This paper describes possible scenarios for the transmission infrastructure supporting the next generation network of the Belgian NREN, BELNET. Also realistic estimations for future traffic demand are described. The evaluation of the scenarios includes a cost-based as well as a functional comparison, indicating how the required L1, L2 and L3 services can be provided over the considered network scenarios and what kind of network protection or restoration can be expected. Investigation of the problem BELNET, the Belgian National Research and Education Network [1], supplies advanced internet services to Belgian universities, colleges, research centers and government services. The BELNET-network (GIGANET) was developed to promote research and training and to stimulate national and international scientific cooperation. Today the national network consists of two star shaped structures, centralized in Brussels, from where data transmission lines of 2.5 Gbps depart to each of the national PoPs. BELNET makes use of managed SDH services for the transmission lines with on top an IP routing layer managed by BELNET. This model has proven its resiliency and cost efficiency but there are some new emerging technologies and bandwidth requirements which could make it more cost-efficient to change to a new model. The considered study includes both the topology design (that should still connect all PoPs) as well as the choice of the appropriate underlying technology. In addition to the current IP (L3) connectivity, the next-generation BELNET network should be capable to provide a L2 and L1 service. The two BELNET core PoPs will be connected to the GEANT2 network (i.e. new network replacing the current pan-European research network, GEANT [2]) to provide an end-to-end L1 service and a L3 IP routing service. For this L1 service BELNET will be connected to an SDH DXC in the GEANT2 PoP via a number of wavelengths carrying STM-16/OC-48, STM-64/OC-192, GbE or 10 GbE signals. The switching granularity of this cross-connect is VC-4Xc/v. This paper elaborates on the considered network and demand scenarios as well as on suggested evaluation method used in the study. Specification of network scenarios The economical interest of several possible future network scenarios for BELNET will be evaluated. Taking into account the presence of multiple telecommunication providers on the small Belgian surface, there is probably no need for BELNET to acquire all physical equipment itself. The following scenarios will be considered: BELNET builds a network on dark fiber (acquisition of own fiber, Indefeasible Right of Use (IRU) or leasing of fiber), BELNET builds a network based on leased wavelengths, BELNET leases a managed SDH network from an external provider or BELNET leases a managed Ethernet network from an external provider. A schematic overview of these scenarios is given in Figure 1. Dark fiber Wavelength network SDH network Ethernet network L3 IP infrastructure L2 MPLS infrastructure supporting VPLS Ethernet infrastructure MPLS network with VPLS sand L1 PWE support L1 SDH DXC-based infrastructure OXC wavelength infrastructure Figure 1 network scenarios Both the dark fiber and the wavelengths network are based on a wavelength infrastructure in core and access PoPs. In Figure 2, a schematic overview of the ownership of the equipment is given. In a dark fiber network, the multiplexers and the transponders are owned by BELNET, whereas in a leased wavelength network the external service provider owns this transmission equipment. Figure 3 shows some important candidate configurations to implement the core PoPs. In the figure, we distinguish the fibers coming from the access PoPs and the fiber(s) to the other core PoP. We also distinguish a cross-connect (XC) that will be responsible for at least assembling the wavelengths to the GEANT2 network. This cross-connect will have direct access to all channels on the fibers from the access PoPs in the double star network configuration (parts b and d of Figure 3). In the other case (i.e., collector rings BroadBand Europe Bordeaux, France 12-14 December 2005 Paper W03B.02 – Verbrugge S. Page 2 of 5 network configuration) (parts a, c and e of Figure 3) each collector will most probably feature an (O)ADM, because this is the most suitable type of equipment deployed in ring networks: the tributary ports of the (O)ADMs are then fed into the cross-connect, before being switched into the wavelengths to the GEANT2 network. However, as shown in part b and d of the figure, the rings might also be terminated directly on the cross-connect avoiding the installation of the intermediate (O)ADMs. Although this might require that the XCs feature for example self healing ring protocols. In Figure 3, a distinction is also made between the OTN and the SDH scenario. The major difference between these scenarios is that in the OTN case an OXC is used and in the SDH case a DXC. The DXC gives as a major advantage that it is possible to groom the traffic (pack low capacity traffic in high capacity streams). This is not possible with an OXC.