Global equipment sales for Metro Ethernet will double between 2006 and 2010 until reaching US$ 18.8 billion as reported by Infonetics Research. And it is true that the technological progress, the possibility of unifying Ethernet as unique protocol from LAN (Local Area Network) to WAN (Wide Area Network) and MAN (Metropolitan Area Network) are fostering this improvement. But such union answers the need of telecom companies to be able to provide any kind of convergent contents quickly, and mainly, of joining devices to improve CAPEX with scale economies, to simplify the operation and reduce OPEX.

For an effective establishment, Metro Ethernet should address the demands of Internet connectivity; clear services between LAN networks, Virtual Private Networks (VPN) between LAN (L2VPN) both to point-to-point and multi-point to multi-point; extranets; VPN Frame Relay and ATM; connectivity to backup centers and to storage networks (SANs); metropolitan transportation (backhaul) and voice over IP. And there is where there are controversies.

The first problem appears in services rendered to big companies, that should be reallocated as from Metro Ethernet to share the infrastructure with residential subscribers, an issue that generates little resentment. Although it is true that private networks were not as such but shared, the difference is that now there will be no difference whatsoever between any public and private network in any of sections. Until the final disappearance, legacy protocols, such as ATM and Frame Relay, will also be mounted over the public network, although with other protocols underneath.

According to 2006 data collected by Vertical Group on United States, from US$ 32 billion generated by business data services, barely 5% belonged to Ethernet, when private lines provided 36%; Frame Relay, 26%; ATM, 8%; Internet dedicated accesses, 15% and dedicated IP services, 10%: A whole world, which will gradually migrate to Ethernet.

Three parts

Metro Ethernet’s scope might be better understood if the network is divided into three parts: first, telephony companies were reaching as long as they needed to take digitalization to the last copper line for the customer in order to improve delivery Internet speed through xDSL. Some niche operators have directly located a switch Ethernet near the customer to travel directly with that protocol over copper in the last mile. Moreover, towers with WiMax have been appearing progressively, as a competitive option or in the case of the incumbents, to improve coverage. Finally, fiber to the home (FTTH) experiences began with GPON or EPON; Ethernet and each access linked to a fiber optic ring that joins (bundles) nodes and grants redundancy to the set. For cable networks, the ring joins the CMTS, Cable Modem Termination System, which adds the voice over IP possibility (VoIP) and Internet to the video own functionality, through coaxial cable under the DOCSIS standard (Data Over Cable Service Interface Specifications) certified by Cable Labs. But the user, in all cases, receives an Ethernet port.

Cisco withdrew from DSLAM, GPON, EPON or PON markets, which it considers tending to become commodities. Instead, it holds small access switches and it is completing the purchase of Navini, WiMax provider. Alcatel holds peer-to-peer connections. Nortel does not compete in the last mile, although it does not reject the possibility of entering it as long as fiber optic options become general.

Vendors’ battle then lies in routers and switches, and in some extent, in the possibility of offering end-to-end solutions. But, greatest problem was posed by Nortel when launching a pure and hard Metro Ethernet for bundling layer, a variable: PBT (Provider Backbone Transport), technology using switches of lower cost (ISO model, layer 3), to solve what MPLS solves with routing in layer 3). So discussion confronting vendors is whether PBT is an option or not.

In the router segment, Cisco’s leadership is evident. According to Synergy Research data belonging to third quarter of 2007, the company holds 71.3% of incomes on bundling nodes; followed by Alcatel Lucent with 12% in a market over US$ 350 million. But Nortel believes published figures are inadequate and that it holds greater presence after the acquisition of Tasman Networks for US$ 95.5 million cash, held in February 2006.

In the border services, Juniper gains presence, with 17% of market share, a similar value to Alcatel-Lucent; but Cisco concentrates more than half of the total in a business of more than US$ 700 million. Nortel halted its BRAS (Broadband Remote Access Server) at the beginning of the year.

The border islands are joined through IP MPLS technology with the core of the network. In core routers, the market is mostly focused on two players: Cisco and Juniper. In WAN switches, multi-protocols (ATM, FR, IP) in 2006, according to the report from Dell Oro, from February 2007, the market is divided between Nortel and Alcatel-Lucent. This is the spot Nortel wants to recover with PBT, and the sport attacked by detractors of new initiative.

End-to-end

Alcatel Lucent's strength comes together with its wide products portfolio, a reality Nortel tries to break with its variant PBT. They believe it is hard to think MPLS might be adjusted to new functionalities and that it is more convenient to have an ad hoc standard. Criticism is focused on the fact that precisely PBT is not a standard by a kind of Ethernet proprietary and that it is not efficient to hold legacy protocols. It only enables to set point-to-point virtual private networks and not multi-points network. Moreover, they consider that it does not allow any other kind of tunnel but Ethernet, neither ATM, nor, X25, nor FR, nor PPP.

Nortel prefers simple IP PBT switches to replace MPLS. The other vendors defend IP and MPLS on top; and they argue that the following steps required by PBT, for example, to run ATM over Ethernet, is like going back ten years and run directly to ATM. PBT to run legacy protocols should use PB (Pseudo Wire), a functionality Nortel assures comes embedded in PBT itself.

Detractors point out that it also requires someone to supply the service from a centralized place, when MPLS can do so automatically. With the same criterion they argue that today all ATMs still operate over X.25 and it is enough, then why not being as we are now. And in this sense, with PBT, a parallel network will be needed, then it will not be a multi-service network, as MPLS is.

But, not just that. They also object that PBT will only be able to provide point-to-point links, and not multi-point, when IPTV precisely requires this kind of broadcasting. In Nortel, they say the solution is already planned and that in order to support it they say that the standard number allocated for this is 802.1aq. They say that discovering a fault in MPLS is hard but instead PBT has tools for a quick diagnosis. And they add, proving the importance, that MPLS Forum 2008 version will be called MPLS & Carrier Ethernet Forum, to include PBT.

Migration

With a software, by Ethernet the speed supplied to customer can be changed from 0 to 1 Gbps and to 10 Gbps over fiber, Exactly what carriers needed. Main players migrate to an overlay model and at the same time, they deploy points of services that are used as distribution nodes and from there fiber is wired in ring or star; though ring is preferred based on costs (shorter laying) and reliability by getting redundancy.

As FR and ATM did so over SDH, that same fiber is used for rings; where there is no laying at all, they mount Ethernet directly over fiber. Theory says that in places where Ethernet runs over SDH, sooner or later, it will be removed in order to avoid duplication of costs of equipment maintenance.

Then, the choice should be made between MPLS in the ring and PBT. IDC published a categorical document in which it advise against PBT, but Nortel made the controversial BT, though took them as a witness case.