FED4FIRE - EXPERIMENTING PERFORMANCE ISOLATION IN VIRTUAL NETWORKS

Fed4FIRE is an Integrating Project under the European Union’s Seventh Framework Programme (FP7) addressing the work programme topic Future Internet Research and Experimentation. The project is performed by a consortium of 17 partner organisations from 8 countries, and is coordinated by iMinds, Belgium. It started in October 2012 and will run for 48 months, until the end of September 2016.

CREATE-NET joined the FED4FIRE project in response to being successful within the second open call of FED4FIRE. A short summary of CREATE-NET activities in FED4FIRE is provided below:

Network virtualization sits firmly on the Internet evolutionary path allowing researchers to experiment with novel clean–slate designs over the production network and practitioners to manage multi–tenants infrastructures in a flexible and scalable manner. In such scenarios, isolation between virtual networks is often intended as purely logical: this is the case of address space isolation or flow space isolation. Logical isolation neglects the effect that network virtualization has on resource allocation network–wide. Preliminary empirical evidence that sub–optimal resource allocation may result from the utilization of state–of–the–art network virtualization technologies has already been provided by the experimenters. Such work leveraged mininet in order to emulate an OpenFlow-controller network composed of a few nodes. A practical on-line feedback mechanism aiming at providing each tenant with the necessary information about per–link available bandwidth has also been implemented and tested.

The contribution of this experiment is twofold. On the one hand, we aim at demonstrating the fundamental limitations of current network virtualization techniques in realistic setting by leveraging a physical infrastructure. On the other hand, we aim at providing an effective and efficient performance isolation solution for virtualized multi-tenants facilities. It is worth noticing that the proposed solution will effectively allow users of a Fed4FIRE-like facility to have a virtual infrastructure that is isolated at the logical and performance level from the other tenants. From the academic standpoint this will allow for reproducible experiments. Similarly, from the commercial viewpoint, an effective performance isolation solution could pave the way to new business models, e.g. virtual data-centres operators. The experiment will also build on top of our previous work w.r.t. the algorithmic approach to be used in providing feedback to the various tenants. This experiment will leverage the two OpenFlow Ofelia islands provided by Fed4FIRE (i2Cat and UBristol) for a medium-scale experiments on real equipment, and the VirtualWall testbed provided by iMinds for large-scale experiments on emulated machines. The outcomes of the proposed experiment will be beneficial to both the Fed4FIRE community and in general to data center operators. The former will effectively benefit from a better characterization of the performance isolation level that can be expected in a virtualized facility, while the latter will be provided with a proof-of-concept implementation of a practical performance isolation scheme for multi-tenants networks. It is worth noticing that the performance isolation scheme could be integrated within the Fed4FIRE virtualization layer providing experimenters with an effective technology for replicable experiments. Finally, providing deterministic performance isolation represents a key milestone toward and effective monetization and long term sustainability of a FIRE facility.