Ethernet services may be moving to an inflection point in the broadband services marketplace faster than many people anticipated, thanks to service providers' resourcefulness in finding ways to expand greatly the usefulness of the venerable LAN protocol at low costs.
While a new generation of Ethernet gear is vital to broader application of the protocol, service providers are discovering they can save money by making sure they leverage excess fiber capacity to maximum advantage before choosing next-generation platforms. The upshot is Ethernet services are likely to play a more significant role in the overall service mix of the enterprise marketplace, much as many investors in next-generation technology anticipated, but without as high a payback to those investors as they had hoped.
There are about as many Ethernet service strategies as there are players in the game. But, despite the variations in network platforms, pricing, markets served and types of services offered, these service providers have a lot in common with respect to their directions as they take advantage of abundant fiber capacity to accomplish their aims. For example:
Ethernet service providers are moving beyond the traditional LAN extension application to make Internet access and outsourced IP-based applications bigger factors in their service mixes;
Service providers are exploiting the emergence of standardized approaches to interfacing IP/Ethernet networks with storage centers to make Ethernet a more potent transport option in this domain;
Ethernet service providers that have focused on supplying unprotected low-cost services are looking for ways to implement higher classes of service based on guaranteed bandwidth and quality of service, while those that have been operating at the high end of the market are embracing unprotected service strategies as well;
and metro service providers are looking to match the capabilities of Ethernet service providers that are in the end-to-end WAN space by exploiting newly installed MPLS capabilities on their long-haul suppliers' data networks to create Ethernet-based wide area private-line service offerings.
Moving Down Market
Nobody better exemplifies the change in thinking brought on by fiber glut than Metromedia Fiber Network Inc.
MFN has long served the high-end enterprise market with OC48 wavelength services, where customers buy one or more wavelengths over a single fiber strand out of the bundle of 44 fibers that typically connects a building to the ring. At 32 waves per fiber that adds up to more than 3.5 terabits of capacity per building, notes Treb Ryan, senior vice president of technology at MFN. "The demand isn't there for all that capacity," he says. "We can lower costs if we just put a gigabit Ethernet connector on each end of the fiber and sell gigabit Ethernet over a single fiber rather than by waves."
There are tradeoffs, of course. The strategy involves moving to Layer 3 router-based protection rather than using the full protection of an OC48 SONET ring. Ryan notes there are a lot of applications where that protection is unnecessary.
Now that MFN is ready to offer the unprotected pure gigE service, its next step, the timing of which remains uncertain, involves moving to a managed gigE service along the lines of the managed wave services the carrier currently offers its enterprise customers. "We can deploy relatively inexpensive gigE switches and interconnect offices using the same rings we use for the managed wave services," Ryan says. "We'll manage and own the switches."
Another carrier moving to add unprotected Layer 3 Ethernet to its portfolio is XO Communications Inc., which has more than enough fiber to go to this mode while maintaining its high-end "five 9s" Ethernet over SONET service. Both levels of service offer the same data rate options, namely, 10mbps, 100mbps or 1gbps. But the new unprotected service, by eliminating a lot of processing required to provide for optical Layer 2 protection, uses a cheaper add/drop multiplexer unit at the customer premises, says Garrett Hess, senior product manager for Ethernet services at XO.
"With some competitors now offering unprotected service at lower costs, it hurts us when the buyer's decision is made on the basis of price alone in instances where they don't need five 9s protection or guaranteed bandwidth," Hess says. "So it makes sense for us to take this step in order to meet the low-price competition."
XO's move retains the company's focus on Ethernet as a port-to-port LAN extension service, but that's likely to change, Hess says. Asked whether he sees a need to offer Internet access and enhanced applications over the Ethernet connections even with IP VPN and other IP services already in play, he answers, "Absolutely." Storage center interconnection, now provided on an account-by-account basis, is another service the carrier is planning to move into the Ethernet space, he says, noting this will be a national offering employing Fibre Channel "under the Ethernet portfolio."
Low-Cost Options
Of course, nobody is more aggressive about expanding the range of Ethernet services than Yipes Communications Inc.
A pioneer of Ethernet switching to provide high-quality Layer 2 performance with Layer 3 functionality, Yipes obviously is not shy about looking at advanced technologies to accomplish its aims. But even Yipes is making sure it avails itself of the benefits of abundant fiber before plunging into next-generation platforms.
For example, while the industry waits for 10 gigabit Ethernet to come on line, Yipes is experimenting with using multiple fibers to accomplish needed capacity gains, says CTO Kamran Sistanizadeh. "Point-to-point (10 gigE) solutions are out there, but, from a networking standpoint where we need interoperable devices, we're at least six to nine months away," he notes.
Yipes has found it can logically aggregate multiple gigE fiber feeds onto a ring so that Ethernet switches located at each node on the ring "think they're looking at 8 gigabits of Ethernet," Sistanizadeh says. Right now, it looks like this is a cheaper way to go than aggregating multiple wavelengths of gigE over a single fiber, he adds. This could mean that Yipes' need for a next-generation box that integrates wavelength management and Layer 2 switching into a single box will be postponed.
"There might be a need for this type of solution for some larger customers over the next nine to 12 months," Sistanizadeh says. But, he adds, the number of such customers may not be as high as once anticipated, given the linear pace in the growth of demand for bandwidth capacity. "From a technology perspective, it makes sense for us to be testing link aggregation using coarse WDM rather than DWDM," he says.
The abundance of capacity also is affecting the opportunities Yipes sees in the long-haul space and how it views the benefits of Resilient Packet Ring technology (see accompanying story, this page) that promises to give Ethernet services the quality of service protection afforded by SONET but at lower costs.
Where long haul is concerned, Yipes will soon implement a "National Area Network" service by exploiting MPLS deployment on the part of its long-haul transport suppliers Level 3 Communications Inc. and Genuity Inc. "We're using MPLS to set up virtual streams across these networks, so that, to the customer, it likes like a Layer 2 network," Sistanizadeh says. "It could be a very powerful replacement to traditional private line service."
As for RPR, the incentive to use the technology is now largely a matter of its ability to deliver high levels of QoS, rather than the two-to-one savings in fiber capacity consumption it offers over SONET, Sistanizadeh notes. The question is whether the failure recovery gain alone merits spending on the new technology, he adds.
While RPR will enable SONET-like performance at less than 50 millisecond recovery rates vs. the 30- to 35-second rate of the current spanning tree technique Yipes uses with Ethernet, "it remains to be seen how cost effective RPR will be," Sistanizadeh says. Even though RPR will be cheaper than SONET for five 9s performance over packet rings, it's uncertain how strong the demand for five 9s capability will be, especially when fast spanning tree recovery at one second or less, being tested by Yipes, becomes commercially available. Because all users are on the Yipes ring, the end-to-end recovery performance of its service must be compared not only to the five 9s of SONET rings but also to less-dependable long copper or fiber lines ILECs use to connect customers to those rings, Sistanizadeh adds.
MFN, which uses Layer 3 routing techniques to achieve two-second recovery on its rings, has similar concerns. "Two seconds isn't going to hurt most people," Ryan says. "Obviously we're looking at all the ways we can make our services cost effective while meeting customer requirements."
MFN, while it's considering RPR, also will be looking at the uptake on its new Layer 3 managed gigE service and how that might be expanded using coarse WDM, where operating just four or eight waves per fiber lowers the costs of lasers. The advantage of operating a Layer 3 managed Ethernet environment for the end user is that everything is much simpler when it comes to where one locates resources and how one extends use of software applications across remote locations on the network, Ryan notes. "I don't know whether we're going to see gigabit Ethernet move to more sophisticated technology platforms or whether keeping it simple using CWDM will turn out to be the best approach," he says.
The Cogent Disruption
One of the unknowns in how things will evolve among all the Ethernet players is the effect Cogent Communications Inc. will have on the marketplace as it expands the reach of its flat-rate 100mbps Ethernet service. At $1,000 per month, the service is $2,000 per month less than the price Yipes charges for its 100mbps MAN service. Cogent has linked its OC48 DWDM metro rings to more than 200 buildings across the 20 domestic markets it serves via a packet-over-SONET infrastructure, says Cogent CEO Dave Schaeffer. The markets are connected by eight lit wavelengths of OC192 (10gbps) operating over a 12,400-mile backbone.
Cogent's business model calls for offering the flat-rate service to small- and medium-size businesses in buildings with at least 20 tenants, are within 1,250 feet of the ring and have office space of at least 100,000 square feet. After a round of intense telemarketing and other research, Cogent selected 3,000 of these buildings as its core market. The buildings represent less than one half of one percent of all commercial buildings but accounts for 32 percent of the commercial square footage for rent and 65 percent of all IP traffic in these cities, Schaeffer says.
With just a year of commercial operations under its belt, Cogent is at the mere beginning point, with multiple customers in most of the buildings it serves and licensing agreements for access to another 1,000 buildings in hand. "Demand is strong, but clearly it's a difficult environment," Schaeffer says. He notes potential customers' biggest concern is whether Cogent has staying power. "The most common question we're asked is, how can you make money on this service?"
It's a question many competitors and industry observers also are asking. Schaeffer is confident Cogent can meet its business plan goal of 20 percent penetration per building across its targeted market. "We're comfortable with where we are and with our goals," he says. And, he adds, Cogent is set with respect to technology needs for the foreseeable future. "There might be incremental improvements that we implement as time goes by, but we haven't seen anything out there that says we can lower our costs by 50 percent or even by 20 or 30 percent," he says.
The Cisco-supplied optical and routing platform Cogent uses delivers two wavelengths for downstream and two for upstream per building on the ring, representing a total throughput of 5gbps in both directions to each building. The service is not SONET protected even though it is SONET framed, relying instead on the Layer 3 routing control and the counter directions of each pair of wavelengths serving the down and upstream flows to provide failover recovery to each building.
"The only reason we use SONET framing is that it gives us the densest packing per wavelength we can get with Ethernet," Schaeffer says, noting the Cisco system applies concatenation of SONET frames to facilitate efficient placement of Ethernet packets into time slots. Like other Ethernet providers that are putting buildings directly on rings, Schaeffer says the end-to-end performance quality associated with Cogent's service is superior to what customers can get from incumbents at about any price, let alone at $1,000 per month.
Interestingly, because Cogent is not concentrating on interconnecting facilities of any one customer in a given marketplace, its model rests not on the traditional LAN extension appeal of Ethernet but, rather, on the cost efficiency of using Ethernet pipes for access to the Internet and IP-based services. Rather than provide such services, Cogent loosely partners with various providers in security, storage, Voice-over-IP and other application spaces without taking a share of revenues for such services, thereby raising the appeal of the Ethernet proposition without losing its focus on being the transport provider, Schaeffer says.
Given the capital funding conditions of the marketplace, it's easy to understand why Ethernet service providers want to know where the sweet spot for next-generation service lies before they go on any spending binges. If they can win market share by convincing end users performance is better over ring-based Ethernet even without SONET protection than other approaches to broadband data services, and Ethernet access to Internet services offers a path to cost savings in storage, security and other outsourced services, they will have little incentive to go to the higher end capabilities offered by vendors.
The good news for vendors is incumbent carriers are beginning to move into more aggressive Ethernet service offerings (please see the February issue of xchange). To the extent that the Cogents and Yipes of the world succeed, the big carriers will more than likely be prime candidates for next-generation gear that will afford them a performance advantage over the upstarts.
RPR STANDARD MAKES MAJOR HEADWAYBy Fred Dawson |
"We left our meeting in Orlando (in late January) feeling very excited that we'd hit all of the milestones we set when we began this process over a year ago, including the big one, which was to get a first draft completed by January of 2002," says Bob Love, chairman of the RPR Alliance and vice chairman of the 802.17 Working Group. RPR technology 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 that RPR optical restoration and rerouting can be done in conjunction with use of SONET framing to carry essential network management information that operates below the physical layer. Ethernet over RPR, for example, makes more efficient use of existing SONET ring infrastructure by using bandwidth in both directions rather than just one while retaining the ability to provide less than 50-millisecond failover. The draft 802.17 specifications offer a wide range of options for implementation of RPR capabilities, reports Raj Sharma, director of strategic technology at Luminous Networks, Inc. For example, he says, the add/drop functionalities at each node can be tailored to a mode of operation that either gives complete priority to the traffic stream flowing over the ring or it can be set to assign priority to incoming traffic over traffic elements in the passing stream, based on the quality of service parameters associated with each. "Another significant area of compromise in bandwidth management has to do with including 'knobs' that allow service providers three different approaches," Sharma says. These range from absolute guaranteed QoS to best effort, with a flexible pro-rata assignment of bandwidth based on service level agreements as the middle ground. The trick, he adds, was to make sure that, in the case of network failure, rerouting on the ring didn't result in delivery of point-to-multipoint signals to different hubs out of synch with each other, which could create unacceptable levels of jitter. This meant "bandwidth steering" had to be used, vs. the SONET-like process of "wrapping," where any given optical ring is wrapped to another so that, in the event of failure, the signal reverses course and travels over the backup ring to its destination. Steering avoids the long delay that can be incurred by going completely around the ring to get to a relatively close point by redirecting from the point of entry onto the network. This shortens the distance traveled to the most distant node from the point of failure. "Over time people realized there were benefits with both options," says Gady Rosenfeld, director of strategic marketing for Corrigent Systems Inc. "With wrapping you minimize packet loss, while steering minimizes jitter. So the service provider can implement whichever approach works for its services, with steering set as the default in the draft standard." The 802.17 Working Group hopes to issue a draft for working group ballot following its meeting in July, which should lead to eventual final approval in a sponsors' ballot by early next year. By year's end vendors should be in a position to deliver standard-compliant product for testing, with any final refinements incorporated in time for commercial rollouts at about the time the standard is officially sanctioned around next March, Love says. "We see RPR as the ideal vehicle for carrying Ethernet," he adds. |
Participants in the standard-making process surrounding the Resilient Packet Ring Alliance report they have met their goal of achieving an initial draft of a standard in their most recent meeting.
CORRECTION
The article "George Schmitt's Juggling Act" in the March issue of xchange, page 68, incorrectly stated that XO Communications Inc. had at that time filed a petition for Chapter 11 bankruptcy protection.
