REPOSITORY > RESULTS

Abstract

Wireless communications are subject to rapid development cycles and ever-increasing application
and user requirements. The next generation of mobile networks is thus expected
to support industrial and traffic safety applications, enhanced multimedia applications and
a large number of connected meters, sensors and other devices that will form living and
working environments in the future. In order to meet the requirements of all these applications,
mobile networks will accommodate both human-type as well as machine-type
communications which will be enhanced through the inclusion of non-cellular short-range
and low-power wide area technologies in the form of capillary networks. This will result in
a multi-technology, high-interference radio environment that can already be found in some
frequency bands in dense urban centers.
Wireless communication systems have been traditionally implemented straight into silicon
chips leaving little opportunity for reconfigurability. The main reasons for this were
the requirements for high power efficiency and strict real-time operation. With the recent
advances in hardware platforms, however, we are now witnessing a trend towards
softwarization of radio functions in wireless communications. This approach enables the
design of highly adaptable and reconfigurable radio platforms as well as rapid prototyping.
However, in the case of wireless embedded devices which are often characterized by
restricted capabilities, this trend has yet to catch up and go beyond prototyping.
Software development and testing is a complex process involving highly skilled people
and dedicated infrastructure. A mix of organizational practices, development practices
and infrastructure contributes to the efficiency of the overall process which is relatively
well understood for general purpose development. However, the development of wireless
embedded firmware and its testing in a real environment is notably more challenging because
the code needs to be uploaded and tested on target embedded hardware, requiring
purposely developed testing infrastructure that needs to be optimally dimensioned.
In this thesis we investigated the dimensioning of embedded infrastructure as a function
of developer activity to ensure predictable test completion times and reliably quantify the
infrastructure size. By modelling developers activity, we gave better insights into the as
yet unstudied part of the current development practices. With such knowledge, a reliable
capacity dimensioning of supporting infrastructure is possible.
We further proposed a framework that can be used by wireless technology developers
to enable continuous integration practices in their testbed infrastructure. We provided a
proof-of-concept reference architecture and implementation of the framework for controlled
testing of multi-technology wireless networks which blends web service technology and
operating system virtualization technologies with emerging internet of things technologies
enabling continuous-integration practices for wireless embedded development and testing.
Finally, we performed the evaluation and testbed dimensioning for different sizes of
development teams. In particular, our capacity dimensioning simulation shows that for a
team of 10 developers, only 1-2 parallel embedded test environments are needed to ensure
test completion in less than ten minutes in 95% of the cases.