Ethernet services have fast emerged as one of the hot growth areas for carriers, second only to IP-VPN as a corporate WAN solution as both technologies supplant frame relay and ATM as the managed data solutions of choice. Analyst firm Ovum is projecting the overall global Ethernet market to grow at a 14% CAGR to $40 billion over the next five years, with multipoint services pegged as the fastest growing sub-segment at 32% CAGR to 2015.
But one of the challenges of offering Ethernet services is the reach of the operator's network, which is an issue because Ethernet services are often sold with SLAs and class-of-service guarantees. For all the talk of Ethernet being a standardized technology, the reality is that operators typically tweak the specs to cope with packet-based issues like delay, jitter and packet loss, and define their SLAs and CoS guidelines accordingly. That's fine for Ethernet traffic that stays on your own network, but disastrous if the customer needs the service via a PoP on someone else's network. Unless each carrier's Ethernet CoS support is aligned, it's impossible to enforce SLAs or guarantee CoS once it goes off the network. And until recently, there was no standard way to off Ethernet traffic from one network to another via an Ethernet network-to-network interface (E-NNI), forcing carriers to negotiate NNIs on a bilateral basis.
That changed in April this year with an initial set of E-NNI standards released by the Metro Ethernet Forum. What's also changed is that carriers have another option for extending Ethernet services outside of their networks: Ethernet exchanges.
The past year has seen a sudden flood of Ethernet exchange services which, as the term suggests, provide interconnection for Ethernet services in an exchange format, similar to bandwidth exchanges. Equinix launched the first carrier-neutral Ethernet exchange service in North America in October 2009, quickly followed by several other players, including CENX, Telx and Neutral Tandem. Now Ethernet exchanges are poised to become a $674 million business worldwide in 2014, according to Vertical Systems Group.
And while most of the Ethernet exchange activity has initially been centered around North America and Europe, it's already starting to expand into Asia-Pacific. Equinix hosts an Ethernet exchange in Tokyo, and plans to launch another exchange in Hong Kong in Q2 next year. CENX opened its first Ethernet exchange in Hong Kong last month to serve as a regional Ethernet interconnect hub.
However, the level of competition also means carriers have to look more closely at different Ethernet exchange offerings to make sure their needs are met. And needs for Ethernet services ?as well as the requirements of customers who buy them ?vary widely enough that, for some carriers, Ethernet exchanges will only be able to do so much.
While Ethernet exchanges all do the same basic thing - interconnect Ethernet services in a way that allows customers to come in and shop around for providers to hand off their Ethernet services traffic to ?they don't all employ the same business model.
Equinix, for example, employs a relatively straightforward private interconnect model, says Chris Sharp, Equinix's VP of interconnection and services innovation.
"We've set up a template that defines and maps all the service attributes of that Ethernet offering, put them on the switch fabric and then carriers have the opportunity to opt in and interconnect with all of those partners," he explains. "Then you can look at a portal, upload your infrastructure building lists and where you have facilities, then that portal acts as a wholesale channel where you can sell your services or buy them efficiently. So they can test new markets much quicker with this portal and connect outside of their home territories in a more ad hoc fashion."
One thing Equinix doesn't do is get involved with the commercial side of Ethernet services as a reseller, he says. "When we first looked at doing this, we brought in seven Tier 1 carriers with our technology guys and asked them if Equinix was credible in offering this type of service, and would they endorse that? The answer was a resounding yes, but one where they expected us to sit down and do the fundamental switch fabric but not get involved in the commercial side."
In fact, he adds, Equinix takes that concept to the point of not even designing its Ethernet exchange as a profit center. "We're not really trying to generate revenue off the switch fabric or the service. We're just trying to cover our costs for the actual development and deployment of it, because our core offering is space and power."
CENX is much more profit-oriented in its Ethernet exchange offering, but otherwise takes a similar approach, says Eric Gillenwater, worldwide development VP at CENX. "The technical component is done by the carriers, and the CENX model supports the technical connectivity, but not the commercial parts."
The reason to stay out of the commercial segment is, simply, to avoid competing against their own carrier customers coming into the exchange, Gillenwater says. "Some exchanges are targeting enterprises with competing services, but we don't want to compete with our customers, which are the carriers."
Telx, by contrast, is firmly in the commercial segment, treating Ethernet interconnect as "another arrow in the quiver" of its portfolio of interconnect services that it already sells to vertical enterprises as well as carriers.
"We do interconnect for direct fiber cross-connect, Cat 5, DSX-3, Sonet/SDH, and so on," says Glenn Calafati, VP and general manager of Ethernet Exchange services at Telx. "Ethernet is another interconnect product we offer, so basically we're a one-stop shop for all those kinds of products."
That approach, he says, allows customers to mix and match connectivity options to match the applications that need to run over the WAN.
"A good example is media and entertainment companies running HD video and SDI-type protocols to interconnect, which has nothing to do with Ethernet ?it's still connected over fiber or coaxial cable," Calafati explains. "If you do need a service that's a modest bandwidth size, or you want a flexible scalable technology to distribute it, you can use our Ethernet product."
However, Calafati denies that Telx is competing against its carrier customers ?at least in terms of transport.
"We do maintain neutrality," he says. "Any of those carriers can come onto the exchange and also be the long-haul provider on the other side of the exchange presence. So if they need to take the traffic to another city or locale, they can use their network to do that. There's no requirement to use a particular network."
Unsurprisingly, the current primary driver for Ethernet exchange services is the need for service providers to fill gaps in their network footprints as demand for Ethernet services grows, says VSG research services director Erin Dunne.
"Geographic coverage is a competitive differentiator, so Ethernet providers are open to more efficient ways to deliver services to sites they can't reach directly," Dunne says, though she adds that at this stage, carriers are coming onto the exchanges more to sell rather than buy Ethernet services.
But, she adds, "Buyer enrollment is expected to increase as the business benefits of Ethernet Exchanges stabilize as compared to bilateral NNI agreements and other alternatives."
That said, the value of Ethernet exchanges as a gap filler also depends on the nature of the market gap being filled, says Diarmid Massey, VP for international carrier services, global markets, at Cable & Wireless Worldwide.
"The main value we see in exchanges is the extended reach. Where we won't use them is in countries where the traffic is so great that a peer-to-peer E-NNI arrangement makes more sense," he says. "It's the type of commercial arrangement where you just don't want an exchange in the middle as volumes grow."
That raises the issue of how much value Ethernet exchanges bring to the E-NNI question in particular.
Calafati of Telx says that his company's offering can help deal with the problem of disparate Ethernet services by providing a platform to "normalize" them.
"We can support customers who are maybe running on an old VLAN, they don't support 802.1q, or they're not ready for MEF 26. When we work with a customer, we'll help normalize that as part of a service harmonization offering, and move them to the standard that way," he says.
Sharp of Equinix says that Ethernet exchanges add value by enabling E-NNIs with multiple carriers. "When you implement the standard yourself to connect with another carrier, it's set in stone and it's a one-to-one connection. What we do is one-to-many, so the same investment they would put in to get to that one network they can put on our platform and be able to light up other carriers."
However, John Hoffman, head of Ethernet product management at Tata Communications, says that the one-to-many capability of exchanges is far more valuable than the E-NNI component.
"Personally, I think the E-NNI part is about 30% of their value add. The main value-add is their database," he says. "If I'm going into a country with multiple carriers in it and I don't have a large presence there, exchanges make a lot of sense, because I can connect into one exchange using standards that everyone accepts, and I can move circuits around a lot of different players, and it gives me the flexibility to find the lowest price. It also gives me the flexibility to come in once and get quotes, as opposed to me calling seven different carriers. And it doesn't have to be eight to ten circuits with one partner, it can two circuits with four different partners. That's more valuable to me."
Complicating the discussion is that when it comes to E-NNIs from a commercial point of view, the marketing is ahead of the technology, says Hoffman.
"Even the [E-NNI] standards are just partial standards," he says. "There's still work that needs to be done on that. So E-NNIs are still a complex undertaking, not just in terms of the technology and CoS, but also as a commercial effort."
Hoffman cites E-NNI monitoring as an example. "A lot of monitoring E-NNIs is done at the port level today, so you can't differentiate between customers coming across that E-NNI. Some people say that's no big deal ?I say it's no big deal until the customer has something go bad, and you call up your vendor or E-NNI partner and they say 'I can't see that'. So our E-NNI will start once we have service level monitoring."
Another problem, says Hoffman, is that the hype over E-NNIs is high enough that carriers may feel pressured to roll out sub-standard solutions that will ultimately experience major failures and burn customers.
"I've had a couple of customers tell me already, you can't put us over E-NNIs," he says. "Which is a little scary because if I'm trying to get into a country where I don't have presence, E-NNI is a good way to do that, so it's tough when they tell me not to do it."
Massey of Cable & Wireless agrees that there's still a considerable gap between the engineers and the marketing department when it comes to Ethernet services and E-NNIs.
"You talk to the exchange guys, and you hear stories about how nothing is the same, which is news to the commercial guys trying to sell Ethernet," he says. "I'm sure at the engineering level they know, but the commercial guys just think they're rolling out a nice new standard, so we might have messed that up a little."
Massey adds that it's worrying that carriers rolling out Ethernet services have put themselves in the position of needing Ethernet exchanges in the first place.
"There's all these Ethernet implementations that have evolved over time, and there are lots of forums managing the development of Ethernet, yet we have a situation where the answer is exchanges," he says. "I thought we'd all got a bit smarter than this. You'd think that in 2010, we carriers would be capable of rolling out a global standard."