EtherLECs-Competitors or Saviors?

By David M. Piscitello
Core Competence, Inc.

Copper, DSL and "everything" first mile delighted Wall Street and enthused consumers for nearly three years. Lately, however, nearly all the discussion involving these data CLECs has focused on whether there is enough time, talent, and money to sustain the copper-based broadband services "economy" beyond 2003, when even the data CLECs with the largest cash reserves are expected to run dry.

Optical networking, however, is sustaining investor, consumer and service provider interest through this dreadful period of technology discontent. Furthermore, it is nurturing new markets and attracting investors in optical technology as well as infrastructure and services companies.

It's not too hard to understand this shift in allegiance. On-site service providers (OSPs)-a.k.a. Building LECs and EtherLECs-primarily address the high-end, high-return business markets. Service and install isn't "geographically challenged"-fewer truck rolls to fewer locations to install 10/100/1000 Megabit per second access lines can yield several multiples of recurring revenue over DSLs. While purportedly only 5-10% of the communications market has access to fiber, the majority of fiber terminates in office buildings. This market alone is clearly large enough and the recurring revenue model lucrative enough to support multiple players. But OSPs have other equally lucrative markets-ISPs, Application and Storage Service Provider and CLECs, all motivated to provide the fastest, most resilient services to customers at the lowest costs.

The OSP Business and Service Model
OSPs attack the incumbent carrier's business fiber component of the first mile market just as DSL CLECs have attacked the copper market. Their business play is to offer more granularly scalable bandwidth at substantially lower recurring and equipment costs by using an IP over Ethernet architecture where ATM/SONET and TDM have traditionally been deployed.

SONET/ATM in the MAN is out…
To appreciate how OSPs expect to pull this off, let's look first at how high-speed data services are offered over fiber today. Beginning at the customer's premises, businesses will need a high-speed router/switch with TDM (e.g., multiple T1) interfaces and muxes, or ATM/SONET interfaces.

These are expensive compared to Ethernet, and transmission rates above the rates where TDM equipment is required come in rigid increments: DS3 at 45 Mbps, OC-3 at 155 Mbps, and OC-12 at 622 Mbps. While VT concatenation, offered by companies such as Geyser Networks, allows bandwidth "grooming" in DS1 increments, the OC-3, -12, and -48 thresholds require interface upgrades.

Telcos build ATM/SONET access circuits using digital access cross-connect and add-drop multiplexing equipment to concatenate individual SONET synchronous transport signals in their backbone networks, where a ring topology is used to provide resiliency. Provisioning of this network is labor-intensive and complicated. The combination of manual operations and pricey equipment leads to an inevitably expensive service.

And IP over Ethernet is in…
Enter Ethernet, the perpetual thorn in ATM/SONET's paw. In the OSP architectures of metro EtherLECs and BLECs like Yipes, Cogent Communications, Intellispace, Telseon, et. al., customers connect to their metro area network service using a 10/100/1000 Mbps Ethernet NIC using Category 5 cabling or multimode fiber.

This Ethernet access circuit is connected to a Gigabit access switch or equivalent service unit in the building (on site). This equipment is in turn connected over fiber to a local or MAN optic ring. The core optical backbone networks vary across EtherLECs, and are constructed using Gigabit Ethernet and layer 3 MPLS-based switching over Dense or Coarse Wave Division Multiplexing equipment (DWDM, CWDM) or multiple connections into ISP backbones using gigabit wire-speed routers and OC-12, -48, and -192 circuits. These backbones are interim; many metro area service providers seek to eventually deploy access switches that can connect directly into an optically switched network core using Ethernet interfaces.

EtherLECs don't have the same difficulty telcos experienced when they first rolled out ATM. There are plenty of vendors vying for this market, from market giants Cisco and Nortel to Agilent, Africa, Ciena, Extreme, Foundry, Geyser Networks, Juniper, Sycamore, and more. Unlike the DSLAM vendors, many if not most publicly traded optical equipment companies have weathered the punishment inflicted on technology stocks by Wall Street and have sufficient funding to continue to deliver product, and many pre-IPO companies have adequate capital to sustain operations until the blush returns to the technology sector.

Acquiring Fiber
On-site Service Providers need fiber just as data CLECs need copper pairs. OSP's employ different business strategies when acquiring fiber. But the vast majority of metro dark fiber comes from other than incumbent LECs. Telseon, for example, has strategic relationships for dark fiber from Level(3) and MFN. Telseon, Yipes! and others also lease from municipalities, utilities and private companies. In selected cases, OSP's will build new fiber routes through contractors.

Many OSPs lease the majority of the dark fiber they utilize. The business environment OSPs have with fiber suppliers is far more cooperative than the adversarial relationships CLECs and DLECs have experienced with incumbents.

"We've found our suppliers to be very cooperative," says Nick Ciancio, VP Infrastructure Development at Yipes! "They want to sell and lease their assets". Ciancio also indicates that reasonably priced fiber was instrumental in Yipes! ability to offer service in 20 markets in less than a year.

Quality of Service and Granularly Scalable Bandwidth
With an Ether-switched backbone, EtherLECs will be able to meet stringent latency and availability metrics and commit to them in service level agreements. In fact, it appears that latency is the killer application in the metro area for EtherLECs and the major attraction for data center, storage, and application service providers.

Cogent Communications offers SLA's with a guaranteed sub-60 millisecond delay and complements this with five nines availability for their 100 Mbps non-oversubscribed service. These are achievable benchmarks for EtherLECs who make use of IEEE 802.1p, IP Class and IP Quality of Service (COS/QOS) traffic management mechanisms to allow customers to assign priorities to traffic. Traffic "markings" in customer IP packets and Ethernet frames are examined by Differentiated Services (DiffServ) capable access switches and forwarded across an MPLS-capable core switching fabric.

These and edge traffic shaping and policing functions are used to allow customers granular (i.e., megabit) increments of bandwidth, with fewer and less expensive interface upgrades. The provisioning of such changes is a matter of reconfiguring switches rather than rings and is so greatly simplified that Telseon offers web-based provisioning of bandwidth. EtherLECs all have their sights fixed on Gigabit and 10 Gigabit Ethernet as future customer interfaces, so scalability is assured as these technologies mature.

Cheap bandwidth and CPE, lower operating costs
Cogent Communications advertises $1000 per month for 100 Mbps non-oversubscribed Internet service. This is lower than T-1 service alone in many metro service areas. Both Yipes! and Telseon's prices are roughly "twice the bandwidth at half the price" compared to SONET although, according to Telseon's Vice President of Product Development and Founder Bob Klessig, "the same bandwidth at half the price" is the more typical Telseon customer take. Telseon,Yipes, and Cogent Communications services allows businesses to continue to use existing Ethernet category 5 connections.

The savings on the CPE alone buys you years of service. Do the math. A Fast Ethernet router port is somewhere between 10 to 20 times less cost than an ATM OC3 router port. For every $12,000 you reduce of equipment costs, you buy a year's service at 100 Mbps. Eliminating a $100,000 midrange router with ATM interfaces keeps you in service for nearly the remainder of the decade.

Finally, there's the matter of daily operations. An Ethernet MAN solution eliminates the need to hire or learn what Klessig aptly describes as "arcane WAN protocols such as OC-n, ILMI, Q2931, DXI, etc." I've been there; arcane is kind.

Is there a downside?
Like any infrastructure play, EtherLECs will burn capital for some time before they can be profitable. Dark fiber and optical networking equipment is way more expensive than DSLAMs and copper pairs, but fiber owners appear to be more amenable to leasing and faster to provision than incumbent LECs, and EtherLECs are betting on Ethernet's track record of dramatic increases in bandwidth and rapid silicon-ization of Ethernet-based technology, with the concomitant decrease in cost.

Prognosis for CLECs
Is this another situation where "my provider is my competitor" for CLECs? Telseon's Klessig insists the answer is no. "The typical bandwidth of a Telseon data connection is well above 10 Mbps. As such, we do not compete with DSL and cable modem service providers. Rather, they are our customers. Once they have aggregated their traffic together, they use Telseon to connect to their business partners, e.g., a DSL CLEC can use Telseon to deliver data to their ISP partners. Our service is typically less costly and much easier for the CLEC business partner to interface to."

For the early multi-tenant unit CLECs, it's quite possible that some companies will opt for higher rate Ethernet services, especially if the price differential is small. But direct competition at the (sub-) megabit market should be small, and the broader residential market, largely bereft of fiber, is safe. Overall, CLECs should expect financial relief from the onerous recurring costs of DS3 and OC-x circuits used to aggregate traffic off Central Office DSLAMs into their data networks, and broader competition for their aggregation circuit business. Or they can extend their play into this same market.

David Piscitello is president of Core Competence, Inc., a network consulting firm and founder of The Internet Security Conference.