Wholesale Ethernet access and the “hairpin turn” dilemma

CEN Feature (Aug 10 2010) Wholesale / Exchange

1. 

   E-LAN services can be a great solution for organizations that have multiple locations and want to use Ethernet connectivity not just for dedicated Internet access, but also to extend the company’s local area network to support capabilities such as application sharing in the organization. But the more locations an organization wants to interconnect, the greater the likelihood that the carrier serving that organization will not have network facilities to every location, requiring the carrier to rely on a wholesale network operator for the last-mile link to serve certain sites.

    

   Increasingly, carriers are opting to use Ethernet network-to-network interfaces (E-NNI) to connect to wholesale Ethernet links for these last-mile connections to off-network locations. But when wholesale circuits are used to support E-LAN services, a unique challenge often arises—a challenge some people refer to as the “hairpin turn” problem.

    

   The problem arises when last-mile links for two or more customer locations are served out of the same wholesale carrier central office or co-location facility. As Mark Durrett, director of marketing for Overture Networks, noted recently in his blog “Wholesale Ethernet access – the E-NNI places three new requirements on the service provider,” this is not an uncommon situation.

    

   Durrett pointed to the example of a competitive local exchange carrier, which recently found that nearly 40 percent of its service orders required connectivity for E-LAN or E-line services to multiple sites served by the same co-location facility.

    

   With E-LAN service, these sites often need to communicate directly with one another. But if two of those sites are served from the same wholesale carrier central office, the traffic often must take a surprisingly circuitous route between the two locations. First it flows into the wholesale provider’s central office, where it is combined onto a single link – at the external network-to-network interface (E-NNI) – with traffic from multiple last-mile circuits and handed off to the competitive service provider, who then carries the traffic deeper into the network to its own Ethernet switch/router. At the switch/router, traffic destined for the second customer location is rerouted back to the central office, where it is handed back to the wholesale carrier which delivers it to the customer. It’s a bit like the French train system, where often the only way to get to a town a dozen or more kilometers away is to go into Paris and back out again.

    

   As Durrett explains, the reason this occurs relates to the edge aggregation switches that competitive carriers co-locate in the wholesale provider’s central office. Ideally these devices should be able to hand traffic destined for the second customer location right back to the wholesale carrier, rather than having to send it to their own core network first. The hitch is that because multiple last-mile links are aggregated onto the same physical interface at the ENNI handoff, the edge aggregation switch has to be able to send the traffic destined for the second customer location back out over the same port it came in on with a different S-VLAN tag. And some edge aggregation switches don’t have that “hairpin turn” capability.

    

   Competitive carriers planning to offer E-LAN services should take care when selecting an edge aggregation switch to choose one with that capability, Durrett says. And to do the job right, the edge aggregation switch also should support queuing and shaping on an individual Ethernet virtual circuit basis. This is required in order to ensure that each customer location receives the service quality that the customer signed up for.

    

   With an E-LAN service, each customer location can receive traffic from any other customer location, including those served completely over the competitive carrier’s facilities and those that rely on a last-mile link from the wholesale operator. At any moment in time, the total traffic destined for an individual customer location can easily exceed the data rate that the competitive carrier has ordered from the wholesale operator for the last-mile link. By using an edge aggregation switch that supports queuing and shaping on an individual Ethernet virtual circuit basis, the competitive carrier can maintain control of how traffic is prioritized. It can ensure it does not hand off more traffic destined to an individual customer than the wholesale carrier has committed to providing. If that were to happen, the wholesale operator would apply its own prioritization scheme to the traffic, most likely discarding traffic along the way that the competitive carrier would not have intended.

   Bookmark or Share this article

Login to comment.