An opportunity remains for Ethernet in carrier networks because the demand for bandwidth continues and bottlenecks remain in access networks, said Anand Parikh, VP of Business Development for Appian Communications. However, while Ethernet is suitable as a transport method for shared, and best-effort Internet services, it is still not ready to be the transport layer for multiservice TDM and packet traffic. Parikh observed that to do so would require significant and expensive MPLS upgrades throughout core the network before Ethernet could really handle this time-sensitive traffic. However, Ethernet makes a lot of sense as a service, Parikh said, especially since customers already understand Ethernet. He believes carriers are more interested in mission specific platforms to deliver this service, rather than a “god box” that requires a rebuild of the SONET infrastructure. This implies a “pay as you grow” model, where the lowest possible price of entry is a critical factor. The Metro Ethernet Forum has been working to define metrics for standard Ethernet services, covering such parameters as explicit rate guarantees, optional bursting, granularities, recovery times, etc. The Appian platform is deployed by NTT Communications in about 40 to 50 central offices for offering Ethernet VLAN services.
Taking the contrary point of view, Vivek Ragavan, CEO of Atrica, argued that a paradigm shift to multi-service Ethernet metro transport would happen. The changeover will not happen at the same pace for all carriers, but demand for packet services will drive Ethernet into the transport network. Why Ethernet in the metro? Apart from the growth in packet traffic, there is a demand for bandwidth flexibility and for a lower cost-of-ownership that optical Ethernet is very well suited to handle. Ragavan believes it will not be cost-effective in the long term to patch-up the existing network to deliver the services that customers want. Ethernet end-to-end, where service is integrated with transport, is not only simpler for the carrier to manage but also significantly reduces the number of network-facing interfaces and saves cost. Such a network would still be able to handle TDM services. The question is not whether it is an overlay or a greenfield deployment, said Ragavan, but what are the services that drive future revenues? Atrica believes the transition is already happening in Asia.
ATM-based ADSL networks in the US are losing to the cable modem networks by a two-to-one margin, said Roger Dorf, CEO of Celite Systems. On top of that, the “early adopter” phase of broadband rollout is over and further deployments will require price cuts. Dorf believes that US carriers will find it difficult to make DSL profitable at under $40 per month because they are currently spending $275 per subscriber for each ATM-based DSL activation, taking into account the DSLAM port cost, POTS splitters and filters and the CO engineering and installation costs. Celite is developing a “DSL headend” box that would sit in a neighborhood remote terminal and connect back to a DSLAM in the central office by bonding 8 ADSL circuits into a 40 Mbps trunk. The DSL headend could support 200 to 600 homes. On the subscriber side, the Celite box would provide rapidly provisioned IP/Ethernet connection over the installed copper phone lines. The company is promising much lower cost per subscriber at even modest take rates. Celite was founded in July 2001 and is based in Austin, Texas. Celite recently started shipping its product for carrier lab trials.
Both the incremental approach and radical re-build approach are likely to find their place in the market, said Alex Balkanski, General Partner at Benchmark Capital. He noted that the Asian market is far more interested in these new Ethernet approaches, while US carriers might have a little more life left in their networks. Long term, he remains optimistic that Ethernet’s cost of advantages will enable it to prevail.
It is a really difficult market, pointed out James Wei, General Partner and co-Founder of Worldview Technology Partners, especially because of the numerous bankrupt carriers that are liquidated their networks for pennies on the invested dollar. This means the equipment may be a very low percentage of the overall cost and the question needs to asked why not continue to deploy low-cost Cerent boxes and live with cheap SONET? A case in point is Cogent Communications, which has acquired extensive infrastructure from Allied Riser and other bankrupt carriers. Its approach is to use new tools for existing networks.
