New FIBRE testbed

The main objective of this project is to create a common space between the EU and Brazil for Future Internet (FI) experimental research into network infrastructure and distributed applications. Currently such facilities already are operated, or are being built following similar designs, by partners in this project from both sides of the Atlantic Ocean. We expect that such a space will enable and encourage closer and more extensive bilateral cooperation in FI research and experimentation, as well as strengthening the participation of both communities in the increasingly important global collaborations in this important area of network research and development.

In the last two decades networks, and especially the Internet, have become part of the critical infrastructure of governments, businesses, homes and schools. The current Internet architecture, initially designed about 30 years ago, has suffered many extensions in recent years, to include new functionalities, which were unforeseen in the original design. Many network experts now consider it is necessary to undertake the study of alternative architectures for the Future Internet as a truly effective way to resolve many of the pressing problems that currently afflict the Internet. Some of the disadvantages of continued persistence in the use of the current architecture include:

• Imminent exhaustion of the currently available space of IPv4 endpoint identifiers, causing a "balkanization" of the Internet, without true global connectivity;
• Increased costs of IP routers, due to the non-scalable nature of the internal routing tables, and of the performance requirements to process IP packet headers at line speed on very high-speed links, thus restraining network growth;
• Immense investments in palliative measures to counter such security problems as are currently caused by spam, denial of service and outright information crimes;
• Difficulties of combining access transparency and application performance for mobile users.

The adoption of an alternative architecture can alter this situation, and it is important to note that the pursuit of such alternatives by network researchers has already begun in several countries. However, one serious obstacle to effective adoption of such innovations has been the inability to validate them convincingly. The reduction in real-world impact of any given network innovation is due to the enormous installed base of equipment and protocols, and the reluctance to experiment with production traffic, which have created an exceedingly high barrier to entry for new ideas. Today, in major countries in the world, there is almost no practical way to experiment with new network protocols in sufficiently realistic settings to gain the confidence needed for their widespread deployment. The result is that most new ideas from the networking research community go untried and untested, leading to the commonly held belief that the Internet infrastructure has “ossified”.

Having recognized the problem, the network research community is developing alternative solutions for experimental FI research, using programmable testbed networks, such as those of GENI in the USA, AKARI in Japan and FIRE in the EU, and similar initiatives have also been launched more recently in other parts of the world. Close attention has been paid to the question of providing effective programmable network elements (routers and switches) at low cost. One approach uses “commercial off-theshelf” (COTS) hardware, in the form of Intel-based PCs. This implies a low performance network element, as the basic hardware is not at all optimised for high-performance I/O and packet switching. At Stanford University solutions have been sought to the problem of producing low-cost programmable network elements of acceptably high performance. An early proposal was NetFPGA, a high-performance I/O extension board for PCs, providing 4 Gigabit Ethernet ports (upgraded to 10G ports in more recent models). A more recent and significant contribution involves an architectural alteration to network element design, where high-performance switching hardware is combined with a table-based implementation of the control plane, which can easily be modified by the user, in this case, an experimental network designer. The resulting architecture, known as OpenFlow (OF), is designed as an extension to production network element design, and a growing number of switch and router manufacturers already sell OpenFlow-capable hardware, or have advanced plans to do so.