CARRIER CHANNEL: Ethernet Hype Dampened by Market Realities, Technical Challenges

By Khali Henderson

Posted: 04/2002

Ethernet Hype Dampened by Market Realities, Technical Challenges

by Khali Henderson

IN SPRING 2000 WHEN THE FIRST commercial deployments of Gigabit Ethernet technology in metropolitan area networks (MAN), sourcing wholesale metro bandwidth seemed to be turned on its head. The introduction of inexpensive enterprise LAN technology into metro access and the metro core clearly represented revolutionary price-performance metrics — five to 10 times less expensive than the reigning metro protocol, Synchronous Optical Network (SONET).

Over the past two years, the Ethernet fanfare has quieted some under the weight of market realities and technical challenges. The latter — lack of support for TDM services and substandard restoration times — the industry has been preoccupied with controlling and the former — an extensive SONET-based metro infrastructure and tightening capital markets — it has so far been unsuccessful controlling.

While there is still plenty of noise to make the metro bandwidth marketplace confusing for metro carrier’s carriers and their wholesale buyers, recent market and technical developments may offer some clarity.

Market Realities

While optical Ethernet can be used for both wholesale metro access and metro core services, some are finding the rationale for the latter to be less compelling. The recent demise of metropolitan optical network providers Sphera Optical Networks Inc. and Sigma Networks Inc. may reinforce their view. Sphera filed Chapter 11, and Sigma Networks is in the process of closing its doors and selling its assets and customers.

“[Metro core players] are competing against the RBOCs and their customer base is small and credit challenged,” explains Dave Schaeffer, founder and CEO, of Cogent Communications. “The segment is not growing — it’s displacing one segment with another.”

Indeed, SONET rings deployed by the Bells and competitive providers are the predominant means of transporting traffic between central offices and carrier hotels within the MAN.

“Traditional telcos want to aggregate TDM traffic over SONET-based local facilities in a core metro ring,” Schaeffer says. “That is not the market we serve.”

With the notable exception of Telseon, neither do most of the so-called “EtherLECs,” the small group of upstart providers that pioneered Ethernet in the MAN. Instead, many are exclusively serving the metro access market, and in a predominantly retail play.

“Carriers are going to try Ethernet because the costs are so low,” says Doug McEuen, an analyst for Pioneer Consulting LLC. “Access is the only place where you can experiment. In the core, you’re going to have SONET.”

The truth is, many metro wholesale buyers aren’t even asking for Ethernet, says Neil Flynn, president of FPL Fibernet LLC. The carrier has 1,000 route miles of metro network throughout Florida — all of it SONET based. Why? “Because the demand isn’t there [for Ethernet]. No one is asking for it,” Flynn says, noting the carrier could offer it using its existing infrastructure, which includes Nortel OpTera switches. “Ethernet is in its infancy and it will grow, but in relation to the existing base of SONET traffic, it’s a small percentage.”

Pioneer Consulting predicts the worldwide Ethernet service provider market will increase from $217.2 million in 2001 to $14.081 billion by 2005.

Analysts at Communications Industry Research (CIR) contend 2001 was the year of hype for optical Ethernet and point out, even in the more promising access market, requisite fiber penetration to customers remains low.

“Carriers such as Yipes (Communications Inc.), Cogent and XO (Communications), which were supposed to drive Ethernet services market, are not viewed by CIR as viable businesses for the long-term given their recently published financial woes,” the analysts wrote in a January 2002 opinion.

Indeed, at press time XO was expected to file for bankruptcy, according to published reports from Reuters. Cogent, on the other hand, just started trading on the American Stock Exchange, and just purchased assets from PSINet out of bankruptcy. And, Yipes, which is privately held, reported 61percent growth last year, and booked $10 million in fourth quarter 2001.

Considering the fate of Sphera and Sigma, however, there may be some questions about the longevity of Ethernet MAN pioneers, but there is little doubt Ethernet in the MAN is here to stay.

RBOCs, which have SONET cores and which dominate the public network landscape, are starting to look at Ethernet technology as are IXCs and more established CLECs.

Metromedia Fiber Network told PHONE+ it has plans to come out with a GigE service soon. Details were not released.

Carrier’s carrier Williams Communications Inc., for example, is rolling out its Ethernet services in second quarter 2002 as part of its metro strategy announced last summer. The company already hinted at the offer in an announced deal in November 2001 with Yahoo! Inc. in which it agreed to provide the portal with local-to-national-to-local optical Gigabit Ethernet service. In addition, the companies are collaborating on the development of Williams’ wide-area Gigabit Ethernet services.

“We literally see Williams as a fresh alternative to new MAN players and legacy CAPS and RBOCs,” says Barbara Fuente, manager of product marketing — transport services for Williams Communications. “Some customers will ask for Ethernet, but we are going to have to consultatively sell and do what’s right for the customer on an individual case basis.”

The customer, she says, could be a carrier, but is more likely to be a content provider with need for huge pipes — a bit of a departure from Williams’ traditional wholesale target but fairly typical of etherLEC targets.

“We are confident that both of the services [SONET and Ethernet] will be in demand. Different types of customers have differe nt QoS needs,” she said.

Access Methods by User Type
Source: Williams Communications Inc.

Technological Advances

Fuente’s comments get to the heart of the technological challenges facing optical Ethernet — QoS.

“The lower costs of GigE equipment make it extremely attractive to financial strapped carriers,” said Paul Kellet, senior director of research for Pioneer Consulting. “However carrier-class reliability and quality of service must be clearly demonstrated to carriers to achieve further market penetration.”

The single most significant difference between optical Ethernet and SONET has been SONET’s capability to restore a connection in less than 50 milliseconds (ms), explains Richard Cunningham, senior analyst, optical networking for Cahners In-Stat/MDR. “In contrast, the best that optical Ethernet could achieve was a restoration time (also called latency) of a few seconds, far too long to be acceptable for voice communications.”

That may be changing.

“Ethernet is closing the gap with SONET from a performance perspective and a reliability perspective,” says Kamran Sistanizadeh, chief technology officer for Yipes. “All the standards groups — IEEE 802.17 and the Metro Ethernet Forum — and different vendors are pushing it.”

The IEEE 802.17 Working Group, for example, closed out its January meeting in Orlando with the first draft of the resilient packet ring (RPR) standards putting vendors on course to deliver compliant product for beta trials by year’s end. The group hopes to issue a draft for ballot following its plenary meeting in July.

RPR is designed to allow traditional TDM and packet-based services to flow over the same optical infrastructure in a more bandwidth-efficient manner than can be done with SONET alone, but it is compatible with SONET in the sense RPR optical restoration and rerouting can be done in conjunction with the use of SONET framing to carry essential network management information. For example, Ethernet over RPR makes more efficient use of existing SONET ring infrastructure by using bandwidth in both directions, rather than just one, while leveraging SONET link protection to provide less than 50 ms fail over.

“RPR is giving Ethernet some of the qualities of SONET,” says Pioneer Consulting’s McEuen. “Whether that will take hold is yet to be seen. Industry hype doesn’t guarantee success.”

Sistanizadeh says we should see actual deployment of RPR by the end of 2002. Yipes, for example, already has completed lab trials with various undisclosed vendors, and will have a limited trial in third quarter and possible a limited deployment by the end of the year.

Others PHONE+ talked to were less optimistic about RPR’s short-term impact. “It will be some period before RPR finds a market, says Greg Wortman, vice president of marketing, Coriolis Networks Inc. “It represents a whole new architecture. You have to build a whole new network. It will be several years before we will see it. We don’t see carrier momentum. What we see is carriers that want to evolve their SONET network to handle data more profitably.”

Coriolis Networks and other startup vendors like ATOGA Systems have come up with solutions to enable legacy carriers with SONET infrastructure to also run Ethernet.

“We are bringing a network platform based on SONET framing but doing it in a fashion so as to bring in new technology to make SONET efficient at Ethernet transport. For carriers focusing on T1, private line and OCn and want to add Ethernet, it’s a good evolutionary approach,” says Wortman, adding it’s an option for providers that want to start with Ethernet and add TDM services later.

Rather than add a Layer 2 Ethernet switch over Layer 1 SONET, which adds expense and management overhead, these new solutions basically combine a SONET ADM with a Layer 2 Ethernet switch over a DWDM layer into one network element.

Coriolis Network’s OptiFlow Network Solution started shipping last summer. Its first customer, Marietta FiberNet was signed in October. Marietta FiberNet, Georgia’s state’s first municipally owned company to be certified as a CLEC, initially is using the OptiFlow Network to deliver optical Ethernet service to other carriers, enterprises and local government organizations in the metropolitan Atlanta area. Wortman says the company also was involved in trials with a large IXC that was going to build our metro markets to service wholesale customers.

“It costs 20 percent more for what we are doing vs. SONET [alone],” says Wortman, “but 40 to 60 percent less when you add in the cost of the Layer 2 switch. And that doesn’t include operational savings of a single network element for provisioning and management.”

Coriolis Network tests show an 82 percent savings in costs of provisioning with the aggregated box.

“The value proposition is that it is a converged solution, says P.G. Menon, vice president of marketing for ATOGA Systems. “You don’t have to invest in [SONET and Ethernet] both. You can retail out TDM now and Ethernet later.”

ATOGA’s platform, the OAR5i, has been in trial with one long-distance carrier and one cableco. Menon expects commercial deployments will be announced in the latter half of this year.

ATOGA also is talking to an EtherLEC about running Ethernet over DWDM, Menon says. “We have integrated tunable DWDM into the platform instead of having an external DWDM device. So we are using DWDM to carry multiple channels of Ethernet over a single strand of light,” he says, explaining this improves speed to market and reduces provisioning time.

In contrast, most native Ethernet providers only offer 1gbps. With DWDM, you can get 40gbps on the same piece of fiber. In addition, aggregation of gigabit traffic with WDM is presently much more cost-effective than moving to 10 gigabit Ethernet transport, according to analysts at CIR.

Parting Shots

To be fair, not all traffic requires SONET-like QoS. And, to be even fairer, optical Ethernet offers advantages over SONET even in the area of QoS.

As proponents are fond of pointing out, 95 percent of the LANs in America are Ethernet-based. Furthermore, adds Ron Young, chairman of the Metro Ethernet Forum and co-founder of Yipes, a high percentage of those are connected to other LANs via SONET or ATM, which he says introduces additional points of failure.

“There is a bottleneck between the LAN and the backbone,” Young says. He explains running traffic over the last mile of the network is akin to a New Yorker taking a stage coach to John F. Kennedy Airport, boarding the Concorde to LAX, and taking yet another stage coach to his ultimate destination in Los Angeles.

Yipes CTO Sistanizadeh says T1 lines, the last-mile connections, do not conform to SONET standards. “For true end-to-end performance, metro Ethernet compares favorably (to SONET),” he says. “In our network, which has a ring architecture, we bring it to the basement. We emulate the functionality of SONET in the premise as opposed to just the backbone.”

Yipes says it guarantees latency within its large metro optical networks to be

less than 10 ms — well under the SONET benchmark.

It’s not surprising then to learn 72 percent of Cogent’s customers use its Ethernet services to replace T1s.

Another key advantage is scalability. “With SONET, going from a T1 to a T3 is a 30 times jump — no one buys that way,” says MEF’s Young. “Ethernet offers fast scalability from 1mbps to 1gbps in 1mbps increments.”

Such scalability is on demand. “If I take and buy a DS3 of SONET and I need

to upgrade, I have to pull a card and buy a new one. With 200mbps Ethernet, I can double it without a hardware change.

It’s a software change up to a gigabit,” explains Andrew Crabtree, senior product manager — Ethernet services, for Williams Communications.

“In lots of cases, even in a wholesale scenario, what’s needed is a portion [of the available bandwidth], not the whole thing or not the whole thing all the time,” says Wortman. “Provisioning it in small increments helps meet that requirement. Ethernet has a lot of burst to it, so being able to offer tiered service with guaranteed bandwidth up to a full GigE, being able to burst is a benefit because you don’t have to pay as much, and for the wholesaler it offers to ability to oversubscribe for a lower cost basis.”

Additional reporting on the IEEE 802.17 Working Group by Fred Dawson.