SOFTWARE MAINTENANCE 24 March 2013 William W. McMillan.

28
SOFTWARE MAINTENANCE 24 March 2013 William W. McMillan

Transcript of SOFTWARE MAINTENANCE 24 March 2013 William W. McMillan.

SOFTWARE MAINTENANCE

24 March 2013 William W. McMillan

Who wants to do software maintenance?

If this is not your cup of tea, why not?

Software “Evolution”

Term is a good one. But can be confused with “evolutionary

development.” Focus here is on what happens to

software after initial delivery. Yes, it evolves, but only after it was

created intentionally. “Maintenance” has been most common

term.

Most Software Effort is Here

Systems may be around for decades. Adding functional increments much

harder in older systems than in brand new.

Have to deal with loss of: Tools and platforms Knowledge (developers, management,

users) Ways of working (e.g. plan-based to agile) Conducive environment (leadership, access

to users, team energy,…)

What difficulties have you had in changing software after initial creation?

Kinds of Maintenance (Old)

Corrective (~ 20%) Fixing errors

Adaptive (up to 25%) New software or hardware environment

“Perfective” (50% +) Enhancements, changes in requirements

Preventative (~ 0%) Cleaning house, tidying up, documenting

Kinds of Maintenance (Newer)

Fault repair (17%) Fixing errors

Environmental adaptation (18%) New software or hardware environment

Functionality addition, modification (65%) Enhancements, changes in requirements

Preventative (~ 0%) Cleaning house, tidying up, documenting

What could be done to encourage preventative maintenance?

The Payoff for Good Engineering Most of software engineering supports

maintenance. More effort early leads to easier

maintenance. One estimate:

over 80% of maintenance effort caused by incorrect requirements and specs.

Key in getting management support for software engineering.

Cost of Maintenance

Proportion of code changed per year times original cost of system.

Annual maintenance effort = annual change traffic * software development time

AME = ACT * SDT (in person-months) Oversimplification – need to consider all

other COCOMO factors.

What Has To Be Maintained

Code. Databases. Documents:

Requirements, user manuals, help files, design documents.

Installations: Workstations, mobile devices, embedded systems.

V & V records, test data sets. Archives

Past system configurations, management docs, changes.

Why is maintenance hard?

Challenges in Maintenance

1. Hard to get resources for good software engineering to begin with.

2. Hard to get resources to maintain old stuff.

3. Lose knowledge of system and domain expertise.

4. People don’t like to do it.5. … ?

Lehman & Belady’s “Laws”

Continuing change Nothing in software is static: OSs, needs,

delivery modes… Increasing complexity

When has a system ever gotten simpler? Continuing growth

Need growing functionality to satisfy users and clients.

Declining quality Counter-examples?

Lehman & Belady’s “Laws”

Large program evolution is “self-regulating” Can’t push change beyond a natural limit.

Conservation of familiarity Systems maintain their basic character … any examples of those that change

drastically?

Any other software “laws” you can think of?

Software Re-Engineering

Reimplementation of major parts or all. Legacy systems have old..

languages, hardware, OSs, coding practices, algorithms, interfaces

Over 1011 lines of code in existence. Heavy reliance on old systems.

Software Re-Engineering

Understand intent of present system. Documents, domain, context of use

Recover architecture and abstractions. Use reverse engineering tools. Find interactions, dependencies,

coupling, redundancies. Understand algorithms.

Especially if speed, memory use or similar at issue.

Interviewing people is critical (if they’re around!).

Software Re-Engineering

Re-implement (part or whole). Translate to new programming language

(auto?). Refactor, restructure system. Improve algorithms. Add exception handling. Improve security.

Software Re-Engineering

Switch paradigm, say to object-oriented or functional.

Improve robustness and quality (reduce coupling, better style, etc.).

Improve V & V. Update documents.

Have you ever had to recover software architecture from code alone?

Data Re-Engineering

Newer file formats, directory structures. Change field types, sizes, etc. Improve numeric precision. Improve validation rules. Establish or change default values. Change units of measurement. Implement new data model (e.g.,

relational). Modify database keys & indexing. Normalize database.

What data re-engineering has been necessary in recent years for many systems?

User Experience Re-Engineering Beyond incremental requirements

enhancement. Changes in platform, user interface,

deployment, connections among users. Examples? Challenges? Advice?

Guidelines for SW Maintenance

1. Expect it, plan for it.2. Use good people and pay them well.3. Understand the system thoroughly:

Architecture, algorithms, V & V done, domain.

4. Isolate changes, change minimally.5. Use regression testing.

Guidelines for SW Maintenance6. Update documents.

Requirements, specs, V & V records

7. Control and release versions carefully.8. Allow user input via bug reports,

change requests.9. Use software tools to support the

process.

Other maintenance guidelines?

Exercise

Your software engineering group is called in to help a company evaluate its software maintenance operation. What six to 12 indicators would you look at to assess the state of their operation? If you can, give quantitative measures you would like to see.