CMMI: Capability Maturity Model Integration
ICSM: International Conference on Software Maintenance
SCM: Software Configuration Management
SQA: Software Quality Assurance V&V Verification and Validation
Y2K: Year 2000
Software development efforts result in the delivery of a software product which satisfies user requirements. Accordingly, the software product must change or evolve. Once in operation, defects are uncovered, operating environments change, and new user requirements surface. The maintenance phase of the life cycle begins following a warranty period or post-implementation support delivery, but maintenance activities occur much earlier.
Software maintenance is an integral part of a software life cycle. However, it has not, historically, received the same degree of attention that the other phases have. Historically, software development has had a much higher profile than software maintenance in most organizations. This is now changing, as organizations strive to squeeze the most out of their software development investment by keeping software operating as long as possible. Concerns about the Year 2000 (Y2K) rollover focused significant attention on the software maintenance phase, and the Open Source paradigm has brought further attention to the issue of maintaining software artifacts developed by others.
In the Guide, software maintenance is defined as the totality of activities required to provide cost-effective support to software. Activities are performed during the pre-delivery stage, as well as during the post-delivery stage. Pre-delivery activities include planning for post-delivery operations, for maintainability, and for logistics determination for transition activities. Post-delivery activities include software modification, training, and operating or interfacing to a help desk.
The Software Maintenance KA is related to all other aspects of software engineering. Therefore, this KA description is linked to all other chapters of the Guide.
The Software Maintenance KA breakdown of topics is shown in Figure 1.
This first section introduces the concepts and terminology that form an underlying basis to understanding the role and scope of software maintenance. The topics provide definitions and emphasize why there is a need for maintenance. Categories of software maintenance are critical to understanding its underlying meaning.
Software maintenance is defined in the IEEE Standard for Software Maintenance, IEEE 1219, as the modification of a software product after delivery to correct faults, to improve performance or other attributes, or to adapt the product to a modified environment. The standard also addresses maintenance activities prior to delivery of the software product, but only in an information appendix of the standard.
The IEEE/EIA 12207 standard for software life cycle processes essentially depicts maintenance as one of the primary life cycle processes, and describes maintenance as the process of a software product undergoing "modification to code and associated documentation due to a problem or the need for improvement. The objective is to modify the existing software product while preserving its integrity." ISO/IEC 14764, the international standard for software maintenance, defines software maintenance in the same terms as IEEE/EIA 12207 and emphasizes the pre-delivery aspects of maintenance, planning, for example.
Software maintenance sustains the software product throughout its operational life cycle. Modification requests are logged and tracked, the impact of proposed changes is determined, code and other software artifacts are modified, testing is conducted, and a new version of the software product is released. Also, training and daily support are provided to users. Pfleeger [Pfl01] states that "maintenance has a broader scope, with more to track and control" than development.
A maintainer is defined by IEEE/EIA 12207 as an organization which performs maintenance activities [IEEE12207.0-96]. In this KA, the term will sometimes refer to individuals who perform those activities, contrasting them with the developers.
IEEE/EIA 12207 identifies the primary activities of software maintenance as: process implementation; problem and modification analysis; modification implementation; maintenance review/acceptance; migration; and retirement. These activities are discussed in topic 3.2 Maintenance Activities.
Maintainers can learn from the developer's knowledge of the software. Contact with the developers and early involvement by the maintainer helps reduce the maintenance effort. In some instances, the software engineer cannot be reached or has moved on to other tasks, which creates an additional challenge for the maintainers. Maintenance must take the products of the development, code, or documentation, for example, and support them immediately and evolve/maintain them progressively over the software life cycle.
Maintenance is needed to ensure that the software continues to satisfy user requirements. Maintenance is applicable to software developed using any software life cycle model (for example, spiral). The system changes due to corrective and non-corrective software actions. Maintenance must be performed in order to:
The maintainer's activities comprise four key characteristics, according to Pfleeger [Pfl01]:
Maintenance consumes a major share of software life cycle financial resources. A common perception of software maintenance is that it merely fixes faults. However, studies and surveys over the years have indicated that the majority, over 80%, of the software maintenance effort is used for non-corrective actions. [Abr93, Pig97, Pre01] Jones (Jon91) describes the way in which software maintenance managers often group enhancements and corrections together in their management reports. This inclusion of enhancement requests with problem reports contributes to some of the misconceptions regarding the high cost of corrections. Understanding the categories of software maintenance helps to understand the structure of software maintenance costs. Also, understanding the factors that influence the maintainability of a system can help to contain costs. Pfleeger [Pfl01] presents some of the technical and non-technical factors affecting software maintenance costs, as follows:
Lehman first addressed software maintenance and evolution of systems in 1969. Over a period of twenty years, his research led to the formulation of eight "Laws of Evolution". [Leh97] Key findings include the fact that maintenance is evolutionary developments, and that maintenance decisions are aided by understanding what happens to systems (and software) over time. Others state that maintenance is continued development, except that there is an extra input (or constraint)–existing large software is never complete and continues to evolve. As it evolves, it grows more complex unless some action is taken to reduce this complexity.
Since software demonstrates regular behavior and trends, these can be measured. Attempts to develop predictive models to estimate maintenance effort have been made, and, as a result, useful management tools have been developed. [Art88], (Bel72)
[Art88:c1s1.2; Lie78; Dor02:v1c9s1.5; IEEE1219-98:s3.1.1,s3.1.2,s3.1.7,A.1.7; ISO14764-99:s4.1,s4.3,s4.10, s4.11,s6.2; Pig97:c2s2.3]
Lientz & Swanson initially defined three categories of maintenance: corrective, adaptive, and perfective. [Lie78; IEEE1219-98] This definition was later updated in the Standard for Software Engineering-Software Maintenance, ISO/IEC 14764 to include four categories, as follows:
ISO/IEC 14764 classifies adaptive and perfective maintenance as enhancements. It also groups together the corrective and preventive maintenance categories into a correction category, as shown in Table 1. Preventive maintenance, the newest category, is most often performed on software products where safety is critical.
Table 1: Software maintenance categories
A number of key issues must be dealt with to ensure the effective maintenance of software. It is important to understand that software maintenance provides unique technical and management challenges for software engineers. Trying to find a fault in software containing 500K lines of code that the software engineer did not develop is a good example. Similarly, competing with software developers for resources is a constant battle. Planning for a future release, while coding the next release and sending out emergency patches for the current release, also creates a challenge. The following section presents some of the technical and management issues related to software maintenance. They have been grouped under the following topic headings:
Limited understanding refers to how quickly a software engineer can understand where to make a change or a correction in software which this individual did not develop. Research indicates that some 40% to 60% of the maintenance effort is devoted to understanding the software to be modified. Thus, the topic of software comprehension is of great interest to software engineers. Comprehension is more difficult in text-oriented representation, in source code, for example, where it is often difficult to trace the evolution of software through its releases/versions if changes are not documented and when the developers are not available to explain it, which is often the case. Thus, software engineers may initially have a limited understanding of the software, and much has to be done to remedy this.
The cost of repeating full testing on a major piece of software can be significant in terms of time and money. Regression testing, the selective retesting of a software or component to verify that the modifications have not caused unintended effects, is important to maintenance. As well, finding time to test is often difficult. There is also the challenge of coordinating tests when different members of the maintenance team are working on different problems at the same time. [Plf01] When software performs critical functions, it may be impossible to bring it offline to test. The Software Testing KA provides additional information and references on the matter in its sub-topic 2.2.6Regression testing.
Impact analysis describes how to conduct, cost effectively, a complete analysis of the impact of a change in existing software. Maintainers must possess an intimate knowledge of the software's structure and content [Pfl01]. They use that knowledge to perform impact analysis, which identifies all systems and software products affected by a software change request and develops an estimate of the resources needed to accomplish the change. [Art88] Additionally, the risk of making the change is determined. The change request, sometimes called a modification request (MR) and often called a problem report (PR), must first be analyzed and translated into software terms. [Dor02] It is performed after a change request enters the software configuration management process. Arthur [Art88] states that the objectives of impact analysis are:
The severity of a problem is often used to decide how and when a problem will be fixed. The software engineer then identifies the affected components. Several potential solutions are provided and then a recommendation is made as to the best course of action.
Software designed with maintainability in mind greatly facilitates impact analysis. More information can be found in the Software Configuration Management KA.
How does one promote and follow up on maintainability issues during development? The IEEE [IEEE610.12-90] defines maintainability as the ease with which software can be maintained, enhanced, adapted, or corrected to satisfy specified requirements. ISO/IEC defines maintainability as one of the quality characteristics (ISO9126-01).
Maintainability sub-characteristics must be specified, reviewed, and controlled during the software development activities in order to reduce maintenance costs. If this is done successfully, the maintainability of the software will improve. This is often difficult to achieve because the maintainability sub-characteristics are not an important focus during the software development process. The developers are preoccupied with many other things and often disregard the maintainer's requirements. This in turn can, and often does, result in a lack of system documentation, which is a leading cause of difficulties in program comprehension and impact analysis. It has also been observed that the presence of systematic and mature processes, techniques, and tools helps to enhance the maintainability of a system.
Organizational objectives describe how to demonstrate the return on investment of software maintenance activities. Bennett [Ben00] states that "initial software development is usually project-based, with a defined time scale and budget. The main emphasis is to deliver on time and within budget to meet user needs. In contrast, software maintenance often has the objective of extending the life of software for as long as possible. In addition, it may be driven by the need to meet user demand for software updates and enhancements. In both cases, the return on investment is much less clear, so that the view at senior management level is often of a major activity consuming significant resources with no clear quantifiable benefit for the organization."
Staffing refers to how to attract and keep software maintenance staff. Maintenance is often not viewed as glamorous work. Deklava provides a list of staffing-related problems based on survey data. [Dek92] As a result, software maintenance personnel are frequently viewed as "second-class citizens" (Lie81) and morale therefore suffers. [Dor02]
Software process is a set of activities, methods, practices, and transformations which people use to develop and maintain software and the associated products. [Pau93] At the process level, software maintenance activities share much in common with software development (for example, software configuration management is a crucial activity in both). [Ben00] Maintenance also requires several activities which are not found in software development (see section 3.2 on unique activities for details). These activities present challenges to management. [Dor02]
Organizational aspects describe how to identify which organization and/or function will be responsible for the maintenance of software. The team that develops the software is not necessarily assigned to maintain the software once it is operational.
In deciding where the software maintenance function will be located, software engineering organizations may, for example, stay with the original developer or go to a separate team (or maintainer). Often, the maintainer option is chosen to ensure that the software runs properly and evolves to satisfy changing user needs. Since there are many pros and cons to each of these options [Par86, Pig97], the decision should be made on a case-by-case basis. What is important is the delegation or assignment of the maintenance responsibility to a single group or person [Pig97], regardless of the organization's structure.
Outsourcing of maintenance is becoming a major industry. Large corporations are outsourcing entire portfolios of software systems, including software maintenance. More often, the outsourcing option is selected for less mission-critical software, as companies are unwilling to lose control of the software used in their core business. Carey (Car94) reports that some will outsource only if they can find ways of maintaining strategic control. However, control measures are hard to find. One of the major challenges for the outsourcers is to determine the scope of the maintenance services required and the contractual details. McCracken (McC02) states that 50% of outsourcers provide services without any clear service-level agreement. Outsourcing companies typically spend a number of months assessing the software before they will enter into a contractual relationship. [Dor02] Another challenge identified is the transition of the software to the outsourcer. [Pig97]
Software engineers must understand the different categories of software maintenance, discussed above, in order to address the question of estimating the cost of software maintenance. For planning purposes, estimating costs is an important aspect of software maintenance.
It was mentioned in sub-topic 2.1.3, Impact Analysis, that impact analysis identifies all systems and software products affected by a software change request and develops an estimate of the resources needed to accomplish that change. [Art88]
Maintenance cost estimates are affected by many technical and non-technical factors. ISO/IEC14764 states that "the two most popular approaches to estimating resources for software maintenance are the use of parametric models and the use of experience" [ISO14764-99:s7.4.1]. Most often, a combination of these is used.
Some work has been undertaken in applying parametric cost modeling to software maintenance. [Boe81, Ben00] Of significance is that data from past projects are needed in order to use the models. Jones [Jon98] discusses all aspects of estimating costs, including function points (IEEE14143.1-00), and provides a detailed chapter on maintenance estimation.
Experience, in the form of expert judgment (using the Delphi technique, for example), analogies, and a work breakdown structure, are several approaches which should be used to augment data from parametric models. Clearly the best approach to maintenance estimation is to combine empirical data and experience. These data should be provided as a result of a measurement program.
Grady and Caswell [Gra87] discuss establishing a corporate-wide software measurement program, in which software maintenance measurement forms and data collection are described. The Practical Software and Systems Measurement (PSM) project describes an issue-driven measurement process that is used by many organizations and is quite practical. [McG01]
There are software measures that are common to all endeavors, the following categories of which the Software Engineering Institute (SEI) has identified: size; effort; schedule; and quality. [Pig97] These measures constitute a good starting point for the maintainer. Discussion of process and product measurement is presented in the Software Engineering Process KA. The software measurement program is described in the Software Engineering Management KA.
Abran [Abr93] presents internal benchmarking techniques to compare different internal maintenance organizations. The maintainer must determine which measures are appropriate for the organization in question. [IEEE1219-98; ISO9126-01; Sta94] suggests measures which are more specific to software maintenance measurement programs. That list includes a number of measures for each of the four sub-characteristics of maintainability:
Certain measures of the maintainability of software can be obtained using available commercial tools. (Lag96; Apr00)
The Maintenance Process subarea provides references and standards used to implement the software maintenance process. The Maintenance Activities topic differentiates maintenance from development and shows its relationship to other software engineering activities.
The need for software engineering process is well documented. CMMI® models apply to software maintenance processes, and are similar to the developers' processes. [SEI01] Software Maintenance Capability Maturity models which address the unique processes of software maintenance are described in (Apr03, Nie02, Kaj01).
Maintenance processes provide needed activities and detailed inputs/outputs to those activities, and are described in software maintenance standards IEEE 1219 and ISO/IEC 14764.
The maintenance process model described in the Standard for Software Maintenance (IEEE1219) starts with the software maintenance effort during the post-delivery stage and discusses items such as planning for maintenance. That process is depicted in Figure 2.
Figure 2 The IEEE1219-98 Maintenance Process Activities
ISO/IEC 14764 [ISO14764-99] is an elaboration of the IEEE/EIA 12207.0-96 maintenance process. The activities of the ISO/IEC maintenance process are similar to those of the IEEE, except that they are aggregated a little differently. The maintenance process activities developed by ISO/IEC are shown in Figure 3.
Figure 3 ISO/IEC 14764-00 Software Maintenance Process
Each of the ISO/IEC 14764 primary software maintenance activities is further broken down into tasks, as follows.
Takang & Grubb [Tak97] provide a history of maintenance process models leading up to the development of the IEEE and ISO/IEC process models. Parikh [Par86] also gives a good overview of a generic maintenance process. Recently, agile methodologies have been emerging which promote light processes. This requirement emerges from the ever-increasing demand for fast turn-around of maintenance services. Some experiments with Extreme maintenance are presented in (Poo01).
As already noted, many maintenance activities are similar to those of software development. Maintainers perform analysis, design, coding, testing, and documentation. They must track requirements in their activities just as is done in development, and update documentation as baselines change. ISO/IEC14764 recommends that, when a maintainer refers to a similar development process, he must adapt it to meet his specific needs [ISO14764-99:s220.127.116.11, 2]. However, for software maintenance, some activities involve processes unique to software maintenance.
There are a number of processes, activities, and practices that are unique to software maintenance, for example:
Maintainers may also perform supporting activities, such as software maintenance planning, software configuration management, verification and validation, software quality assurance, reviews, audits, and user training.
Another supporting activity, maintainer training, is also needed. [Pig97; IEEE12207.0-96] (Kaj01)
3.2.3. Maintenance planning activity [IEEE1219-98:A.3; ISO14764-99:s7; ITI01; Pig97:c7,c8]
An important activity for software maintenance is planning, and maintainers must address the issues associated with a number of planning perspectives:
At the individual request level, planning is carried out during the impact analysis (refer to sub-topic 2.1.3 Impact Analysis for details). The release/version planning activity requires that the maintainer [ITI01]:
Whereas software development projects can typically last from some months to a few of years, the maintenance phase usually lasts for many years. Making estimates of resources is a key element of maintenance planning. Those resources should be included in the developers' project planning budgets. Software maintenance planning should begin with the decision to develop a new system and should consider quality objectives (IEEE1061-98). A concept document should be developed, followed by a maintenance plan.
The concept document for maintenance [ISO14764-99:s7.2] should address:
The next step is to develop a corresponding software maintenance plan. This plan should be prepared during software development, and should specify how users will request software modifications or report problems. Software maintenance planning [Pig97] is addressed in IEEE 1219 [IEEE1219-98] and ISO/IEC 14764. [ISO14764-99] ISO/IEC14764 provides guidelines for a maintenance plan.
Finally, at the highest level, the maintenance organization will have to conduct business planning activities (budgetary, financial, and human resources) just like all the other divisions of the organization. The management knowledge required to do so can be found in the Related Disciplines of Software Engineering chapter.
The IEEE Standard for Software Maintenance, IEEE 1219 [IEEE1219-98], describes software configuration management as a critical element of the maintenance process. Software configuration management procedures should provide for the verification, validation, and audit of each step required to identify, authorize, implement, and release the software product.
It is not sufficient to simply track Modification Requests or Problem Reports. The software product and any changes made to it must be controlled. This control is established by implementing and enforcing an approved software configuration management (SCM) process. The Software Configuration Management KA provides details of SCM and discusses the process by which software change requests are submitted, evaluated, and approved. SCM for software maintenance is different from SCM for software development in the number of small changes that must be controlled on operational software. The SCM process is implemented by developing and following a configuration management plan and operating procedures. Maintainers participate in Configuration Control Boards to determine the content of the next release/version.
It is not sufficient, either, to simply hope that increased quality will result from the maintenance of software. It must be planned and processes implemented to support the maintenance process. The activities and techniques for Software Quality Assurance (SQA), V&V, reviews, and audits must be selected in concert with all the other processes to achieve the desired level of quality. It is also recommended that the maintainer adapt the software development processes, techniques and deliverables, for instance testing documentation, and test results. [ISO14764-99]
More details can be found in the Software Quality KA.
This subarea introduces some of the generally accepted techniques used in software maintenance.
Programmers spend considerable time in reading and understanding programs in order to implement changes. Code browsers are key tools for program comprehension. Clear and concise documentation can aid in program comprehension.
Reengineering is defined as the examination and alteration of software to reconstitute it in a new form, and includes the subsequent implementation of the new form. Dorfman and Thayer [Dor02] state that reengineering is the most radical (and expensive) form of alteration. Others believe that reengineering can be used for minor changes. It is often not undertaken to improve maintainability, but to replace aging legacy software. Arnold [Arn92] provides a comprehensive compendium of topics, for example: concepts, tools and techniques, case studies, and risks and benefits associated with reengineering.
Reverse engineering is the process of analyzing software to identify the software's components and their interrelationships and to create representations of the software in another form or at higher levels of abstraction. Reverse engineering is passive; it does not change the software, or result in new software. Reverse engineering efforts produce call graphs and control flow graphs from source code. One type of reverse engineering is redocumentation. Another type is design recovery [Dor02]. Refactoring is program transformation which reorganizes a program without changing its behavior, and is a form of reverse engineering that seeks to improve program structure. (Fow99)
Finally, data reverse engineering has gained in importance over the last few years where logical schemas are recovered from physical databases. (Hen01)