Software Testing and

Implementation

Review

Testing
o
o
o
o
o
o

Test cases
White & black box testing
Unit testing
Integrated testing
System & acceptance testing
Implementation

Overview

Introduction.
Testing Techniques:
o

black-box & white-box testing.

Testing Strategies:
o

unit, integration, and system testing.

Introduction

The design phase produces:

o
o

module structure e.g. structure chart;

module specifications e.g. data structures and
algorithms (pseudo-code).

Every module in the design is implemented

according to its specification.
Objectives of the development phase:
o
o

to transform the design into program code;

to unit test and debug each module in isolation.

Development

Development is a relatively short phase in the

software lifecycle:
o

it is likely to become less and less important as

CASE tools improve.

Testing requires the most effort of all the

development phases:
o

in general, the largest number of staff will be

needed during testing;

Testing

Aim: to identify defects in a software product.

The input domain of most non-trivial software
is too large (often infinite) to test completely:
o

a system with just six inputs, each with ten

possible values, has 106 possible test cases!

It is usually impractical, if not impossible, to

test a software product exhaustively:
o

hence it is rarely possible to guarantee that

software is entirely free of defects!

Test Cases & Test Suites

Testing a program involves:

using a set of carefully designed test cases as
input to the program;
o analyzing the resulting output to determine
whether the program behaved as expected.
o

A test case is a triplet [I, S, O]:

o
o
o

I is the data to be input to the system;

S is the system state for which I will be input;
O is the expected output from the system.

A set of test cases is called a test suite.

Designing Test Cases

There are two complimentary

approaches to the design of test cases:
o
o

black-box testing;
white-box (or glass-box) testing.

Black-box testing is sometimes called

functional testing.
White-box testing is sometimes called
structural testing.

Black Box Testing

Black-box testing

Designing black box test cases:

o requires

knowledge of the
software specification;
o does not require knowledge of
the software design or
implementation.

Software Testing Strategies

Software is usually tested at several levels of

complexity:
Unit Testing test each module in isolation;
Integration Testing test a growing set of
integrated modules each time a new one is added;
o System Testing test the complete system once
integration is finished.
o
o

Why not just test the software once when

integration is complete?

Unit Testing

Testing each module in isolation (before

integration) reduces debugging effort:
testing can begin before all of the modules in the
system are ready;
o errors detected when several modules are tested
together will be difficult to locate.
o

Regression Testing: the module must be

completely re-tested when a bug is fixed.
o

Who: the module programmer?

Integration Testing

A software product begins to take shape as its

modules are integrated:
o

integration can begin as soon as two related

modules have been coded and unit tested.

Modules are integrated and tested according

to a well-defined plan, usually one at a time:
Regression Testing: the system must be
completely re-tested after each integration.
Who: an independent test team.

System Testing

System testing begins once all the

modules have been integrated
successfully.
There are several kinds of system
testing
Alpha Testing: users (supervised).
Beta Testing: users (unsupervised).
Stress Testing: the developers.
Acceptance Testing: the customer.

System Testing

Alpha Testing:
off-the-shelf software is tested by the developer
observing users at work;
o however, this is done in a controlled environment.
o

Beta Testing:

the software is then tested by (friendly?) users in

an uncontrolled environment i.e. field testing;
o the developer is usually not present.
o

Beta testing is the final live testing before the

software is officially released.

Stress Testing

Stress testing involves using abnormal inputs

to push a system to its limits.
The system is evaluated using performance
measures such as:
o
o
o

input and output quantity and frequency;

processing and response times;
bandwidth and memory utilization.

The system should be tested well beyond the

designed capacity.

Acceptance Testing

In contract-based projects acceptance testing

is carried out by the customer.
The aim is to determine whether the system
meets its specification:
o

this involves validating the system against the

original SRS document.

After successful system testing the customer

takes delivery and you finally get paid!

References

Essential Reading:
o Pressman Chapters 17 & 18.
Background Reading:
o Pfleeger Chapters 7, 8 & 9;
o Sommerville Chapters 19 & 20.

Implementation

When a new ICT system has been

developed, it is ready is be
implemented into the working
environment, replacing the original
system. There are four strategies that
can be used to do this:
1.direct
2.phased
3.pilot
4.parallel running

Direct implementation

Advantages:
o
o

The most rapid of all the strategies, provided it works!

Less risk of confusion between old and new systems.

Disadvantages:
Most stressful for the users - the old system has gone so there is no going back if they
have difficulties.
o Most difficult to train staff on as the new system was not in place to learn on before the
change over.
o Most stressful for the developers - all the data and files from the old system will have to be
converted ready for use on the new one.
o Most risky - if the new system does not work properly then there is nothing to fall back on.
o

PHASED implementation

Advantages:
o
o
o

Very structured, each phase can be fully evaluated before moving onto the next one.
Lower risk, a well planned and controlled introduction of the new system.
Easy to train staff by letting them learn new skills on each phase as it is introduced.

Disadvantages:
o
o

Slower than direct implementation.

Although each phase is easy to evaluate, you have to wait until all the phases are
complete before you can evaluate the whole change over.

PILOT implementation

Advantages:
o
o
o
o

Easy to control, the pilot can be halted at any time.

Easy to evaluate because the new and old systems are both running.
Low risk, if a small-scale pilot fails then not too much has been lost.
Easy to train staff by letting them learn new skills on the pilot system.

Disadvantages:
o
o

It can be slow to get a pilot to completely replace the old system.

A pilot may not show up problems that a full-scale implementation would. This is because a system can
work well as a small-scale pilot but has difficulties when it is scaled up to a full operating system with
more realistic volumes of data to be processed

PHASED implementation

Advantages:
o
o
o
o

If there are initial problems with the new system then the old one can still be used.
Both systems can easily be compared.
Easy to train staff by letting them learn new skills on the parallel system.
Easy to evaluate because the new and old systems are both running.

Disadvantages:
o
o

Expensive - both systems are being run as fully operating versions so both are doing the same job.
This may mean duplication of staff and hardware.
Some risk - there is a greater chance of confusion or errors if the two different systems are being run
side-by-side.