design documentation, as in the design phase. Traceability analysis involves tracing the source code to corresponding design specifications and vice versa.

One category of static analysis techniques performed on code is complexity analysis. Complexity analysis measures the complexity of code based on specific measurements (e.g., number of parameters passed, number of global parameters, number of operands/operators). Although not an error detection technique, complexity analysis can be used as an aid in identifying where use of error detection techniques should be concentrated and also in locating test paths and other pertinent information to aid in test case generation.

Other static analysis techniques used during implementation which aid in error detection include software sneak circuit analysis and slicing. Software sneak circuit analysis is a rigorous, language-independent technique for the detection of anomalous software (i.e., "sneaks") which may cause system malfunction. The methodology involves creation of a comprehensive "pictorial" database using quasi-electrical symbology which is then analyzed using topological and application "clues" to detect faults in the code [PEYTON]. Slicing is a program decomposition technique used to trace an output variable back through the code to identify all code statements relevant to a computation in the program [LYLE]. This technique may be useful to demonstrate functional diversity.

Dynamic analysis techniques help to determine the functional and computational correctness of the code. Regression analysis is used to reevaluate requirements and design issues whenever any significant code change is made. This analysis ensures awareness of the original system requirements. Sizing and timing analysis is performed during incremental code development and analysis results are compared against predicted values.

A formal method used in the implementation phase is proof of correctness, which is applied to code.

Dynamic analysis in the test phase involves different types of testing and test strategies. Traditionally there are four types of testing: unit, integration, system, and acceptance. Unit testing may be either structural or functional testing performed on software units, modules, or subroutines. Structural testing examines the logic of the units and may be used to support requirements for test coverage, that is, how much of the program has been executed. Functional testing evaluates how software requirements have been implemented. For functional testing, testers usually need no information about the design of the program because test cases are based on the software requirements.

Integration testing is conducted when software units are integrated with other software units or with system components. During integration testing, various strategies can be employed (e.g., top-down testing, bottom-up testing, sandwich testing) but may depend on the overall strategy for constructing the system. Integration testing focuses on software, hardware, and operator interfaces.

Both system testing and acceptance testing execute the complete system. The primary difference is that the developer conducts system testing, usually in the development environment, while the customer conducts acceptance testing (or commissions the developer to conduct the acceptance testing in the presence of the customer). Acceptance testing is supposed to occur in a fully operational customer environment, but in some cases (e.g., nuclear power plants, flight control systems), some parts of the environment may need to be simulated.

For all four types of testing, different strategies may be used, according to the project's characteristics. Some strategies include stress testing, boundary value testing, and mutation testing. Operational profile testing allows testers to select input data that are of special interest to the customer. For example, input data that causes execution of the most frequently used functions in operation may be the most important profile for testing for some systems. In other cases, it may be more important to choose an input profile that should not occur in reality. For nuclear power plants, this means choosing a profile that causes the software safety system to react; the system responses can be examined to determine system behavior in adverse circumstances.

A major problem with testing is knowing when to stop. Software reliability estimation techniques, such as those described in section 5 of this report, can be used to estimate the number of errors still present in the system, and to determine how much more testing is needed. Sensitivity analysis, a promising technique emerging from the research community and entering the marketplace, is intended to indicate where to test, and hence to determine how much to test [VOAS]. Because sensitivity analysis is derived from mutation testing which is intended for detecting small changes to a program, this technique depends on code that is already "close to correct" for its effectiveness.

During this phase, it is necessary to validate that the software operates correctly with the operational hardware system and with other software, as specified in the interface specifications. It is also necessary to verify the correctness and adequacy of the installation procedures and certify that the verified and validated software is the same as the executable code approved for installation. There may be several installation sites, each with different parameters. It is necessary to check that the programs have been properly tailored for each site.

The most commonly used dynamic analysis techniques for this phase are regression analysis and test, simulation, and test certification. When any changes to the product are made during this phase, regression analysis is performed to verify that the basic requirements and design assumptions affecting other areas of the program have not been violated. Simulation is used to test operator procedures and to isolate installation problems. Test certification, particularly in critical software systems, is used to verify that the required tests have been executed and that the delivered software product is identical to the product subjected to software verification and validation (V&V).

During operation of an on-line continuous system, test cases may be constructed that will check periodically if the system is behaving as expected. For any software maintenance activity, error detection techniques should be selected as if the maintenance activity were a new development activity, but considering the impact of new changes to the system. Use of traceability analysis on the software products, including test documentation, is crucial to identifying the extent of use of any selected error detection technique on the total software system. Regression testing must be applied in this phase.

In the early days of computers, static analysis of software involved hours of tedious manual checking of code for structural errors, syntax errors, and other types of errors. Today automation handles the tedious bookkeeping in both design and code activities. In the past, manual reading by an individual not only took longer, but may not always have been thorough. Design tools force consistency to some extent and support other static analyses. Techniques such as control flow analysis were difficult to perform manually, but with modern Computer Aided Software Engineering (CASE) tools, most static analyses can be performed more quickly and efficiently. As the power and ease of use of the tools improve, then static analyses become more effective.

A tool commonly used to perform static analysis is the compiler, which can detect syntactical code errors. The direct costs are the amount of electrical power and resources needed to conduct the compilation. However, not everyone agrees on the usefulness of compilers for producing error-free code. Supporters of Cleanroom engineering, a methodology for developing and assuring the quality of software, argue that the costs of rework and recompile are significant and should not be ignored. They believe that complete reliance on the tools to perform some of the intellectual work may reduce quality, because clean compilations can give a false sense of complete correctness [MILLS]. With Cleanroom engineering, programmers do not compile their code. Instead they spend more time on design, using a "box structure" method, and on analyzing their own work. When the programmers are confident of their work, it is submitted to another group who then compiles and tests the code. The Software Engineering Laboratory (SEL) at the National Aeronautics and Space Administration Goddard Space Flight Center collected sufficient data over 16 years to establish baselines for error and productivity rates. In recent years, two experiments were conducted on the Cleanroom approach [GREEN]. Results of the two Cleanroom experiments compared with SEL baselines show a lower error rate in the finished product and an increase in productivity across the lifecycle.

Software inspection, another static technique, is time consuming because it requires line by line, or graph by graph reading of the software element. Data collected from over 203 software inspections at the Jet Propulsion Laboratory in Pasadena, California, showed a significantly higher density of defects during requirements inspections [KELLY]. However, the defect densities of the products decreased exponentially from the requirements phase to the coding phase, implying that testing and rework will take less time. Code reading is another static analysis technique that has been shown in another SEL study to be quite effective [BASILI]. Researchers found that effectiveness increased with the experience level of the code readers, the reason being that experienced personnel were mentally executing the code. This technique may be difficult to schedule and implement formally; usually it is used when a programmer seeks help from a peer, and is conducted on small sections of code at a time. Also, errors found by code reading may not always be handled with a formal anomaly report.

Inspection and code reading have one drawback in common. The maximum benefits for these techniques are achieved when they are performed on all elements of the design and code, which tends to be time-consuming. Because of the time factor, they are usually conducted on only small portions of a program at a time, usually in three- or four-hour meetings. When the objective is to examine the entire program or many modules for global features, then other techniques with specific objectives (e.g., interface consistency, control flow, logic flow analysis) are more appropriate. Many of the static analysis techniques are intended to be conducted by individual members of a team, perhaps over days, and probably with automated support. There may be interaction among the team, especially to discuss possible anomalies. These types of techniques are effective for examining the integration of design or code modules.

Dynamic analyses tend to use large amounts of computer resources and involve human effort to prepare, execute and analyze tests. Testing can never guarantee that a system is completely correct, but it can demonstrate exactly what will occur under specific circumstances. Testing helps to establish an operational profile under which the system will work correctly. Testing also helps to uncover errors that were previously not discovered. Acceptance testing assures a customer that the software system will behave appropriately under specific circumstances as the customer has requested. Some CASE tools provide test aids (e.g., test case generators, test result capture, documentation). Although the power of modern computers has reduced execution time of test cases, nevertheless, exhaustive testing with all possible inputs under all circumstances is still not possible. In order to obtain maximum benefits from testing, careful planning and development of test goals, and strategies to achieve those goals are required.

While some static and dynamic analyses have become easier to perform with CASE tools, CASE technology has not eliminated all problems of software development and assurance. There are the problems of cost, methodology-dependence, and difficulty in understanding and using them. Two other major problems with CASE include restrictions on using the environments when they are not built on the concept of open systems [NIST187] and when information developed by the tools cannot be readily exchanged among tools of different vendors.

While journal articles and other literature describing usage of formal methods are available, they do not provide sufficient information to draw conclusions about the cost/quality benefits of using formal methods for the assurance of the quality of software. A study of formal methods was funded by NIST, the U.S. Naval Research Laboratory, and the Atomic Energy Control Board of Canada, to determine whether the benefits of using formal methods are significant relative to the costs of using them. The results of this study are published in [NISTGCR]. In the United Kingdom, there is an existing standard for safety critical software used in defense equipment, which requires the use of formal languages for specifications [MOD55].

The cost benefits of using specific error detection techniques or classes of techniques will differ from project to project. A balanced error detection program will depend on many factors, including the consequences of failure caused by an undetected error, the complexity of the software system, the types of errors likely to be committed in developing specific software, the effort needed to apply a technique, the automated support available, and the experience of the development and assurance staff. Another factor to consider is the interplay among techniques (e.g., whether the output of one technique can be used readily by another technique). If a specific error type is likely, then a technique known for finding that type of error should be selected. The application of formal verification techniques is appropriate when failure of the software would be disastrous. For planning a balanced program, an important requirement should be to ensure that analyses are applied to all the software products at all phases of the lifecycle in an orderly manner. The program should be evaluated frequently to ensure that the analyses are being used correctly and are aiding in error detection. The SPC techniques described in section 5 aid in this evaluation.

A final consideration for selecting techniques based on their cost benefit takes into account who will be conducting the analysis, and under what circumstances. For auditors, techniques which examine interfaces across the entire program, control flow, and critical operational paths are more appropriate than those involving detailed line by line analysis (e.g., software code inspection). When an anomaly is found, however, the auditors may choose to examine in greater detail the areas suspected of contributing to the anomaly.

This section describes the process of analyzing anomalies and removing errors. This is performed after an anomaly has been discovered using any error detection technique, such as those discussed in section 3. Analysis of an anomaly will not only aid in the removal of errors related to the anomaly, but will also help to detect other similar errors which have not yet manifested themselves. In addition, information obtained from this analysis can provide valuable feedback that may improve subsequent efforts and development processes in future projects.

The handling of an anomaly generally follows three steps: identification, investigation, and resolution. However, exact procedures for dealing with an anomaly will depend on many factors. First, it may be that the anomaly is not actually an error.⁽⁴⁾ For example, the anomaly may be a result of misinterpreting test results. In these situations, an explanation about why the anomaly is not an error should be recorded, and no further action is required. Second, the procedures will depend on the activity used to detect the anomaly. For example, anomalies discovered during walkthroughs and code reading are often fixed immediately, without having to go through the formal error resolution process. During integration testing, all anomaly reports may be collected and then addressed to locate probable cause and recommend fixes. Third, the severity level of the anomaly will determine how soon the error should be fixed. Generally, the more severe the error, the sooner it needs to be fixed.

The general policy for handling anomalies should include rules/regulations concerning the administration of the entire error removal activity (e.g., who must fill out problem reports, where or to whom this information is distributed, how to close out problem reports, who enters the collected information into the error database). These issues are not addressed in this report, because the policy will be specific to an organization.

General project information which supports the error removal process should be maintained. This information may include, but is not limited to, descriptions of the design methodology, the verification plan used in design, the test plan, the configuration control plan, identification of tools used to design and test software (e.g., CASE tools), and the programming language used.

As soon as an anomaly is detected, information about it should be recorded to help identify, analyze, and correct the anomaly. Typically, this information is presented in an anomaly, or problem report. While the formats may differ, reports should include the following types of information.

Locator. Identify the person(s) who discovered the anomaly including name, address, phone number, email address, fax number, and company identification.

Date and Time. Specify the date and time that the anomaly occurred and/or was discovered. Time can be specified by wall clock time, system time, or CPU time. For distributed systems, specify the time zone.

Activity. Identify the activity taking place at the time the anomaly was discovered. These activities include error detection activities employed during the development and release of a product, including static and dynamic analysis, review, inspection, audit, simulation, timing analysis, testing (unit, integration, system, acceptance), compiling/assembling, and walkthrough.

Phase Encountered. Identify the lifecycle phase in which the anomaly was encountered (e.g., requirements, design, implementation, test, installation and checkout, and operation and maintenance). If possible, specify the activity within the phase (e.g., during preliminary design in the design phase).

Operational Environment. Specify the hardware, software, database, test support software, platform, firmware, monitor/terminal, network, and peripherals being used.

Status of Product. Specify the effect of the problem on the product (e.g., unusable, degraded, affected, unaffected).

Repeatability. Determine if the anomaly is a one-time occurrence, intermittent, recurring, or reproducible.

Symptom. Describe the symptoms, which are indications that a problem exists (e.g., inconsistency between the design and the requirements, violation of a constraint imposed by the requirements, operating system crash, program hang-up, input or output problem, incorrect behavior or result, error message, inconsistent logic or behavior, infinite loop, and unexpected abort).

Location of Symptom. The location of the anomaly can be in the actual product (hardware, software, database, or documentation), the test system, the platform, or in any development phase product (e.g., specification, code, database, manuals and guides, plans and procedures, reports, standards/policies). Identify the documentation that was analyzed, or the code that was executed, the tools and documentation used to support the activity. Identify the specific location(s) where the anomaly is evident in documentation, or the test case in which the anomaly occurred.

Severity. Severity is a measure of the disruption an anomaly gives the user when encountered during operation of the product. Severity can be divided into several levels, with the highest level being catastrophic, and the lowest being at the annoyance level. A severity classification system should be tailored to particular systems or class of systems. The number of levels and the corresponding descriptions of the levels may vary. An example of a severity classification is given below:

Level 6 Critical. Major feature not working, system crashes, loss of data

Level 5 Serious. Impairment of critical system functions, no workaround

Level 4 Major. Workaround is difficult

Level 3 Moderate. Workaround is simple

Level 2 Cosmetic. Tolerable, or fix is deferrable

Level 1 User misunderstanding

Level 0 No problem (e.g., testing error)

Following the identification stage, all errors should be investigated to obtain further information on the nature and cause in order to propose solutions for resolution action or corrective action. Information that should be recorded during this stage include the following:

Phase Introduced. Identify the lifecycle phase in which the error was introduced. If possible, specify the activity within the phase (e.g., during preliminary design in the design phase).

Type. This attribute describes the type of error found, e.g., logic error, computational error, interface/timing error, data handling error, data error, documentation error, document quality error (e.g., out of date, inconsistent, incomplete), and enhancement errors (e.g., errors caused by change in requirements, errors caused by a previous fix).

Location of Error. The location of the error may be the same as the location of the symptom. See Location of Symptom in section 4.1.

Cause. Typical causes of an error include human errors (e.g., misunderstanding, omission errors) and inadequate methodology (e.g., incomplete inspection process,inappropriate design methodology).

Units Affected. This attribute identifies the software unit(s) affected by the error and its related fix, e.g., which components, modules, or documents are affected.

Priority. Priority is the degree of importance that is given to the resolution of an error. Based on the priority level, it is determined whether the error should be fixed immediately, eventually, or not at all (e.g., if error becomes obsolete as result of other changes). However, fixes should be performed according to the software configuration management policies of the project. The relationship between the priority scale and the severity scale should be specified by the administrative policy. An example of a priority scale is:

Level 5 Resolve error immediately

Level 4 Error gets high attention

Level 3 Error will be placed in normal queue

Level 2 Use workaround or fix in the interim

Level 1 Will be fixed last

Level 0 Will not be fixed

Error resolution consists of the steps to correct the error. The policy of the project determines if the person who investigates the error will also correct the error. The procedures for distribution and retention of the error information is also identified by the policy. Typically, the recipients of the error information are the project manager, SQA manager, corporate database manager, and the customer. The amount of formalism (e.g., whether the plan needs to be documented) depends on the scope, risk, and size of the project. For small errors in small projects, this scheme may not be necessary.

The proposed procedures for resolution action should be documented in a resolution plan.

Item to Fix. Identify the item to be fixed (e.g., name, ID number, revision), the component within the item, text description of the fix.

Estimated Date or Time. Specify the proposed dates for start and completion of the fix.

Personnel. Identify the manager, engineer, or other members responsible for performing the fix and for follow-up.

The resolution action is the actual fix, i.e., making changes in the product to correct and remove the error. The following information should be provided by the person(s) who perform the resolution action, upon completion.

Date Completed. Specify the date when resolution action (fix) was completed.

Personnel. Identify the person(s) who fixed the error.

Time Length. Specify the number of minutes or hours required for the fix.

Size of Fix. Specify the size of the fix in terms of the number of source lines of code (SLOC) added / modified, or the number of document pages added / modified.

The corrective action stage is optional, because not all anomalies will require individual corrective actions.

Standards, Policies or Procedures. Identify the standards, policies, or procedures to be revised, created, or reinforced.

Other Action. This includes other revisions to the development process (e.g., implementing training, reallocation of people or resources, and improving or enforcing audit activities).

For all errors, there should be a follow-up to verify that the necessary resolution action or corrective action has been performed correctly.

Personnel. Identify the person or organization that performed follow-up of resolution action and/or corrective action.

Verification of Action. Confirm that the "right fix" has been performed, that the error has actually been removed, and that the fix has not adversely affected other areas of the software.

Disposition. Describe the status of the anomaly, whether it is closed (resolution was implemented and verified, or not within scope of project), deferred, merged with another problem, or referred to another project.

Closeout. Identify procedures for retention of the error data.

The data that is collected for the purpose of removing a single error can be used for other purposes. This data can aid in removing all errors similar to the original error. In addition to making improvements in the product, data on single errors can be used to improve the current development process. For instance, if many errors are found to be requirements errors, this may prompt a change to the requirements specification process. Data on single errors can also be used in measurement and statistical process control activities such as those discussed in section 5. For example, the data can be used to calculate measures or it can be used as input to control charts. Finally, individual error data can be entered into an error database, in order to maintain an error history of all projects in the organization.

Techniques for collecting and analyzing sets of error data during the lifecycle aid in understanding, evaluating and improving the development and maintenance process or aid in evaluating or estimating product quality. Software measures provide insights about both process and product. Measures may feed into statistical process control (SPC) techniques; SPC techniques may be used for both process and product evaluation. Software reliability estimation techniques are usually applied only to the product. Most of these techniques operate on error history profiles of error data discovered by error detection techniques.

This section addresses only the error aspects of these techniques. Other information may need to be collected when making major decisions to change a policy or development activity (e.g., project information, customer requirements, company policy, methodologies being used, tools being used, number of people using a certain technique). These types of information are not discussed in this report.

An error history profile is needed to perform error analysis effectively. An organizational database can be used both to track the status of a project and to track the use of error analysis techniques. Data collected for the purpose of resolving single errors (e.g., source, cause, type, severity), should be placed in the error database to enable the establishment of anomaly histories. Other data collected specifically for the purpose of measurement or statistical process control should also be entered into the database. The database serves as a baseline for validation as well as for improvement. Past mistakes can be avoided from lessons learned. Maintaining a database serves the following purposes:

• -- To identify which development processes work well (or poorly) for an application domain,
  
  
• -- To support process improvement for the current project as well as for future projects,
  
  
• -- To identify whether the processes are being applied properly (or at all),
  
  
• -- To identify error classes by cause,
  
  
• -- To estimate the error growth rate throughout development, and therefore to be able to adjust plans for assurance activities, and
  
  
• -- To measure the quality of the software at delivery.

Error data collected from an error detection technique in one phase can support process improvement in an earlier lifecycle phase (for future projects), as well as in a later phase. For example, in a presentation at a COMPASS Conference, ⁽⁵⁾ one panel member explained that analysis of the data collected from code inspections at his organization revealed that a high percentage of the code errors were the result of errors in the software requirements specification. In response to this finding, the organization began investigating the use of formal languages for specifying software requirements. This example demonstrates the necessity of collecting and analyzing data for both error removal and process improvement.

Data histories can help managers to recognize when there is a significant deviation from project plans during development. Past error data can be used to estimate the number of expected errors at different times in the development cycle. For instance, if the reported error count for a particular product was smaller than was expected, compared with similar past projects, this may suggest that the development team generated an unusually low number of errors. However, further investigation may reveal that the project was behind schedule, and to save time, planned inspections were not held. Thus, many existing errors remained undetected, so that the low error count did not reflect the true quality of the product. This example illustrates how a history profile enables an organization to recognize and correct a process problem to avoid delivering a product with residual errors.

This section describes the process of collecting error data for a specific purpose (e.g., to use with control charts). Some of the error data may include data previously collected during error detection. This data can be retrieved from the organizational database, or can be collected directly upon discovery or during resolution of an error. Data must be collected properly in order for any error analysis technique to be effective. The recommended steps of the data collection process are listed below [AIAA]:

1. Establish the objectives.
   
   
2. Prepare a data collection plan. The plan may include the following recommended elements:
   

   Data definition and type. Specify/define the data to be collected and the type (i.e., attribute or variable data). An attribute is a characteristic that an item may or may not possess. It is obtained by noting the presence or absence of a characteristic and counting occurrences of the characteristic with a specified unit. For example: a module may or may not contain a defect. This type of data takes on only discrete (integer) values. Variable data is obtained by recording a numerical value for each item observed. Variable data can be either continuous or discrete. Examples: cost of fixing an error (continuous), lines of code (discrete).
   
   

   Analysis technique. Identify the technique requiring the data. Each technique has unique data requirements, so the technique to be used should be specified prior to data collection.
   
   

   Measurement method. Measurements can be taken by equipment, observation, or selecting data from existing records. The reliability, determined by accuracy and precision, of the measurement method must be established prior to data collection.
   
   

   Sampling Procedure. The data collection interval, amount of data to be collected, and the sampling method should be specified (e.g., random sampling using a random number table). When determining the data collection interval, issues such as process volume, process stability, and cost should be considered.
   
   

   Personnel. Identify persons responsible for specific data collection tasks.
   
   

   Forms for data reporting (e.g., electronic spreadsheet, paper forms, etc.).
   
   

   Recording and processing of data. One method for processing data is blocking, the separating of data into potential comparison categories during the recording of data. Blocking can be accomplished by recording each category separately, or through labeling information that enables future sorting.
   
   

   Monitoring. Describe how the data collection process is to be monitored.

   
   
3. Apply tools. Automated tools should be considered whenever possible, in order to minimize impact on the project's schedule. Factors to consider include the following: availability of the tool, reliability of the tool, cost of purchasing or developing the tool, and whether it can handle any necessary adjustments.
   
   
4. Provide training. Once tools and plans are in place, training should be provided to ensure that data collectors understand the purpose of the measurements and know explicitly what data is to be collected.
   
   
5. Perform trial run. A trial run of the data collection plan should be made to resolve any problems or misconceptions about the plan. This can save vast amounts of time and effort.
   
   
6. Implement the plan. Collect the data and review them promptly, so that problems can be resolved before the disappearance of information required to resolve them (e.g., if test results on a screen are not saved).
   
   
7. Monitor data collection. Monitor the process to ensure that objectives are met and that procedures are implemented according to the data collection plan.
   
   
8. Use the data. Use the data as soon as possible to achieve maximum benefit.
   
   
9. Provide feedback to all involved. Those involved need to know what impact their efforts had, and the end result. This will enable them to understand the purpose of their efforts and agree to undertake similar tasks in the future.

Additional recommendations for the data collection process are listed below [ROOK]:

• -- Data collection should be integrated into the development process (e.g., as part of the quality management system).
  
  
• -- Data collection should be automated whenever possible.
  
  
• -- Time-scales between data collection and data analysis should be minimized.
  
  
• -- Data should be treated as a company resource and facilities should be available to keep historical records of projects as well as to monitor current projects.
  
  
• -- The problem of motivating personnel to keep accurate records should not be underestimated. Proper training and quick analysis facilities are essential, but are not sufficient.

Within the software engineering community, there is much confusion and inconsistency over the use of the terms metric and measure. In this report, a metric is defined to be the mathematical definition, algorithm, or function used to obtain a quantitative assessment of a product or process. The actual numerical value produced by a metric is a measure. Thus, for example, cyclomatic complexity is a metric, but the value of this metric is the cyclomatic complexity measure.

Data on individual errors (see sec. 4) can be used to calculate metrics values. Two general classes of metrics include the following:

management metrics, which assist in the control or management of the development process; and

quality metrics, which are predictors or indicators of the product qualities

Management metrics can be used for controlling any industrial production or manufacturing activity. They are used to assess resources, cost, and task completion. Examples of resource-related metrics include elapsed calendar time, effort, and machine usage. Typical metrics for software estimate task completion include percentage of modules coded, or percentage of statements tested. Other management metrics used in project control include defect-related metrics. Information on the nature and origin of defects are used to estimate costs associated with defect discovery and removal. Defect rates for the current project can be compared to that of past projects to ensure that the current project is behaving as expected.

Quality metrics are used to estimate characteristics or qualities of a software product. Examples of these metrics include complexity metrics, and readability indexes for software documents. The use of these metrics for quality assessment is based on the assumptions that the metric measures some inherent property of the software, and that the inherent property itself influences the behavioral characteristics of the final product.

Some metrics may be both management metrics and quality metrics, i.e., they can be used for both project control and quality assessment. These metrics include simple size metrics (e.g., lines of code, number of function points) and primitive problem, fault, or error metrics. For example, size is used to predict project effort and time scales, but it can also be used as a quality predictor, since larger projects may be more complex and difficult to understand, and thus more error-prone.

A disadvantage of some metrics is that they do not have an interpretation scale which allows for consistent interpretation, as with measuring temperature (in degrees Celsius) or length (in meters). This is particularly true of metrics for software quality characteristics (e.g., maintainability, reliability, usability). Measures must be interpreted relatively, through comparison with plans and expectations, comparison with similar past projects, or comparison with similar components within the current project. While some metrics are mathematically-based, most, including reliability models, have not been proven.

Since there is virtually an infinite number of possible metrics, users must have some criteria for choosing which metrics to apply to their particular projects. Ideally, a metric should possess all of the following characteristics:

• -- Simple - definition and use of the metric is simple
• -- Objective - different people will give identical values; allows for consistency, and prevents individual bias
• -- Easily collected - the cost and effort to obtain the measure is reasonable
• -- Robust - metric is insensitive to irrelevant changes; allows for useful comparison
• -- Valid - metric measures what it is supposed to; this promotes trustworthiness of the measure

Within the software engineering community, two philosophies on measurement are embodied by two major standards organizations. A draft standard on software quality metrics sponsored by the Institute for Electrical and Electronics Engineers Software Engineering Standards Subcommittee supports the single value concept. This concept is that a single numerical value can be computed to indicate the quality of the software; the number is computed by measuring and combining the measures for attributes related to several quality characteristics. The international community, represented by the ISO/IEC organization through its Joint Technical Committee, Subcommittee 7 for software engineering appears to be adopting the view that a range of values, rather than a single number, for representing overall quality is more appropriate.

Metrics enable the estimation of work required in each phase, in terms of the budget and schedule. They also allow for the percentage of work completed to be assessed at any point during the phase, and establish criteria for determining the completion of the phase.

The general approach to using metrics, which is applicable to each lifecycle phase, is as follows: [ROOK]

• -- Select the appropriate metrics to be used to assess activities and outputs in each phase of the lifecycle.
  
  
• -- Determine the goals or expected values of the metrics.
  
  
• -- Determine or compute the measures, or actual values.
  
  
• -- Compare the actual values with the expected values or goals.
  
  
• -- Devise a plan to correct any observed deviations from the expected values.

Some complications may be involved when applying this approach to software. First, there will often be many possible causes for deviations from expectations and for each cause there may be several different types of corrective actions. Therefore, it must be determined which of the possible causes is the actual cause before the appropriate corrective action can be taken. In addition, the expected values themselves may be inappropriate, when there are no very accurate models available to estimate them.

In addition to monitoring using expected values derived from other projects, metrics can also identify anomalous components that are unusual with respect to other components values in the same project. In this case, project monitoring is based on internally generated project norms, rather than estimates from other projects.

The metrics described in the following subsections are defined in [ROOK], [IEEE982.2] and [AIRFORCE], [SQE], and [ZAGE] and comprise a representative sample of management and quality metrics that can be used in the lifecycle phases to support error analysis. This section does not evaluate or compare metrics, but provides definitions to help readers decide which metrics may be useful for a particular application.

Primitive metrics such as those listed below can be collected throughout the lifecycle. These metrics can be plotted using bar graphs, histograms, and Pareto charts as part of statistical process control. The plots can be analyzed by management to identify the phases that are most error prone, to suggest steps to prevent the recurrence of similar errors, to suggest procedures for earlier detection of faults, and to make general improvements to the development process.

Problem Metrics

Primitive problem metrics.


Number of problem reports per phase, priority, category, or cause

Number of reported problems per time period

Number of open real problems per time period

Number of closed real problems per time period

Number of unevaluated problem reports

Age of open real problem reports

Age of unevaluated problem reports

Age of real closed problem reports

Time when errors are discovered

Rate of error discovery

Primitive cost and effort metrics.

Time spent

Elapsed time

Staff hours

Staff months

Staff years

Primitive change metrics.


Number of revisions, additions, deletions, or modifications

Number of requests to change the requirements specification and/or design during

lifecycle phases after the requirements phase

Primitive fault metrics. Assesses the efficiency and effectiveness of fault resolution/removal activities, and check that sufficient effort is available for fault resolution/removal.



Number of unresolved faults at planned end of phase

Number of faults that, although fully diagnosed, have not been corrected, and number of outstanding change requests

Number of requirements and design faults detected during reviews and walkthroughs

The main reasons to measure requirements specifications is to provide early warnings of quality problems, to enable more accurate project predictions, and to help improve the specifications.

1. Identify the purpose and the characteristics of the process to be monitored.
2. Select the appropriate type of control chart based on the type of characteristic measured, the data available, and the purpose of the application.
3. Determine the sampling method (e.g., number of samples (n), size of samples, time frame).
4. Collect the data.
5. Calculate the sample statistics: average, standard deviation, upper and lower control limits.
6. Construct the control chart based on sample statistics.
   
7. Monitor the process by observing pattern of the data points and whether they fall within control limits.

A run chart is a simplified control chart, in which the upper and lower control limits are omitted. The purpose of the run chart is more to determine trends in a process, rather than its variation. Although very simple, run charts can be used effectively to monitor a process, e.g., to detect sudden changes and to assess the effects of corrective actions. Run charts provide the input for establishing control charts after a process has matured or stabilized in time. Limitations of this technique are that it analyzes only one characteristic over time, and it does not indicate if a single data point is an outlier. This section is based on [OPMC] and [CAPRIO].

Implementation

1. Decide which outputs of a process to measure.
2. Collect the data.
3. Compute and draw the average line.
4. Plot the individual measurements chronologically.
5. Connect data points for ease of interpretation.

A bar graph is a frequency distribution diagram in which each bar represents a characteristic/attribute, and the height of the bar represents the frequency of that characteristic. The horizontal axis may represent a continuous numerical scale (e.g., hours), or a discrete non-numerical scale (e.g., Module A, Module B or Requirements Phase, Design Phase). Generally, numerical-scale bar charts in which the bars have equal widths are more useful for comparison purposes; numerical-scale bar charts with unequal intervals can be misleading because the characteristics with the largest bars (in terms of area) do not necessarily have the highest frequency. This section is based on [SMITH].

Implementation

1. Define the subject and purpose.
2. Collect the data. Check that the sample size is sufficient.
3. Sort the data by frequency (or other measure) of characteristics.
4. For numerical-scale bar charts, determine the number of bars and the width of the bars (class width), by trying series of class widths, avoiding too fine or too coarse a granularity. The class widths in a chart may be all the same, or they may vary (as in fig. 5-2b), depending on how one wants to show the distribution of the data.
5. Construct the chart and draw the bars. The height of a bar represents the frequency of the corresponding characteristic.

A Pareto diagram is a special use of the bar graph in which the bars are arranged in descending order of magnitude. The purpose of Pareto analysis, using Pareto diagrams, is to identify the major problems in a product or process, or more generally, to identify the most significant causes for a given effect. This allows a developer to prioritize problems and decide which problem area to work on first. This section is based on [OPMC] and [CAPRIO].

Implementation

1. Follow the steps for constructing a bar graph, except that the bars should be in descending order of magnitude (height).
2. Determine the "vital few" causes by drawing a cumulative percent line and applying the 20/80 rule.
3. Compare and identify the major causes. Repeat process until root cause of the problem is revealed.

Figure 5-3. Example Pareto Chart - Percentage of Defects by Type.

A histogram is a frequency distribution diagram in which the frequencies of occurrences of the different variables being plotted are represented by bars. The purpose is to determine the shape of the graph relative to the normal distribution (or other distributions). It is often confused with a bar graph, in which the frequency of a variable is indicated by the height of the bars. In a histogram, the frequency is indicated by the area of the bar. Histograms can only be used with variable data, which require measurements on a continuous scale. Only one characteristic can be shown per histogram, and at least 30 observations representing homogenous conditions are needed. This section is based on [OPMC], [CAPRIO], and [FREEDMAN].

Implementation

1. Define the subject and purpose.
2. Collect the data and organize from smallest to largest values. Check that sample size is sufficient.
3. Calculate the range (r), i.e. the difference between the largest and smallest values.
4. Decide arbitrarily on the number of bars (k), usually between 7 and 13.
5. To make bars of equal width, use the equation, w = r/k to calculate the interval or width (w) of the bars.
6. Sort the data into the appropriate intervals and count the number of data points that fall in each interval.
7. Calculate the frequencies (actual counts or percent) for each interval.
8. Draw the bars. The height of the bar is calculated by dividing the frequency by w, the width of the bar (in horizontal units).

A scatter diagram is a plot of the values of one variable against those of another variable to determine the relationship between them. This technique was popularized by Walter Shewhart at Bell Laboratories. Scatter diagrams are used during analysis to understand the cause and effect relationship between two variables. They are also called correlation diagrams. This section is based on [KITCHENHAM], [OPMC], and [CAPRIO].

Implementation

1. Define the subject and select the variables.
2. Collect the data.
3. Plot the data points using an appropriate scale.
4. Examine the pattern to determine whether any correlation exists (e.g., positive, negative). For a more precise specification of the relationship, regression, curve fitting or smoothing techniques can be applied.

• -- Complexity vs. defect density (example shown in fig. 5-5)
• -- Effort vs. duration (of an activity)
• -- Failures vs. time
• -- Failures vs. size
• -- Cost vs. time

This technique can be used in conjunction with scatter diagrams to obtain a more precise relationship between variables. It is used to determine the equation of the regression line, i.e., the line that "best fits" the data points. With this equation, one can approximate values of one variable when given values of the other. The equation of the line is Y = a + bX, where a and b are constants which minimize S, the sum of squares of the deviations of all data points from the regression line. For any sample value x_i of X, the expected Y value is a + bx_i. This section is based on [OPMC], [CAPRIO], and [SMITH].

Implementation

1. Collect n data values for each of the 2 variables, X and Y, denoted by x₁, x₂,..., x_n and y₁, y₂,..., y_n.
   
   
2. Minimize S = (y_i - a - bx_i)² by first taking the partial derivative of S with respect to a and then with respect to b, setting these derivatives to zero, and then solving for a and b.
   
   
3. The results obtained from steps should be the following⁸, where X_B = x_i /n and Y_B = y_i /n:
   b = (X_i - X_B)(Y_i - Y_B)a = Y_B - bX_B
   --------------------------
   (X_i - X_B)²
   

"Reliability" is used in a general sense to express a degree of confidence that a part or system will successfully function in a certain environment during a specified time period [JURAN]. Software reliability estimation models can predict the future behavior of a software product, based on its past behavior, usually in terms of failure rates. Since 1972, more than 40 software reliability estimation models have been developed, with each based on a certain set of assumptions characterizing the environment generating the data. However, in spite of much research effort, there is no universally applicable software reliability estimation model which can be trusted to give accurate predictions of reliability in all circumstances [BROCKLEHURST].

It is usually possible to obtain accurate reliability predictions for software, and to have confidence in their accuracy, if appropriate data is used [ROOK]. Also, the use of reliability estimation models is still under active research, so improvements to model capability are likely. Recent work by Littlewood (1989), for example, involves the use of techniques for improving the accuracy of predictions by learning from the analysis of past errors [ROOK], and recalibration [BROCKLEHURST].

Some problems have been encountered by those who have tried to apply reliability estimation models in practice. The algorithms used to estimate the model parameters may fail to converge. When they do, the estimates can vary widely as more data is entered [DACS]. There is also the difficulty of choosing which reliability model to use, especially since one can not know a priori which of the many models is most suitable in a particular context [BROCKLEHURST]. In general, the use of these models is only suitable for situations in which fairly modest reliability levels are required [ROOK].

There are three general classes of software reliability estimation models: nonhomogeneous Poisson process (NHPP) models, exponential renewal NHPP models, and Bayesian models. Some of the more common reliability estimation models are described below [DUNN], [LYU].

• -- Jelinski-Moranda (JM). One of the earliest models, it assumes the debugging process is purely deterministic, that is, that each defect in the program is equally likely to produce failure (but at random times), and that each fix is perfect, i.e., introduces no new defects. It also assumes that the failure rate is proportional to the number of remaining defects and remains constant between failures. This model tends to be too optimistic and to underestimate the number of remaining faults; this effect has been observed in several actual data sets.
  
  
• -- Goel-Okumoto (GO). This model is similar to JM, except it assumes the failure rate (number of failure occurrences per unit of time) improves continuously in time.
  
  
• -- Yamada Delayed S-Shape. This model is similar to GO, except it accounts for the learning period that testers go through as they become familiar with the software at the start of testing.
  
  
• -- Musa-Okumoto (MO). This NHPP model is similar to GO, except it assumes that later fixes have a smaller effect on a program's reliability than earlier ones. Failures are assumed to be independent of each other.
  
  
• -- Geometric. This model is a variation of JM, which does not assume a fixed, finite number of program errors, nor does it assume that errors are equally likely to occur.
  
  
• -- Schneidewind. Similar to JM, this model assumes that as testing proceeds with time, the error detection process changes, and that recent error counts are usually more useful than earlier counts in predicting future counts.
  
  
• -- Bayesian Jelinski-Moranda (BJM) Model. This model is similar to JM, except that it uses a Bayesian inference scheme, rather than maximum likelihood. Although BJM does not drastically underestimate the number of remaining errors, it does not offer significant improvement over JM. Actual reliability predictions of the two models are usually very close.
  
  
• -- Littlewood. This model attempts to answer the criticisms of JM and BJM by assuming that different faults have different sizes, i.e., they contribute unequally to the unreliability of the software. This assumption represents the uncertainty about the effect of a fix.
  
  
• -- Littlewood-Verrall (LV). This model takes into account the uncertainty of fault size and efficacy of a fix (i.e., a fix is of uncertain magnitude and may make a program less reliable), by letting the size of the improvement in the failure rate at a fix vary randomly.
  
  
• -- Brooks and Motley (BM). The BM binomial and Poisson models attempt to consider that not all of a program is tested equally during a testing period and that only some portions of the program may be available for testing during its development.
  
  
• -- Duane. This model assumes that the failure rate changes continuously in time, i.e., it follows a nonhomogeneous Poisson process. The cumulative failure rate when plotted against the total testing time on a ln-ln graph follows a straight line. The two parameters for the equation of the line can be derived using the method of least squares.

The following is a generic procedure for estimating software reliability [AIAA]. It can be tailored to a specific project or lifecycle phase; thus some steps may not be used in some applications.

1. Identify the application. The description of the application should include, at a minimum, the identification of the application, the characteristics of the application domain that may affect reliability, and details of the intended operation of the application system.
   
   
2. Specify the requirement. The reliability requirement should be specific enough to serve as a goal (e.g., failure rate of 10-9 per hour).
   
   
3. Allocate the requirement. The reliability requirement may be distributed over several components, which should be identified.
   
   
4. Define failure. A specific failure definition is usually agreed upon by testers, developers, and users prior to the beginning of testing. The definition should be consistent over the life of the project. Classification of failures (e.g., by severity) is continuously negotiated.
   
   
5. Characterize the operational environment. The operational environment should be described in terms of the system configuration (arrangement of the system's components), system evolution and system operational profile (how system will be used).
   
   
6. Select tests. The test team selects the most appropriate tests for exposing faults. Two approaches to testing can be taken: testing duplicates actual operational environments as closely as possible; or testing is conducted under more severe conditions than expected in normal operational environments, so that failures can occur in less time.
   
   
7. Select the models. The user should compare the models prior to selection based on the following criteria: predictive validity, ease of parameter measurement, quality of the model's assumptions, capability, applicability, simplicity, insensitivity to noise, and sensitivity to parameter variations.
   
   
8. Collect data. See section 5.2.
   
   
9. Determine the parameters. There are three common methods of estimating the parameters from the data: method of moments, least squares, and maximum likelihood. Each of these methods has useful attributes, but maximum likelihood estimation is the most commonly used approach. However, for some models, the maximum likelihood method does not yield equations for the parameters in closed form, so instead numerical methods (e.g., Newton's method) must be used [ROME]. As stated previously, some datasets may cause the numerical methods not to converge. There exist automated software reliability engineering tools, which are capable of performing parameter estimation.
   
   
10. Validate the model. The model should be continuously checked to verify that it fits the data, by using a predictive validity criteria or a traditional statistical goodness-of-fit test (e.g., Chi-square).
    
    
11. Perform analysis. The results of software reliability estimation may be used for several purposes, including, but not limited to, estimating current reliability, forecasting achievement of a reliability goal, establishing conformance with acceptance criteria, managing entry of new software features or new technology into an existing system, or supporting safety certification.

A disadvantage of these models is that they rely on testing and hence are used rather late in the development life cycle. The models are usually time based, that is, the probability is based on time to failure. Research is needed to identify how to use more valuable parameters with these models. See [ROOK]

Application Examples
Applicability of the models should be examined through various sizes, structures, functions and application domains. An advantage of a reliability model is its usability in different development and operational environments, and in different lifecycle phases. Software reliability models should be used when dealing with the following situations:

• -- Evolving software (i.e., software that is incrementally integrated during testing)
• -- Classification of failure severity
• -- Incomplete failure data
• -- Hardware execution rate differences
• -- Multiple installations of the same software
• -- Project environments departing from model assumptions

Error analysis for software consists of many activities to assure the quality of delivered software. The activities include error detection, analysis, resolution on an individual level and also on a collective level. In the latter case, the collective data may be used to locate common errors within a product, to identify areas for improvement in software development and maintenance, and to identify areas for improvement in error analysis.

Many activities of error analysis may be conducted during the early phases of the software development lifecycle to prevent error propagation and to reduce costs of fixing the errors at a later time in the lifecycle. Finding the root cause of an error discovered in system test may require analysis of code, design specifications, software requirements documentation, and perhaps analysis and test documentation. Correction of the error results in additional verification and testing activities through the lifecycle products. The time spent initially in preparing correctly stated software requirements will pay off in reduced time needed for rework in the later phases.

The error detection techniques described in this report are a representative sampling of the most widely-used error detection techniques and those most frequently referenced in standards, guidelines and technical literature. The report also describes some techniques which represent new approaches and are not yet widespread. The techniques include those that examine software products without executing the software, those that execute (or simulate the execution of) the software, and those that are based on mathematical analysis and proof of correctness techniques. Evidence of the effectiveness of any of these techniques may be hard to find. Journal articles report success of some techniques, but most often anecdotal evidence is provided through conference presentations and discussions among colleagues.

With many techniques to choose from, appropriate selection for a specific project depends on the characteristics of the project, such as the types of problems most likely to occur. Other selection criteria, which are outside the scope of this report, include parameters like development processes, skill of the technical staff, project resources, and quality requirements of the project.

The study of standards for high integrity software reported in [NUREG, NIST204] indicated that these standards are beginning to require techniques of all kinds with some guidelines attempting to base the requirement on the quality requirements and problem types of the software project. An examination of approximately 50 standards, draft standards, and guidelines indicates that these documents vary widely in their recommendations for error analysis.

Error detection is only one activity of error analysis. Information about the detected error must be reported and delivered to any persons responsible for correcting the error, managing the project, analyzing the data for process improvement and identifying similar problems in the product. Individual problem reports may be collected and analyzed using statistical process control techniques, to determine and monitor the efficiency and adequacy of the development process. Findings which result from using SPC techniques should be used as feedback to improve the development process for the current, as well as future, products. Data on single errors is also used in estimating software reliability and in predicting the number of errors (at later stages in the lifecycle).

Vendors should use error analysis to collect error data and to build corporate memory databases for use across projects. They may use the error information to identify appropriate techniques for similar projects and to better understand how to produce quality software systems.

Regulatory agencies should also consider establishing and maintaining a database of error analysis data for software systems. Willingness of vendors to participate must include mechanisms to assure confidentiality of proprietary information and that vendor data will not be used in a recriminatory sense. The database must contain both developmental and operational error data for effective use. Vendors, auditors, regulators, and the software engineering community may all benefit from use of error databases.

Auditors may use the information in the database to identify the most error-prone features of specific high integrity systems and may ensure that their audits examine these features carefully. The auditors may use the data to identify acceptance limits on different aspects of the high integrity software safety system.

Regulators may use the information from a database in several ways. First, over time, it may become apparent that some error analysis techniques are more effective than others with respect to a given type of problem. It may also become apparent that problems in these areas occur most often with certain development practices or occur less frequently with other development practices. This knowledge may influence recommendations in regulatory documents.

Finally, careful analysis of the information in the database may enable the software engineering community in general to identify research needs for software development practices to prevent specific problems from occurring and error analysis techniques to locate problems as soon as possible in the software lifecycle.

[AIAA]

R-013-1992, "Recommended Practice for Software Reliability," American Institute of Aeronautics and Astronautics, Space-Based Observation Systems Committee on Standards, Software Reliability Working Group, 1992, c/o Jim French, AIAA Headquarters, 370 L'Enfant Promenade, SW, Washington, DC 20024-2518W.

[AIRFORCE]

AFSCP 800-14, Air Force Systems Command, Software Quality Indicators, "Management Quality Insight," Department of the Air Force, January 20, 1987.

[ALBRECHT]

Albrecht, Allan J. and John E. Gaffney, Jr., "Software Function, Source Lines of Code, and Development Effort Prediction: A Software Science Validation," IEEE Transactions on Software Engineering, Vol. SE-9, No. 6, November, 1983.

[ANS104]

ANSI/ANS-10.4-1987, "Guidelines for the Verification and Validation of Scientific and Engineering Computer Programs for the Nuclear Industry," American Nuclear Society, May 13, 1987.

[ASMENQA2]

ASME NQA-2a-1990 Part 2.7, "Quality Assurance Requirements for Nuclear Facility Applications," The American Society of Mechanical Engineers, November 1990.

[BASILI]

Basili, V.R. and R.W. Selby, "Comparing the Effectiveness of Software Testing Strategies," IEEE Transactions on Software Engineering , IEEE Computer Society, Volume SE-13, Number 12, December 1987, pp. 1278-1296.

[BEIZER]

Beizer, Boris, Software Testing Techniques, Van Nostrand Reinhold Company, New York, 1983.

[BROCKLEHURST]

Brocklehurst, S., P. Y. Chan, Bev Littlewood, and John Snell, "Recalibrating Software Reliability Models," IEEE Transactions on Software Engineering , Vol. 16, No. 4, 1990.

[CAPRIO]

Caprio, William H., "The Tools for Quality," Total Quality Management Conference, Ft. Belvoir, Virginia, July 13-15, 1992.

[CLARK]

Clark, Peter, and Bard S. Crawford, Evaluation and Validation (E&V) Reference Manual, TASC No. TR-5234-3, Version 3.0, February 14, 1991.

[DACS]

"Software Reliability Models," DACS Newsletter, Data & Analysis Center for Software, Volume X, Number 2, Summer, 1992.

[DEMILLO]

DeMillo, Richard A. et al. Software Testing and Evaluation , The Benjamin/Cummings Publishing Company, Inc., Menlo Park, CA, 1987.

[DEMMY]

Demmy, W. Steven and Arthur B. Petrini, "Statistical Process Control in Software Quality Assurance," Proceedings of the 1989 National Aerospace and Electronics Conference, NAECON, May 22-26, 1989, Dayton, OH, IEEE, Inc., Piscataway, NJ, p. 1585-1590.

[DUNN]

Dunn, Robert. Software Defect Removal, McGraw-Hill, Inc., 1984.

[EWICS3]

Bishop, P. G. (ed.), Dependability of Critical Computer Systems 3 - Techniques Directory, The European Workshop on Industrial Computer Systems Technical Committee 7 (EWICS TC7), Elsevier Science Publishers Ltd, 1990.

[FAGAN]

Fagan, M. E., "Design and Code Inspections to Reduce Errors in Program Development," IBM Systems Journal, Volume 15, Number 3, 1976.

[FLORAC]

CMU/SEI-92-TR-ZZZ, "Software Quality Measurement: A Framework for Counting Problems, Failures, and Faults," William Florac, The Quality Subgroup of the Software Metrics Definition Working Group and the Software Process Measurement Project Team, Software Engineering Institute, Carnegie Mellon University, Pittsburgh, PA, Draft, 1992.

[FREEDMAN]

Freedman, David, Robert Pisani, and Roger Purves, "Statistics," W.W. Norton & Company, Inc., New York, 1978.

[GRADY]

Grady, Robert B. and Caswell, Deborah, Software Metrics: Establishing a Company-Wide Program, Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1987.

[GREEN]

Green, Scott E. and Rose Pajerski, "Cleanroom Process Evolution in the SEL," Proceedings of the Sixteenth Annual Software Engineering Workshop, National Aeronautics and Space Administration, Goddard Space Flight Center, Greenbelt, MD 20771, December 1991.

[IEC65A94]

IEC 65A(Secretariat)94, "Draft British Standard 0000: Software for Computers in the Application of Industrial Safety-Related Systems," WG9, December 6, 1989.

[IEC65A122]

IEC 65A(Secretariat)122, "Software for Computers in the Application of Industrial Safety- Related Systems," WG9, Version 1.0, September 26, 1991.

[IEEEGLOSS]

ANSI/IEEE Std 610.12, "IEEE Standard Glossary of Software Engineering Terminology," The Institute of Electrical and Electronics Engineers, February, 1991.

[IEEEP1044]

IEEE P1044, "Draft Standard of: A Standard Classification for Software Errors, Faults, and Failures," The Institute of Electrical and Electronics Engineers, August 1991.

[IEEE982.2]

ANSI/IEEE Std 982.2-1988, "Guide for the Use of IEEE Standard Dictionary of Measures to Produce Reliable Software," The Institute of Electrical and Electronics Engineers, June, 1989.

[IEEE1012]

ANSI/IEEE Std 1012-1986, "IEEE Standard for Software Verification and Validation Plans," The Institute of Electrical and Electronics Engineers, Inc., November 14, 1986.

[JURAN]

Juran, J. M. (ed.), Juran's Quality Control Handbook, 4th ed., McGraw-Hill, Inc., New York, 1988.

[KELLY]

Kelly, John C., Joseph S. Sherif, and Jonathan Hops, "An Analysis of Defect Densities Found During Software Inspections," Proceedings of the Fifteenth Annual Software Engineering Workshop, National Aeronautics and Space Administration, Goddard Space Flight Center, Greenbelt, MD 20771, November, 1990.

[KITCHENHAM]

Kitchenham, B. A. and B. Littlewood, Measurement for Software Control and Assurance, Elsevier Science Publishers Ltd, London and New York, 1989.

[LYLE]

Lyle, Jim, "Program Slicing," to be published in Encyclopedia of Software Engineering, John Wiley Publishing Co., New York, New York.

[LYU]

Lyu, Michael and Allen Nikora, "Applying Reliability Models More Effectively," IEEE Software, Vol. 9., No. 4, July, 1992.

[MAKOWSKY]

Makowsky, Lawrence C., Technical Report, USA-BRDEC-TR//2516, "A Guide to Independent Verification and Validation of Computer Software," United States Army, Belvoir Research, Development and Engineering Center, June 1992.

[MILLER]

Miller, Keith W., et al, "Estimating the Probability of Failure When Testing Reveals No Failures," IEEE Transactions on Software Engineering, Vol.18, No.1, January 1992.

[MILLS]

Mills, H. D., M. Dyer, and R. C. Linger, "Cleanroom Software Engineering," IEEE Software, September, 1987, pp. 19-25.

[MOD55]

Interim Defence Standard 00-55, "The Procurement of Safety Critical Software in Defence Equipment," Parts 1 and 2, Ministry of Defence, UK, April 5, 1991.

[MYERS]

Myers, Glenford J., The Art of Software Testing, John Wiley & Sons, New York, 1979.

[NBS93]

Powell, Patricia B., NBS Special Publication 500-93, "Software Validation, Verification, and Testing Technique and Tool Reference Guide," U.S. Department of Commerce/National Bureau of Standards (U.S.), September 1982.

[NIST187]

NIST SPEC PUB 500-187, "Application Portability Profile (APP) The U.S. Government's Open System Environment Profile OSE/1 Version 1.0," U.S. Department of Commerce/National Institute of Standards and Technology, April 1991.

[NIST204]

Wallace, D.R., L.M. Ippolito, D.R. Kuhn, NIST SPEC PUB 500-204, "High Integrity Software Standards and Guidelines," U.S. Department of Commerce/National Institute of Standards and Technology, September, 1992.

[NISTGCR]

Craigen, Dan, Susan Gerhart, Ted Ralston, NISTGCR 93/626, "An International Surveyof Industrial Applications of Formal Methods," Volumes 1 and 2, U.S. Department of Commerce/National Institute of Standards and Technology, March, 1993.

[NISTIR]

Wallace, D.R., W.W. Peng, L.M. Ippolito, NISTIR 4909,"Software Quality Assurance: Documentation and Reviews," U.S. Department of Commerce/National Institute of Standards and Technology, 1992.

[NUREG]

Wallace, D.R., L.M. Ippolito, D.R. Kuhn, NUREG/CR-5930, "High Integrity Software Standards and Guidelines," U.S. Nuclear Regulatory Commission, September, 1992.

[OPMC]

The Organizational Process Management Cycle Programmed Workbook, Interaction Research Institute, Inc., Fairfax, Virginia.

[PARK]

Park, Robert, CMU/SEI-92-TR-20, ESC-TR-92-20, "Software Size Measurement: A Framework for Counting Source Statements," Software Engineering Institute, Carnegie Mellon University, September, 1992.

[PEYTON]

Peyton and Hess, "Software Sneak Analysis," IEEE Seventh Annual Conference of the Engineering in Medicine and Biology Society , The Institute of Electrical and Electronics Engineers, 1985.

[PUTNAM]

Putnam, Lawrence H. and Ware Myers, Measures for Excellence, Reliable Software On Time, Within Budget, Prentice-Hall, Inc., Englewood Cliffs, NJ, 1992.

[RIFKIN]

Rifkin, Stan and Charles Cox, "Measurement in Practice," Technical Report, CMU/SEI- 91-TR-16, ESD-TR-91-16, Carnegie Mellon University, 1991.

[ROME]

Iuorno, Rocco and Robert Vienneau, "Software Measurement Models," Draft, Prepared for Rome Air Development Center, Griffiss Air Force Base, NY, July 1987.

[ROOK]

Rook, Paul, Software Reliability Handbook, Elsevier Science Publishers Ltd, Crown House, London and New York, 1990.

[SCHULMEYER]

Schulmeyer, G. Gordon. Zero Defect Software, McGraw-Hill, Inc., 1990.

[SHOOMAN]

Shooman, Martin L., "A Class of Exponential Software Reliability Models," Workshop on Software Reliability, IEEE Computer Society Technical Committee on Software Reliability Engineering, Washington, DC, April 13, 1990.

[SQE]

"Software Measurement," Seminar Notebook, Version 1.2, Software Quality Engineering, 1991.

[SMITH]

Smith, Gary, Statistical Reasoning, Allyn and Bacon, Boston, MA, 1991.

[STSC]

MAC010/STSC, "Software Management Guide," Software Technology Center, Hill AFB, UT 84-56, October, 1990.

[ROME]

Proceedings of the 3rd Annual Software Quality Workshop; Alexandria Bay, New York: August 11-15, 1991, Rome Laboratory, Griffiss AFB, New York.

[VOAS]

Voas, Jeffrey M. and Keith W. Miller," A Model for Improving the Testing of Reusable Software Components," 10th Pacific Northwest Software Quality Conference, October 19- 21, 1992, Portland, Oregon.

[WALLACE]

Wallace, Dolores R. "Software Verification and Validation," to be published in Encyclopedia of Software Engineering, John Wiley Publishing Co., New York, NY.

[WALLACEFUJII]

Wallace, Dolores R. and Roger U. Fujii, "Verification and Validation: Techniques to Assure Reliability," IEEE Software, May 1989.

[ZAGE]

Zage, Wayne M. "Code Metrics and Design Metrics; An ACM Professional Development Seminar," November 19, 1991.

Description:
The purpose is to determine the correctness of an algorithm with respect to its intended use, to determine its operational characteristics, or to understand it more fully in order to modify, simplify, or improve. The analysis involves rederiving equations or evaluating the suitability of specific numerical techniques. Algorithms are analyzed for correctness, efficiency (in terms of time and space needed), simplicity, optimality, and accuracy. Algorithm analysis also examines truncation and round-off effects, numerical precision of word storage and variables (e.g., single- vs. extended-precision arithmetic), and data typing influences.

Advantages:

• -- Effective and useful in general

• -- A particular analysis depends on the particular model of computation (e.g, Turing machine, random access machine). If the assumptions of the model are inappropriate, then the analysis will be inaccurate.

• -- Incorrect, inappropriate, or unstable algorithms
• -- Program does not terminate
• -- Inability to operate on the full range of data - e.g., trigonometric routine only works in the first quadrant
• -- Incorrect analysis of computational error (effect of round-off and truncation errors)
• -- Incompatible data representations - e.g., input in lbs., program processes kilograms
• -- Incompatibility with hardware or software resources

Description:
This technique is used to detect test failures by comparing the output of two or more programs implemented to the same specification. The same input data is applied to two or more program versions and their outputs are compared to detect anomalies. Any test data selection strategy can be used for this type of testing, although random testing is well suited to this approach. Also known as comparison testing.

Advantages:

• -- Permits a large number of tests to be made with little effort
• -- Rapid fault detection

• -- Requires the construction of at least one secondary program, although this may be available as part of the overall development
• -- Discrepancies must be analyzed manually to determine which program is at fault (it is not sufficient to assume that majority is always correct)

• -- Does not detect specific errors, only anomalies or discrepancies between programs

Description:
The purpose is to detect and remove errors occurring at parameter limits or boundaries. The input domain of the program is divided into a number of input classes. The tests should cover the boundaries and extremes of the classes. The tests check that the boundaries of the input domain of the specification coincide with those in the program. The value zero, whether used directly or indirectly, should be used with special attention (e.g., division by zero, null matrix, zero table entry). Usually, boundary values of the input produce boundary values for the output. Test cases should also be designed to force the output to its extreme values. If possible, a test case which causes output to exceed the specification boundary values should be specified. If output is a sequence of data, special attention should be given to the first and last elements and to lists containing zero, one, and two elements.

Advantages:

• -- Verifies that program will behave correctly for any permissible input or output

• -- No significant disadvantages in itself but for programs with many types of input cannot test all combinations of input and therefore cannot identify problems resulting from unexpected relationships between input types

• -- Algorithm errors
• -- Array size
• -- Specification errors

Description:
This technique is most applicable to real time and data driven systems. Logic and data requirements are transformed from text into graphic flows, which are easier to analyze. Examples of control flow diagrams include PERT, state transition, and transaction diagrams. For large projects, control flow analysis using control flow diagrams that show the hierarchy of main routines and subfunctions are useful to understand the flow of program control. Purpose is to detect poor and potentially incorrect program structures. The program is represented by a directed graph, which is analyzed for the errors below.

Advantages:

• -- Simple to apply
• -- Readily automated

• -- Results require some interpretation. Identified anomalies may not be faults.
• -- Sometimes difficult to deal with "aliasing" where different variables are associated with the same locations in memory

• -- Inaccessible/unreachable code
• -- Knotted code - If code is well-structured the directed graph can be reduced to a single node. If code is poorly structured, it can only be reduced to a "knot" composed of several nodes.

Description:
Database analysis is performed on programs with significant data storage to ensure that common data and variable regions are used consistently between all call routines; data integrity is enforced and no data or variable can be accidentally overwritten by overflowing data tables; data access through indirect access is checked; and data typing and use are consistent throughout all program elements. Useful for programs that store program logic in data parameters. The purpose is to ensure that the database structure and access methods are compatible with the logical design. Diagrams are useful for understanding user privileges.

Advantages:

• -- Supports interface analysis

• -- May require manual examination of diagrams for access problems

• -- Inconsistent use of data types
• -- Incorrect access protections

Description:
The purpose is to detect poor and potentially incorrect program structures. Data flow analysis combines the information obtained from the control flow analysis with information about which variables are read or written in different portions of code. May also be used in the design and implementation phases.

Advantages:

• -- Readily automated
• -- Easy to apply

• -- Requires some interpretation
• -- Some anomalies may not be faults

• -- Undefined input/output data or format
• -- Incorrect data flow
• -- Variables that are read before they are written (likely to be an error, and is bad programming practice)
• -- Omitted code - indicated by variables that are written more than once without being read
• -- Redundant code - indicated by variables that are written but never read

Description:
Data flow diagrams are used to describe the data flow through a program in a diagrammatic form. They show how data input is transformed to output, with each stage representing a distinct transformation. The diagrams use three types of components:

1. Annotated bubbles - bubbles represent transformation centers and the annotation specifies the transformation
2. Annotated arrows - arrows represent the data flow in and out of the transformation centers, annotations specify what the data is
3. Operators (AND, XOR) - used to link the annotated arrows

• -- They show transformations without making assumptions about how the transformations are implemented.

• -- Inability to provide information about the transformation process

• -- Incorrect data input/output
• -- Inconsistencies in data usage

Description:
The purpose is to provide a clear and coherent analysis of complex logical combinations and relationships. This method uses two-dimensional tables to concisely describe logical relationships between boolean program variables.

Advantages:

• -- Their conciseness and tabular nature enable the analysis of complex logical combinations expressed in code.
• -- Potentially executable if used as specifications

• -- Tedious effort required

• -- Logic

Description:
Code is read by an expert, other than the author of the code, who performs any of the following: looking over the code for obvious defects, checking for correct procedure interfaces, reading the comments to develop a sense of what the code does and then comparing it to its external specifications, comparing comments to design documentation, comparing comments to design documentation, stepping through with input conditions contrived to "exercise" all paths including those not directly related to the external specifications, checking for compliance with programming standards and conventions, any combination of the above.

Advantages:

• -- Inexpensive
• -- Capable of catching 30% of all errors, if performed meticulously
• -- Can be more effective than functional testing or structural testing (NASA Goddard Space Flight Center experiment, see [BASILI])

• -- Requires enormous amount of discipline
• -- Few people are able to use this technique effectively
• -- Usually less effective than walkthroughs or inspections

• -- unreachable code
• -- improper nesting of loops and branches
• -- inverted predicates
• -- incomplete predicates
• -- improper sequencing of processes
• -- infinite loops
• -- instruction modification
• -- failure to save or restore registers
• -- unauthorized recursion
• -- missing labels or code
• -- unreferenced labels

• -- missing validity tests
• -- incorrect access of array components
• -- mismatched parameter lists
• -- initialization faults
• -- anachronistic data
• -- undefined variables
• -- undeclared variables
• -- misuse of variables (locally and globally)
• -- data fields unconstrained by natural or defined data boundaries
• -- inefficient data transport

• -- calls to subprograms that do not exist
• -- improper program linkages
• -- input-output faults
• -- prodigal programming
• -- failure to implement the design

Description:
The purpose of this technique is to determine whether a set of test cases is adequate. Some known error types are inserted into the program, and the program is executed with the test cases under test conditions. If only some of the seeded errors are found, the test case set is not adequate. The ratio of found seeded errors to the total number of seeded errors is approximate equal to the ratio of found real errors to total number of errors, or

          Number of seeded         Number of real
          errors found        =    errors found    
          ----------------         --------------
          Total number of          Total number of
          seeded errors            real errors

• -- Provides indication that test cases are structured adequately to locate errors.

• -- For this method to be valid and useful, the error types and the seeding positions must reflect the statistical distribution of real errors.

• -- Does not detect errors, but determines adequacy of test set

Description:
The purpose is to define or implement the control structure of a system. Many systems can be defined in terms of their states, inputs, and actions. For example, a system is in state S1, receives an input I, then carries out action A, and moves to state S2. By defining a system's actions for every input in every state we can completely define a system. The resulting model of the system is a finite state machine (FSM). It is often drawn as a state transition diagram, which shows how the system moves from one state to another, or as a matrix in which the dimensions are state and input. Each matrix entry is identified by a state and input, and it specifies the action and new state resulting from receipt of the input in the given state.

Advantages:

• -- Allows important properties to be checked mechanically and reliably.
• -- Simple to work with

• -- None that are major

• -- Incomplete requirements specification - check that there is an action and new state for every input in every state
• -- Inconsistent requirements - check that only one state change is defined for each state and input pair

Description:
The purpose is to check whether software fulfills its intended function. Involves the use of theoretical and mathematical models to prove the correctness of a program without executing it. The requirements should be written in a formal specification language (e.g., VDM, Z) so that these requirements can then be verified using a proof of correctness. Using this method, the program is represented by a theorem and is proved with first-order predicate calculus. A number of assertions are stated at various locations in the program, and are used as pre and post conditions to various paths in the program. The proof consists of showing that the program transfers the preconditions into the postconditions according to a set of logical rules, and that the program terminates.

Advantages:

• -- Allows for rigorous statement concerning correctness
• -- Possibly the only way of showing the correctness of general WHILE loops

• -- Time consuming to do manually
• -- Requires enormous amount of intellectual effort
• -- Must be checked mechanically for human errors
• -- Difficult to apply to large software systems
• -- If formal specifications are not used in the design, then formal verification (proof of correctness after implementation tends to be extremely difficult
• -- Only applicable to sequential programs, not concurrent program interactions

Description:
An extension of data flow analysis, in which the actual data flows (both within and between procedures) are compared with the design intent. Normally implemented with an automated tool where the intended data flows are defined using a structured comment that can be read by the tool.

Advantages:

• -- Simple to apply
• -- Readily automated

• -- Results require some interpretation
• -- Some anomalies may not be faults

• -- Undefined input / output data or format
• -- Incorrect flow of information

• -- Provides comprehensive statistics on classes of errors
• -- Studies have shown that inspections are an effective method of increasing product quality (e.g., reliability, usability, maintainability)
• -- Effective for projects of all sizes
• -- Qualitative benefits: less complex programs, subprograms written in a consistent style, highly visible system development, more reliable estimating and scheduling, increased user satisfaction, improved documentation, less dependence on key personnel for critical skills

• -- Inspectors must be independent of programmers
• -- Programmers may feel inspection is a personal attack on their work
• -- Time consuming, involving several staff (2 or 3 pages of not-difficult code may take 3 h to inspect)

• -- Weak modularity
• -- Failure to handle exceptions
• -- Inexpansible control structure
• -- Nonexisting or inadequate error traps
• -- Incomplete requirements
• -- Infeasible requirements
• -- Conflicting requirements
• -- Incorrect specification of resources

Description:
This technique is used to demonstrate that the interfaces of subprograms do not contain any errors or any errors that lead to failures in a particular application of the software or to detect all errors that may be relevant. Interface analysis is especially important if interfaces do not contain assertions that detect incorrect parameter values. It is also important after new configurations of pre-existing subprograms have been generated. The types of interfaces that are analyzed include external, internal, hardware/hardware, software/software, software/hardware, and software/database. Interface analysis may include the following:

• -- Analysis of all interface variables at their extreme positions
• -- Analysis of interface variables individually at their extreme values with other interface variables at normal values
• -- Analysis of all values of the domain of each interface variable with other interface variables at normal values

• -- Can locate problems that would prevent system from functioning due to improper interfaces
• -- Especially useful for software requirements verification and design verification
• -- When used with prototyping or simulation, can find many critical errors that would be costly to correct in the delivered system
• -- Software design tools exist for analysis of interfaces, during design phase

• -- Manual effort may be time-consuming

• -- Input / output description errors (e.g., values of input variables altered)
• -- Actual and formal parameters mismatch (precision, type, units, and number)
• -- Incorrect functions used or incorrect subroutine called
• -- Inconsistency of attributes of global variables (common, etc.)
• -- Incorrect assumptions about static and dynamic storage of values (i.e., whether local variables are saved between subroutine calls)
• -- Inconsistencies between subroutine usage list and called subroutine

Description:
Similar to interface analysis, except test cases are built with data that tests all interfaces. Interface testing may include the following:

• -- Testing all interface variables at their extreme positions
• -- Testing interface variables individually at their extreme values with other interface variables at normal values
• -- Testing all values of the domain of each interface variable with other interface variables at normal values
• -- Testing all values of all variables in combination (may be feasible only for small interfaces).

• -- Locates errors that may prevent system from operating at all or locates errors in timing of interface responses (e.g., slow system response to users was a factor in the failure of the THERAC medical system)

• -- Without automation of design, or code modules, manual searching for interface parameters in all design or code modules can be time consuming

• -- Input / output description errors
• -- Inconsistent interface parameters

Description:
The purpose is to determine the thoroughness with which a program has been tested, and in the process, detect errors. This procedure involves producing a large set of versions or "mutations" of the original program, each derived by altering a single element of the program (e.g., changing an operator, variable, or constant). Each mutant is then tested with a given collection of test data sets. Since each mutant is essentially different from the original, the testing should demonstrate that each is in fact different. If each of the outputs produced by the mutants differ from the output produced by the original program and from each other, then the program is considered adequately tested and correct.

Advantages:

• -- Applicable to any algorithmic solution specification
• -- Results are good predictors of operational reliability

• -- Likely to require significant amounts of human analyst time and good insight
• -- Requires good automated tools to be effective
• -- Reliable only if all possible mutation errors are examined
• -- Cannot assure the absence of errors which cannot be modeled as mutations

• -- Any errors that can be found by test

Description:
The purpose is to measure how well the software system executes according to its required response times, cpu usage, and other quantified features in operation. These measurements may be simple to make (e.g., measuring process time relative to volumes of input data) or more complicated (e.g., instrumenting the code to measure time per function execution).

Advantages:

• -- Useful for checking timing synchronization of functions, memory locations, memory requirements and other performance features.

• -- Caution required in instrumenting code to ensure the instrumentation itself does not interfere with processing of functions or with locations of bytes under examination

• -- Timing, synchronization, and memory allocation

Description:
Purpose is to check the feasibility of implementing a system against the given constraints and to communicate the specifier's interpretation of the system to the customer, in order to locate misunderstandings. A subset of system functions, constraints, and performance requirements are selected. A prototype is built using high level tools, is evaluated against the customer's criteria, and the system requirements may be modified as a result of this evaluation.

Advantages:

• -- Better communication with customer
• -- Early detection of problems
• -- Check feasibility of new ideas or techniques

• -- Unnecessary and expensive if problem is well understood
• -- Tools are needed for quick implementation

• -- User related
• -- Interface related
• -- Omitted functions
• -- Undesired functions
• -- Poorly defined functionality (e.g., specifications that do not cover all expected cases)
• -- Errors in the specification that lead to inconsistent states, failure conditions, erroneous results
• -- Contradictions in requirements
• -- Impossibility, infeasibility of requirements

• -- Expense can appear to be prohibitive, especially for small changes. However, it can often be determined which functions may be affected by a given change, so that the amount of testing can be reduced in these cases

• -- Errors caused by system modifications or corrections

Description:
This technique involves the examination of individual requirements statements to ensure that each statement is complete, readable, accurate, and consistent with other requirements. The manual technique requires analysis of the attributes of a statement: initiator of action, the action, the object of the action, conditions (e.g., when positive state reached), constraints (limits), source of action, destination (e.g., printer, screen, plotter), mechanism, reason for the action. When the data from a set of requirements is examined collectively, the results of the analysis may aid in identifying trends.

Advantages:

• -- Promotes clarity, correctness, completeness, testability, and accuracy
• -- Can help to establish an English base from which to write a formal specification

• -- Very narrow look at each requirement. Could detract from analysis of more global examination of how the requirements fit together.

• -- Inconsistency in requirements
• -- Incomplete requirement
• -- Untestable requirement

Description:
A review is a meeting at which the requirements, design, code, or other products of a software development project are presented to the user, sponsor, or other interested parties for comment and approval, often as a prerequisite for concluding a given phase of the software development process. Usually held at end of a phase, but may be called when problems arise. Often referred to as "Formal Review" versus desktop review of materials.

Advantages:

• -- Provides opportunity to change course of a project before start of next phase.
• -- Because scope of review is usually broad, gives opportunity to recognize global problems

• -- If participants do not have materials ahead of time and spend time preparing, review will accomplish little or nothing.
• -- Attention focus on major issues

Description:
Sensitivity analysis is a new method of quantifying ultra-reliable software during the implementation phase. It is based on the fault-failure model of software and attempts to approximate this model. There are three necessary conditions:

1. The fault must be executed.
2. The fault must affect the computational data state directly after the fault location.
3. The affected data state must propagate to an output variable.

• -- While software testing provides some quantification of reliability, sensitivity analysis provides quantification of software testing.
• -- Presents a different approach to software reliability assessment, which usually is based on an exponential distribution of remaining faults over time.
• -- Promising research that those needing ultra-reliable software should continue to follow.

• -- Intended to operate on code that has already been formally verified
• -- Still new, so that effectiveness of this technique is not yet known and use is not yet widespread
• -- Developed and currently promoted by only one company
• -- Tool is not sold, but rather the service of operating the tool on a client's code is sold

• -- None, but is an aid to error detection techniques

Description:
Simulation is used to test the functions of a software system, together with its interface to the real environment, without modifying the environment in any way. The simulation may be software only or a combination of hardware and software. Simulation is used to evaluate the interactions of large, complex systems with many hardware, user, and other interfaces. A model of the system to be controlled by the actual system under test is created. This model mimics the behavior of the controlled system, and is for testing purposes only. It must provide all inputs of the system under test which will exist when that system is installed; respond to outputs from the system in a way which accurately represents the controlled system; have provision for operator inputs to provide any perturbations with which the system under test is required to cope. When software is being tested, the simulation may be a simulation of the target hardware with its inputs and outputs. In the installation phase, it is used to test operator procedures and to help isolate installation problems.

Advantages:

• -- Enables simulation of critical aspects of the software that would not be practical to analyze manually.
• -- Provides means of performing functional tests of the system's behavior in the event of catastrophes which could not otherwise not be tested
• -- Can provide a means of achieving "long term" test in a short period
• -- Can provide a means of investigating behavior at critical points by slowing the timescale or single stepping

• -- Difficult to achieve independently of the software being tested
• -- May require considerable resource both to create the simulation and to operate it
• -- Dependent on a model of the system which itself could contain safety related flaws

• -- Interface errors
• -- Logical errors
• -- Errors in performance

Description:
Sizing/timing analysis is performed during incremental code development by obtaining program sizing and execution timing values to determine if the program will satisfy processor size and performance requirements allocated to the software. Significant deviations between actual and predicted values is a possible indication of problems or the need for additional examination. This technique is most applicable to real-time programs having response time requirements and constrained memory execution space requirements.

Advantages:

• -- Opportunity to study sequence of operations, especially interfaces between humans and response to user commands
• -- Indicator of whether integrated system will perform appropriately

• -- Time consuming when conducted by poring over control flow or data flow diagrams
• -- When timing results are collected by executing code, must be careful that test code does not alter timing as it would be without the test code

• -- Unacceptable processor load
• -- Control structure ignores processing priorities

Description:
Slicing is a program decomposition technique that is based on extracting statements relevant to a specific computation from a program. It produces a smaller program that is a subset of the original program. Without intervening irrelevant statements, it is easier for a programmer interested in a subset of the program's behavior to understand the corresponding slice rather than to deal with the entire program. This technique can be applied to program debugging, testing, parallel program execution and software maintenance. Several variations of this technique have been developed, including program dicing, dynamic slicing and decomposition slicing.

Advantages:

• -- Readily automated
• -- Reduces time needed for debugging and testing

• -- Resource consuming tool/ method

• -- Aids in finding root of errors during debugging and testing, by narrowing the focus of investigation

Description:
This technique is used to detect an unexpected path or logic flow which causes undesired program functions or inhibits desired functions. Sneak circuit paths are latent conditions inadvertently designed or coded into a system, which can cause it to malfunction under certain conditions.

To perform sneak circuit analysis, source code is first converted, usually by computer, into an input language description file into topological network trees. Then the trees are examined to identify which of the six basic topological patterns appear in the trees. Analysis takes place using checklists of questions about the use and relationships between the basic topological components.

Advantages:

• -- Effective in finding errors not usually detected by desk checking or standard V&V techniques
• -- Applicable to programs written in any language
• -- Applicable to hardware, software, and the combined system

• -- Labor intensive
• -- Likely to be performed late in the development cycle, so changes will be costly

• -- Unintended functions/outputs

Description:
Involves testing the response of the system to extreme conditions (e.g., with an exceptionally high workload over a short span of time) to identify vulnerable points within the software and to show that the system can withstand normal workloads. Examples of testing conditions that can be applied include the following:

• -- If the size of the database plays an important role, then increase it beyond normal conditions.
• -- Increase the input changes or demands per time unit beyond normal conditions.
• -- For the most extreme case, put all influential factors to the boundary conditions at the same time.

• -- Often the only method to determine that certain kinds of systems will be robust when maximum numbers of users are using the system, at fastest rate possible (e.g., transaction processing) and to identify that contingency actions planned when more than maximum allowable numbers of users attempt to use system, when volume is greater than allowable amount, etc.

• -- Requires large resources

• -- Design errors related to full-service requirements of system and errors in planning defaults when system is over-stressed

Description:
This is an evaluation technique in which program execution is simulated using symbols rather than actual values for input data, and program output is expressed as logical or mathematical expressions involving these symbols.

Advantages:

• -- No input data values are needed
• -- Results can be used in program proving
• -- Useful for discovering a variety of errors

• -- Result will consist of algebraic expressions which easily get very bulky and difficult to interpret
• -- Difficult to analyze loops with variable length
• -- For most programs, the number of possible symbolic expressions is excessively large
• -- Unlikely to detect missing paths
• -- Studies have shown that in general, it is not more effective than the combined use of other methods such as static and dynamic analyses

• -- None, but is an aid for detecting errors. A person must verify the correctness of the output generated by symbolic execution in the same way that output is verified when generated by executing a program over actual values.

Description:
The purpose is to ensure that reported test results are the actual finding of the tests. Test related tools, media, and documentation shall be certified to ensure maintainability and repeatability of tests. This technique is also used to show that the delivered software product is identical to the software product that was subjected to V&V.

Advantages:

• -- Assurance that the test results are accurately presented
• -- Assurance that the corrected version of product is in compliance with test findings

• -- Often mistaken as a certification of system quality

• -- Incorrect test results reported
• -- Tests reported that never occurred
• -- Incorrect version of product shipped

Description:
There are several types of traceability analysis, including requirements trace, design trace, code trace, and test trace. Traceability analysis is the process of verifying that each specified requirement has been implemented in the design / code, that all aspects of the design / code have their basis in the specified requirements, and that testing produces results compatible with the specified requirements.

Advantages:

• -- Highly effective for detecting errors during design and implementation phases
• -- Valuable aid in verifying completeness, consistency, and testability of software
• -- Aids in retesting software when a system requirement has been changed

• -- No significant disadvantages

• -- Omitted functions
• -- Higher-order requirement improperly translated
• -- Software specification incompatible with other system specifications

• -- Omission or misinterpretation of specified requirements
• -- Detailed design does not conform to top-level design
• -- Failure to conform to standards

• -- Omission or misinterpretation of specified requirements
• -- Code does not conform to detailed design
• -- Failure to conform to standards

• -- Software does not perform functions and produce outputs in conformance with the requirements specification

Description:
A walkthrough is an evaluation technique in which a designer or programmer leads one or more other members of the development team through a segment of design or code, while the other members ask questions and make comments about technique, style, and identify possible errors, violations of development standards, and other problems. Walkthroughs are similar to reviews, but are less formal. Other essential differences include the following:

• -- Participants are fellow programmers rather than representatives of other functions
• -- Scope - standards usually ignored. Successful static analysis results generally not confirmed
• -- Checklists are rarely used
• -- Follow-up is often ignored

• -- Less intimidating than formal reviews
• -- Identifies the most error-prone sections of the program, so more attention can be paid to these sections during testing
• -- Very effective in finding logic design and coding errors

• -- Designers or programmers may feel walkthrough is an attack on their character or work

• -- Interface
• -- Logic
• -- Data
• -- Syntax