In SOMETIME we investigated the addition of Available Bandwidth (ABw) estimation to the set of metrics collected by MONROE. We have investigated implementations using active measurements and leveraging the Software Defined Network (SDN) paradigm, to consider interference with node-local processes (a more realistic scenario compared with mutually exclusive measurements), and to mitigate such interference.
ABw can be considered as the spare capacity on a network path, and is an important metric with many applications: optimized routing, adaptive encoding, just to name the most common. It is related to other path capacity metrics, such as TCP Bulk Transfer Capacity, but is different from them, with different measurement procedures and practical applications. Specifically, ABw provides an upper bound to available network capacity, dependent on network infrastructure and usage, but independent from the transport protocol and the application. This makes ABw an highly valued metric to characterize operational conditions of a network path, useful to inform at the same time different kinds of applications without per-application tests. As a counterbalance of its importance, ABw estimation is not trivial, especially in wired-cum-wireless scenarios (like MONROE testbed), that also pose constraints related to volume-based billing, and capping for cellular data connections.
Real-world applications mimicked by experiments in MONROE (e.g. multimedia streaming to smartphones, tablets, in-vehicle-infotainment systems) present the additional challenge of possible sharing of computing and communication resources during the measurements: to emulate, assess and mitigate this interference we adopted an SDN-based approach to manage traffic in the mobile terminal (MONROE node).
In our project we evaluated several available bandwidth estimation tools, analyzing their performance in experiments leveraging the MONROE platform, also evaluating the impact of virtualization (both lightweight and full). This experimental study has led to a selection of more reliable tools for the challenging MONROE scenario, and to reimplementation and modifications to existing state-of-art tools (Pathload and Yaz, respectively).
We run measurements managed by an SDN switch (OVS) running in the same environment of the experiment active tool, with also a local OpenFlow controller (Ryu), and validated the feasibility of the approach on current MONROE HW and SW configuration. We released the software as open-source and the experimental data as open-data, we published our findings in international conferences and journals specifically focused on network measurements. The MONROE project, its platform, and our activities in the Open Call have been presented in Master courses in computer science at University of Napoli Federico II. One interdisciplinary departmental research project has been initiated partially based on the findings of our SOMETIME project, and one national-level research project proposal has been presented in collaboration with other participants to the MONROE OC1, on themes derived from our experiences in MONROE.
Further research activities have been inspired by our SOMETIME project, and subject to funding availability, we plan to keep using the MONROE platform as regular users for future experiments.
- Giuseppe Aceto, Fabio Palumbo, Valerio Persico, Antonio Pescapè: An experimental evaluation of the impact of heterogeneous scenarios and virtualization on the available bandwidth estimation tools, IEEE International Workshop on Measurement and Networking (M&N), 2017, Napoli, Italy
- Giuseppe Aceto, Valerio Persico, Antonio Pescapè, Giorgio Ventre: SOMETIME: SOftware defined network-based Available Bandwidth MEasuremenT In MONROE IEEE/IFIP Workshop on Mobile Network Measurement (MNM’17), June 20th, Dublin, Ireland
- Péter Megyesi, Alessio Botta, Giuseppe Aceto, Antonio Pescapè, Sándor Molnár: “Challenges and solution for measuring Available Bandwidth in Software Defined Networks”, Computer Communications (2016)
Available bandwidth vs. Achievable throughput in MBB networks
Modified version of Yaz
Traffic Research Group: