The objective of WP2 is to create and operate FIBRE-BR, a Future Internet (FI) testbed network in Brazil that will allow multiple researchers and developers simultaneously to experiment with new applications and to control the network itself dynamically. This facility should be highly automated, instrumented and easy to use and manage. FIBRE-BR will be initially deployed as nine (9) geographically distributed institutional facilities (IF) connected by long-distance network connections that will later on be integrated to form a single wide-area FI testbed facility. Each IF will have a minimum common infrastructure, composed by a rack including a number of preassembled 1U servers with NetFPGA board(s), one or more industrial grade OpenFlow switch(es), a number of Orbit nodes and a powerful server machine to host the control and monitoring frameworks. Besides, each facility will include an additional set of specific equipment and instrumentation depending on

the classes of experiment (Use case) it will be associated with. Thus, we will organize the facilities in three types:

• Type 1 will target mobile wireless experiments and applications and will have, in addition to the common OpenFlow enabler and WiFi access point providing spatial coverage for experimental wireless communication, laptops and/or programmable handheld devices with WiFi and Bluetooth (BT) interfaces, plus instrumentation for traffic generation and analysis and mobility emulation. At least one facility of type 1 will also include a WiMAX base station and various multi-interface laptops and/or handheld devices (WiMAX, WiFi and BT enabled)

• Type 2 will target wired networking experiments and applications and will have, in addition a multicast-enabled optical reconfigurable WDM infrastructure plus content server(s) to stream high definition video and a projection system to display high definition streams

• Type 3 will target wireless/wired integrated experiments and applications and will include features from both the other two types of facilities.

This organization is not intended to limit the facilities, but to help in task attribution and budget planning. Furthermore, by means of interoperation of facilities of type 1 and type 2 it will be possible to support wireless/wired integrated experiments and applications as well. OpenFlow will be the common substrate of all nine institutional facilities in Brazil, providing a standard way for dynamically programming flow tables in the different forwarding technologies. This means that forwarding-related configurations and statistics in the network devices should be preferably done via OpenFlow protocol. This implies that all network devices should be OpenFlow-enabled and, hence, that porting of OpenFlow to network devices that currently do not understand OpenFlow will be required. And this porting of OpenFlow might require extensions to the OpenFlow protocol specification and, consequently, to FlowVisor and to NOX. FlowVisor will be used as the network virtualization layer regardless of the network technology underneath, allowing for the physical network to be sliced by the control framework, and for each slice to be controlled by the OpenFlow controller(s) associated with the corresponding experiment. Although there are a number of OpenFlow controllers available today, NOX is the strongest candidate because it is open source and its programmatic interface is general enough to support a large variety of network management applications, translating the abstractions into OpenFlow protocol messages.

All facilities will also adopt a control and monitoring framework (CMF) that will allow authorised and authenticated experimenters to allocate resources, run experiments and collect measurements in a given facility and across facilities. Each facility will run its own instance of the CMF. Nevertheless, because of the heterogeneity of the facilities, it is possible that there will be a particular control and monitoring framework technology for each type of experiment.

There are a number of CMF implementations already available that are potential candidates to serve as a basis for the development of the CMFs of the Brazilian facilities. A strong candidate is OMF, which has already been in use for some years and was designed to work with computing, storage and communications resources in general, although its deployment domain primarily encompasses testbeds and technologies in the access network, or what is often referred to as the mobile & wireless edge, as well as home gateways.

Another candidate is based on OCF (Expedient plus aggregate managers), since this solution has been adopted by OFELIA. This solution is already in pre-production use, and is being extended to support new features and hardware. The class of type 3 experiment deserve special attention as there is no control framework today dealing with a mix of wireless and wired resources. For these facilities involving a mix of wireless and wired resources, a candidate framework could be that based on on-going work in NICTA in Australia and in GENI (Global Environment for Network Innovations) in the USA to combine the OMF CMF with the OpenFlow ecosystem.

As regards the monitoring aspects of the CMF are concerned, a strong candidate is the ORBIT measurement library (OML), which enables distributed real-time collection of data in a large-scale testbed environment, which can also be contextualized per experiment. While OML has been developed in the context of OMF, it has been spun out as an independent project and has already been deployed in non-OMF testbeds. It is also now being increasingly adopted for monitoring operational network and service deployments. Hence, if different institutional testbed facilities adopt control frameworks other than OMF, this is already covered by this more agnostic approach. On the other hand, perfSONAR, a middleware for monitoring tools, is a suitable candidate to interoperate and federate the monitoring information across facilities, both in Brazil and the EU, since it provides standard formats for exchanging data measurements, based on Open Grid Forum (OGF) recommendations. A similar approach is being planned for use in ProtoGENI (a GENI control framework) by the perfSONAR development team.