Measurements & evaluation tools

End-user Internet access has traditionally been dominated by wired connectivity (ADSL, cable). This landscape is changing, with mobile broadband (3G/4G) complementing or fully replacing the wired access. Whereas wired network access links typically provide a stable link capacity and delay, mobile wireless access adds another level of complexity. Aspects such as coverage, radio signal strength, handover and resource sharing all contribute to a varying channel capacity, packet delivery delay and delay variations. Besides purely technical aspects, the user interacts and uses mobile devices in different ways than stationary computers, with apps, content sharing, and messaging.

The complexity of wireless and mobile systems therefore makes them hard to evaluate. Typically, four common approaches are used, analytical modeling, simulation, emulation, and live network experiments. Each of these approaches provides different trade offs with regards to the amount of detail considered, model validation requirements, degree of experimental control, reproducibility, and so on. In many cases the approaches are complementary and can provide different insights about the topic under study. While analytical modeling and simulations provide for a high repetitive and scaling of experiments, assumptions must be made with regards to user scenarios such as mobility patterns, application usage, and packet flow distributions. This is also true to a large extent for network emulation, where parts of the system is simulated, but interacts with real network or software components. Live network measurements encapsulates these complexities, at the cost of a less controllable and less repeatable experiment environment. In a sense, being able to use actual measured data as input to simulation and emulations will provide the best of both world, and allow for repeatable experiments with representative user behavior, network utilization and network conditions experienced by end-users. Then, in a feedback process, service optimizations can be screened in an emulator, and then validated in a live network. Possible differences in expected results can then be investigated to further refine the emulation to match reality.

Measurements are an integral part to support the research into scalable signalling support and service optimization. We have long term experience with measurement-based experimental network research, and have a collection of tools to collect and analyze such data. In cooperation with the industrial partners, we intend to learn of each others tools, construct new tools, and share experiences. So far, measurement data has typically been used in individual research project, but not shared over time or by researchers. We will in this project build and operate a data measurement repository where data is collected, described and can easily be shared.