Topic 7 Software Development and Reliability

Software development

When a feasibility study is being carried out to decide whether an ICT system is warranted, software comes in at an early stage.  There are a number of different approaches that we will look at:

• The software is written by the end user.  This is viable for a small project, probably carried out by an enthusiastic employee who has good computer skills.  They know the precise requirements of the department and the software can be developed very quickly.  In effect the employee does something very similar to your Module 6 Project.  However there is often little documentation.  If that employee leaves, then it is quite likely that the software will rapidly fall into disuse.  This approach is suitable for projects with only a limited lifespan.
• Writing the software in-house.  Large organisations have many specialist computer staff whose job it is to maintain the existing software and to develop new software.  The advantage of this is that any confidential information is kept in-house and does not risk getting into the wrong hands, as might be the case if external consultants were used.  The disadvantage is that specialists may need to be recruited on short-term contracts.
• Getting in consultants.  This is a more expensive option.  A software house will design, construct, test, and implement the software.  Quotations may be sought from several different companies, but cost should not be the only consideration.  If the project is complex, there may well be many pitfalls.  A good track-record in implementing complex systems may well justify the extra money.

Question 1  What criteria would a company use in deciding between several different software houses when a new system is going to be delivered?  ANSWER

The alternative to solutions devised in-house or by a software house is to buy a package.  It is a safe and relatively less expensive way  of buying software compared to a bespoke solution:

• The software is much less expensive, several hundred euros compared to hundreds of thousands of euros for a bespoke solution;
• It is immediately available and thoroughly tested, so there are likely to be fewer bugs in it;
• There will be extensive documentation (often a .pdf file), tutorials, or third-party training;
• Technical support is available from a website or a phone line (often premium rate);
• Other users will be able to give views about the package;
• Regular upgrades may be available.

There are, of course, drawbacks:

• The package may not do exactly what the organisation wants it to do;
• It may not run on the hardware;
• It may not be compatible with other software in the organisation.

It may be possible to lease the software, in which case regular upgrades get made available.  The initial expense is less, but it may be more expensive in the long run.

Some organisations try to modify proprietary software, but this can be dangerous; the functionality may well be reduced, or worse...

Criteria for Selection of a Solution

The following criteria can be applied:

• Functionality - how well the software will perform key tasks.  We have seen how weightings can be applied, and a numerical ranking can be produced for several different packages;
• Robustness - error free, handling large amounts of data without crashing;
• Performance - how quickly queries, etc. are handled;
• Support - both initial training, and support to deal with any problems that do arise;
• Portability - how well the files can be used by different applications;
• Transferability - how well existing data are transferred to the new system without having to type it all in again;
• Security - to protect sensitive data;
• Upgradability - the package will be in use for a long time, so it will need upgrading from time to time.

Question 2 A large market research company is considering several different software packages to assist with the analysis of data that are collected on behalf of clients.  Give three criteria that should be considered when evaluating software packages.  For each criterion, explain why it may be important to the company.  (AQA Past Question) ANSWER

Software Testing

All software is tested rigorously before it's released.  You have done testing as an important part of your projects (you did, didn't you?).  You may remember using normal data, boundary data, out of range data, or invalid data.  There are five key stages:

1. Unit testing - each component (subroutine, code) is tested.
2. Module testing - the module is made up of several components that are put together.  These are tested.
3. Subsystem testing - collections of modules are tested.  The Purchase Order system is a subsystem of the Accounts system.
4. System testing - testing all the sub-systems put together to make up the whole system.  The interaction of the subsystems may produce unpredicted results.  Also the testing is carried out to ensure that it matches all the requirements in the specification.  Do you remember doing that in your project?
5. Acceptance testing - as installed at the user's site.  Rather than dummy data, real data from the user are used.

Testing is an iterative process whereby results at each stage can be used to inform earlier stages (like the Waterfall Model).  For example a problem at Module Testing may show that one of the units needs to be re-thought.

Acceptance testing is sometimes called Alpha (a) testing.  Alpha testing reveals errors and omissions.  The user may well find that the system does not do various functions because they were not explained clearly enough in the original specification.

Beta (b) testing is carried out before a package is released for general sale.  The package is given to a number of different end users for testing in real environments.  Errors get shown up and these can be corrected before the product is released.

The more complex software is, the more likely it is to contain errors.  If you look at a good A2 project, you will see how much testing is done for even a relatively simple project.  A2 projects are simple compared to large corporate databases.  Almost all software contains errors which may well not get shown up in testing:

• Desire to be the first on the market with a product may lead to skimping on the testing;
• There may be a deadline for the completion, so not all the tests are done;
• The test plan or data may have been inadequate.  Some aspects may not even be tested at all.
• The user may well use the software in a manner not anticipated by the developer, therefore not tested.

Question 3  Explain what is meant by alpha and beta testing.  In your answer explain which type of testing applies to bespoke and proprietary software.  ANSWER

Question 4  Some bespoke software is tested by the software house, but when it is installed at the client's site, its functionality is poor.  Give two reasons why this might have happened.  ANSWER

Software Maintenance

Software needs to be maintained:

• Errors may well be found;
• Requirements will change as needs change;
• Hardware developments will allow software developments;
• New legislation may alter the requirements that the software runs under.

There are three main categories of maintenance:

• Corrective maintenance - gets rid of bugs that might take time to show up.  Often this is done by downloading corrective patches from websites;
• Adaptive maintenance - systems may be adapted to changing requirements, for example changing single-user software to multi-user software.  A new operating system may require some adaptive maintenance to be done;
• Perfective maintenance - the system can be made to work even better without changing its functionality.  For example reports can be made in a clearer format.

Generally the maintenance time is about 65 % perfective, 18 % adaptive, and 17 % corrective.

The manufacturers of software respond to feedback from users, and will carry out modifications before releasing a later version (a release).  Minor changes are shown by Versions 3.0, 3.1, etc.  Major changes are shown in Versions 4.0, 5.0, etc.  Or even a completely new name.

Maintenance is rather looked down on by developers, so ends up in the hands of less experienced staff, which is why a high proportion of software costs goes on maintenance.  In large systems, maintenance can lead to other errors, which necessitate more changes.  Sometimes it's not worth the trouble.

Generally maintenance is one of the highest costs of running software, so it is well worth-while to make the systems as easy as possible to maintain.

Question 5  State three different types of maintenance.  Explain why maintenance is such a high proportion of the cost of running software.  ANSWER

Now try the Topic Quiz