Library and Information Management

FTE Category A Staff Submitted: 50

% internationally recognised staff: 85%

% Early Career Researchers (ECR): 24%

Overall Quality Profile (percentage of research activity at each level)
Four Star Quality that is world-leading in terms of originality, significance and rigour.	20
Three Star Quality that is internationally excellent in terms of originality, significance and rigour but which nonetheless falls short of the highest standards of excellence.	30
Two star Quality that is recognised internationally in terms of originality, significance and rigour.	35
One star Quality that is recognised nationally in terms of originality, significance and rigour.	15

Robert Hierons is Professor of Computing in the School of Information Systems, Computing and Mathematics and a member of the Centre for Information and Knowledge Management (CIKM).

“We all know that software plays a crucial role in society and that software failures can lead to problems ranging from (potentially massive) financial loss to sensitive data being released and even injury or death. For example, global financial markets use software, many of us use online banking systems and our cars can contains tens of processors with software running on each processor.

This raises a crucial problem - how can we ensure that this software actually does what it should? Many techniques are used in order to tackle this problem testing is one of the most important of these. In testing we execute the system and check that the observed behaviour is acceptable (there are no failures).

Unfortunately, testing is highly complex and extremely expensive. In addition,
human/manual testing is error prone.

There has thus been much interest in automating testing by developing tools that will test the software for us. One of the most promising approaches is Model Based Testing (MBT) in which we produce a model of the aspect of the system we are testing and a tool automatically tests the system based on this model and then automatically checks the observed behaviour against the model. Test automation allows us to test software more efficiently and effectively, potentially leading to cheaper software that contains fewer faults.

I have been working on an MBT problem related to the testing of distributed systems. In testing such systems we typically have test runs that may involve long sequences of inputs to the system interspersed with resultant outputs provided by the system.

For example, when testing an internet banking system, a test tool might start by clicking on the login button (an input), expect to see a new screen (an output), enter a user name and a password (inputs) and then expect to see a resultant new screen (an output). At the end of the test run the tool checks that the sequence observed was one of the allowed/acceptable behaviours and this last step is usually relatively straightforward even though a test run can be quite long (hundreds or thousands of inputs and outputs).

Things change, however, if we are testing a system that interacts with its environment/users at several physically distributed locations and, for example, many web-based systems are like this. Here we observe sequences of interactions (inputs and outputs) at each interface and have to somehow check this set of sequences against the allowed behaviours. The problem is that we don’t actually know the (global) sequence of inputs and outputs that occurred, only a local sequence at each interface, but the required behaviours (defined by the model) are usually described as global sequences.

We could simply form all global sequences that are consistent with the local sequences we have observed and this approach works in principle. However, the number of such sequences can be huge even for a short test run and so this approach just does not scale to testing more complex systems or using longer test runs. I am currently investigating conditions under which this problem can be solved efficiently and developing practical solutions for such situations.”