Metro Optical Ethernet

A multibillion dollar market * Ethernet vs. Sonet * Pioneering service providers * How vendors stack up

November 13, 2000

32 Min Read
Metro Optical Ethernet

As a new generation of IP applications and services moves into metro, so too does Ethernet. The metro area is currently the focus of intense network development and competition, with new service providers appearing on a monthly basis.

No longer is the traditional CLEC (competitive local exchange carrier) the one to watch. Their business model of bringing savings to customers by bypassing the incumbent carrier’s local loop has not proven viable for new entrants. The field is too crowded and differentiation is hard to demonstrate (see Capillary Bleeds Out).

New era LECs are beginning to understand that as applications become increasingly IP-centric so too must the network, and that begins in the metro. Applications service providers, storage service providers, and Web hosting centers are popping up in metros across the U.S., encouraging further disaggregation in the carrier markets.

This disaggregation is creating a number of new LECs to watch, including those that provide capacity on a wholesale basis, those that create and manage services to both ISPs and end users, and those in the retail business of providing low-cost access to these new services.

Coincident with this rise of metro IP services and service providers is metro optical Ethernet. This is not your LAN’s Ethernet, though it benefits from Ethernet’s low cost and ubiquity. It is a “carrier-grade” Ethernet, able to (at least if you believe the vendors) support classes of service, quality of service (QOS), security, scaleability to 10 Gbit/s, and Sonet-like network survivability.

Metro networks employing this equipment will be able to create and offer Gigabit-level connections between customer access gear and service provider POPs (points of presence) in their “native” mode, from one gigabit Ethernet switch or router to another. By eliminating the Sonet and/or ATM layers between those connections, new era CLECs will be positioned to compete very effectively with incumbents on price/performance metrics as well as true service differentiation.

In this report we’ll examine metro optical Ethernet from all the angles: the technology, applications, economics, and service providers, as well as their vendors.

Hyperlinks to different sections are given below:

Ethernet Rising
Ethernet Gets MANly
Crunching the Numbers
How Big Can It Get?
Ethernet is networking’s great survival story. Taking blows from token ring, FDDI (fiber distributed data interface), and ATM (asynchronous transfer mode) in the enterprise like a seasoned prizefighter, Ethernet has emerged from each bout looking stronger and more relaxed. Though not necessarily technologically superior to its challengers, its low cost and ease of use has made it the most resilient technology of the past two decades.

The latest market research on LANs (local area networks) claims 320 million connections will be Ethernet in 2000, while contenders’ market share fades away into oblivion. The result is a ubiquitous, low-cost networking technology that is resident on over 95 percent of all corporate desktops and quickly emerging as the de facto interface for broadband access in the home.

Beyond the access market, Ethernet is proving its scaleability, with millions of gigabit Ethernet interfaces shipping in 2000 for switch and router interfaces that will be appearing not only in the enterprise, but in service provider networks at the edge and core of the network. Soon we’ll have 10-gigabit Ethernet, which will begin putting pressure on OC192 as a low-cost alternative for transporting aggregated IP flows across MAN (metro area network) and WAN (wide area network) backbones.

This gets to the heart of the recent buzz around metro optical Ethernet. If the interface on LAN equipment and service provider POP equipment is Ethernet, then why not find a way to eliminate all that costly Sonet, frame relay, and ATM in between?

Prior to 2000, the answer was simple: Ethernet traffic cannot be transported natively beyond a couple of kilometers and has none of the performance monitoring, latency, and jitter control and restoration capabilities of Sonet, thus no SLA (service-level agreement) support for service providers. In addition, the IP traffic that Ethernet is carrying will be “best effort,” which means only the lowest-revenue access services can be offered, rather than high-margin.

In a time when carriers are looking to add value to basic transport services, point-to-point Ethernet looks a bit unprofitable. ATM and Sonet added network survivability, oversubscription for more efficient metro bandwidth utilization, and a means of managing data traffic with the richness that ATM can support.

But these benefits come with a high price tag. They can be economically justified if IP traffic is bound for the WAN. But if that traffic is meant to stay within the metro (say, for a transparent LAN connection between two offices, or a high-speed connection between a corporation and its ASP), then Sonet and ATM appear to be overkill, especially as IP evolves to be more robust than its previous incarnations.

The question a number of startups plan to answer is how do you stay all-IP in the MAN while supporting QOS, survivability, service provisioning, security, and scaleability? The answer for this group is metro optical Ethernet, but the vendors differ on just how to implement that.

Vendors today can be roughly grouped into two camps: the pure-play packet, and the “Sonetized” packet switch.

The companies generating the buzz around metro Ethernet are the pure-play vendors, including Cisco Systems Inc. (Nasdaq: CSCO), Lantern Communications, Luminous Networks Inc., and Nortel Networks Corp. (NYSE/Toronto: NT), all part of the Resilient Packet Ring efforts underway in The Institute of Electrical and Electronics Engineers Inc. (IEEE). These vendors support bridged VLANs (virtual LANs) and/or routing, and make provisioning a snap around their metro fiber rings.

Other pure-play vendors include a number of Layer 2/3 switch vendors, including Extreme Networks Inc. (Nasdaq: EXTR), Foundry Networks Inc. (Nasdaq: FDRY), Riverstone Networks, World Wide Packets Inc., and Zuma Networks. These companies are adding latency and flow control capabilities to their gigabit Ethernet switches to support multiple classes of IP services from POP to POP or from access points to POPs.

The “Sonetized” vendor of note today is Appian Communications, which takes Ethernet services into its box and maps packet data into Sonet virtual tributaries, directing them to the proper switch in the metro POP. The packet switch at the core of the Appian solution gives users the flexibility of self-provisioning bandwidth and service characteristics, while keeping equipment costs and cost of ownership well below its full Sonet rivals in the access network.

It’s worth noting here that these players aren’t the only ones supporting metro Ethernet. Ethernet services can be carried over Sonet, ATM, wavelengths, or a variety of proprietary methods. What’s important about these new entrants is their singular focus on Ethernet. This is both their strength and their vulnerability. They can provide Ethernet services more cheaply than anyone, but that’s about all right now. Their multiservice capabilities remain in question until beta trials are there to prove them.

Next-gen Sonet players can carve gigabit Ethernet up into virtual trunks and switch them through a Sonet ring, with Layer 2 traffic management and oversubscription. These are ideal if the majority of traffic is circuit and a carrier wants to add an Ethernet service option. But carriers like Yipes Communications Inc. want to be different, and change the pricing of access services altogether. For that they need pure Ethernet, and they have inspired a whole new breed of carriers with Ethernet on their mind.

Metro Ethernet networks are designed to do much more than provide cheap Internet access. That’s just a way in the door. The real value, and the lasting revenue stream for new “EtherLECs,” will be value-added services support. With the arrival of 10-gigabit Ethernet in 2001, the range of applications for metro Ethernet networks is theoretically all of them, though true multiservice support may be longer in coming than most vendors would have us believe.

Early applications will be those that require efficient low-cost transport of IP data among multiple sites in a metro. These “interPOP” applications will consist of interconnecting data centers, colocation facilities, cable data headends, and carrier data POPs. In these cases, 10-gigabit Ethernet switches will aggregate multiple gigabit Ethernet streams from data equipment and provide the lowest-cost transport possible in a metro. This will create stiff competition to metro DWDM (dense wavelength-division multiplexing) players and OC192 Sonet, as the price/performance ratio can’t be beat.

More advanced applications will require multiservice capabilities. That’s where true metro optical Ethernet will prove its mettle. These applications will include:

EtherVPNs Virtual private networks continue to be touted as the next wave in metro networking, but provisioning these remains a challenge for service providers. It often comes down to cheap, flexible bandwidth, which is in short supply in today’s networks. Some of the metro Ethernet vendors propose setting up secure VLANs in the metro, all provisioned and managed easily over a distributed packet switch architecture. In the Lantern solution, for example, thousands of VLANs can be supported on a single ring, giving operators a great deal of flexibility in assigning and prioritizing bandwidth to customers.

ASP access Application service providers today have a difficult time keeping customers. It’s not their fault, they claim; it’s the network’s. ASPs need high bandwidth connectivity with their customers at low costs. They also need flexible network infrastructures that respond with a high degree of agility to shifting customer usage patterns. Metro Ethernet may do a much better job than today’s hard-wired private lines.

SSP access Much the same as the ASP, the storage service provider requires high-speed connections between customers and server farms, with very low bit-error rates. Gigabit Ethernet is becoming a popular SAN (storage area network) protocol, and metro Ethernet networks could easily connect their customers with data warehouses, disaster recovery centers, and hosting centers.

Real-time monitoring and provisioningConsider this example from Riverstone Networks. A company holding a one-hour Web-based videoconference might need extra bandwidth for that purpose. To obtain the additional bandwidth, the company’s network manager uses the service provider’s Web-based self-provisioning interface to enter a request, and within seconds the bandwidth is provided. After the teleconference is over, the bandwidth automatically scales back down to the customer’s normal allocation. Or a company network administrator can use a self-provisioning interface to turn up varying levels of bandwidth for individual applications, offices, or VPN connections.

Ethernet is a LAN technology, and though it does a fine job interconnecting desktops to servers and switches in an enterprise, it was never “hardened” for service provider networks. That was left to Sonet, ATM, and MPLS (multiprotocol label switching), which provide the necessary traffic management and SLA verification capabilities to allow service providers to both create and manage data services to their customers.

But as the market for high-speed data services moves well beyond the largest enterprises, the strict guidelines of frame relay or ATM SLAs seem less important. Companies want broadband access and they want it fast. Once they get this broadband access they’ll start asking for more advanced IP services, and the service provider with a way to provision these rapidly at the lowest possible cost will always come out on top.

What will they need?

Low latency and jitter This will allow service providers to emulate a TDM (time-division multiplexing) circuit, and thus support VOIP (voice over IP), video, and other multimedia services. Circuit emulation really never caught on with ATM because of its cost and complexity, but it may have a second chance at success with low-cost Ethernet. This capability could open the door to real all-IP networks.

Survivability In the metro area fibers get cut. Backhoes are always going where they shouldn’t, and Sonet has proven over the years it has the most rugged protection mechanism around. Not only can it restore around fiber cuts in less than 50 milliseconds, it can tell you where it happened and when, through advanced fault location techniques. Metro Ethernet needs something similar, and that’s what is being undertaken by the Resilient Packet Ring Study Group in the IEEE. This spec will support multiple grades of restoration, from Sonet-like to varying levels of priority. In the meantime, optical layer automatic protection switching will have to suffice.

Reach Metro Ethernet needs to go metro distances, which often span 100km or more. Today 40km reaches are being specified for 10-gigabit Ethernet, but if you put it over a wavelength, then distances of over 120km can be reached in the metro, and much longer in the core. By and large, long-haul gigabit Ethernet services will be the province of DWDM players, and companies like Ciena Corp. (Nasdaq: CIEN), Nortel, and Sycamore Networks Inc. (Nasdaq: SCMR) are building systems today to support managed gigabit Ethernet services over DWDM networks.

Traffic and congestion management ATM is great at this; Ethernet isn’t. Metro Ethernet solutions must adopt some method of traffic management if they want to reap any of the benefits of oversubscription. Cogent Communications Inc., on the other hand, has built an Ethernet network without any oversubsciption, giving them the luxury of forgetting about congestion and traffic management altogether. Everyone gets their own 100 Mbit/s through the network, whether they are using it or not. That’s easy today, but does it scale?

Security Ethernet isn’t very secure, and sending packets out onto a public shared network will worry the business user. Today Yipes is using equipment from Netscreen Technologies Inc. to ensure security of individual accounts. In the future this may be built into metro Ethernet equipment.

Standards There are two important standards efforts underway to make Ethernet more suitable for the public network. One is to standardize the way in which Ethernet or packet switches can be deployed survivable in metro rings, and the other is to standardize 10-gigabit Ethernet, both for LANs and WANs. In September, the IEEE P802.3ae (10-gigabit Ethernet) Task Force completed the first draft of the 10-gigabit Ethernet specification. Final ratification for the standard is expected in early 2002. (For more on standards, see IEEE Nails Down 10-Gig Ethernet Basics,Ten Gig Meeting Sparks Optical Flap, and 10-Gigabit Ethernet Gets Marketing Muscle).

Dark fiber This may be a real sticking point for the rapid uptake of metro Ethernet in the access network. Because services start at gigabit Ethernet, they need to run over fiber, and though fiber is being drawn into metro conduits at a pace of millions of kilometers a year, it is still rare to find it out in the access network, entering the basements of office buildings. But it is coming, and the last-mile challenge only limits the metro Ethernet market by a fraction.

One of the main attractions of Ethernet is that it's very inexpensive compared to Sonet and is likely to remain so for the foreseeable future, according to a study conducted by the Dell'Oro Group, a market research firm.

Dell'Oro's study compares the average selling price (ASP) of fast Ethernet (100 Mbit/s), gigabit Ethernet, and 10-gigabit Ethernet switches with OC3 (155 Mbit/s), OC12 (622 Mbit/s), OC48 (2.5 Gbit/s), and OC192 (10 Gbit/s) Sonet gear. These prices are converted into dollars per gigabit of bandwidth so that a common price/performance metric can be used for all technologies and bandwidths.

With fast Ethernet, for instance, 10 ports would be needed to deliver 1 gigabit of bandwidth, each port costing $475 in the year 2000. So the ASP per gigabit is $4,750. With gigabit Ethernet only a single port is required, so the price comes down to $1,550 this year.

These prices are based on Layer 3, modular-based, fiber ports. Including copper products would make Ethernet pricing much more cost effective but would have skewed the comparison, according to Dell'Oro.

In the future, 10-gigabit Ethernet will undercut gigabit by another significant margin, according to Dell'Oro. All prices will come down over time, but the differential between them is likely to remain much the same. Dell'Oro reckons that the ASP per gigabit of fast Ethernet will be $3,116 in 2004. Gigabit Ethernet will cost $838 by then, and 10-gigabit Ethernet will cost $485.

ASP Per Gigabit of BandwidthDell'Oro applied the same methodology to Sonet equipment. In order to buy a gigabit worth of bandwidth using OC3 ports, about 6.5 ports would be needed, costing $6,229 apiece. So the total comes to $40,118. Higher-speed Sonet gear reduces the ASP per gigabit by about 50 percent.

Once again, prices are likely to come down over time, but the cost differentials among the various port speeds remain about the same. The OC3 gear comes down to $26,367 in 2004, while OC192 gear is $3,143.

ASP Per Gigabit of BandwidthWhen comparing the two technologies, it's clear that Ethernet bandwidth will be 85 percent cheaper than Sonet bandwidth for the foreseeable future. Putting that another way, service providers can spend $150,000 on Ethernet equipment or $1 million on Sonet gear to get the equivalent bandwidth -- and that saving isn't going to change much over time.

Here's the comparison at 10 Gbit/s. In 2001 the ASP per gigabit of 10-gigabit Ethernet is $950, compared to OC192 Sonet's $7,653. The comparable figures in 2004 are $485 and $3,143.

ASP Per Gigabit of BandwidthCogent Communications

Cogent Communications Inc. is all about business model and has built a network around the sole proposition of providing 100-Mbit/s Ethernet services to tenants of office buildings for $1000 per month, roughly the price of a traditional T1 (1.5 Mbit/s) line. To that end, Cogent has partnered with Cisco to build its network from core to edge with Cisco gear.

In the building basement Cisco gigabit Ethernet switches interface with 15454 transport muxes and passive metro DWDM units. These are aggregated back at metro POPs, which contain the famous GSR 12000s. These GSRs are interconnected via DWDM systems from Pirelli, now part of the Cisco fold. There is a little room left in the contract for Chromatis (now Lucent) to use its 32-channel multiservice DWDM systems on rings with more than 16 buildings.

The network looks solid, but the lingering questions surrounding Cogent center on money. Can they make enough to pay for this infrastructure? Cisco provided much of the financing, but Cogent has to start getting customers signed up quickly to get the revenue flowing. At T1 pricing, Cogent needs volume, not just a few tenants per building. And they need access, which means building fiber out to buildings currently served by copper. This is a time-consuming process, with no guarantees of success.

Thus far, Cogent is executing on its goals, building out its network to hundreds of buildings nationwide and signing up customers more quickly than any other competitor. They should break the thousand mark this year, putting them well ahead of their metro-only rivals.

(More about Cogent on The Fat Pipe Formula, Cogent Banks on T1 Replacement, and Next Gen Carriers Cash In.)


Intellispace is one of the more aggressive and innovative "Building LECs," wiring major office buildings in major metros and offering tenants low-cost,Ethernet-based access to the Internet and service provider networks.

Like a number of other BLECs, Intellispace got its start in New York City, wiring 55 Broad Street with fiber and high-speed switching gear. Since then Intellispace has added over 400 buildings to their network, and expanded their service offerings. The company's networks are primarily concentrated in the northeast U.S., but they have since expanded to the west coat and to England and France in Europe.

To create these multiservice in-building networks, Intellispace is using routers from Riverstone, network management from Aprisma, service management software from Dirig Software, enterprise switching equipment from Cabletron, metro fiber from Metromedia Fiber Networks, and Ethernet switches from Extreme Networks.

With their switches and routes in building basements, Intellispace is able to offer dedicated high-speed Internet access, firewalling, VPN, email accounts and Web hosting. This isn't quite the range that Everest is planning to offer, but it is based on a similar Ethernet-centric architecture that allows the highest capacity access to network service providers at the lowest cost per transported bit.

Intellispace is one of the first carriers to announce the availability of granular access self-provisioning, called “eValve", which allows customers to instantaneously increase or decrease Internet access speeds. Users can access speeds from 64 kilobits per second to 1 gigabit per second. To increase or decrease their bandwidth, customers call a regional service office, which makes the requested speed available in 60 seconds or less. The service will have a web interface in the coming months, allowing customers complete online control of their access speeds.

(More about Intellispace on Intellispace Inks $200M Deal With MFN and Riverstone Announces Metro Router)

(More about other "Building LECs" on The Big Idea)


Telseon is positioning itself as a provider of managed gigabit Ethernet services and to that end has built metro networks out of dark fiber and equipment from Riverstone networks to begin providing services. Telseon will be adding equipment from Foundry to scale its core POPs.

Telseon says it’s on target to offer optical Ethernet service in 20 metropolitan areas by year’s end. Expansion into international markets begins Q1 2001.

Telseon offers three types of connections: Point-to-Point for inter-POP connectivity or access connectivity; Point-to-Multipoint for service providers connect to multiple customer sites or for multicast distribution; Multipoint-to-Multipoint in which a single organization connects to campuses, multiple carrier POPs, or trading communities. Customers connect to the Telseon network via a standard Ethernet interface. Network equipment is attached to a Telseon Service Interface Unit (SIU) via standard 10/100 Mbps Ethernet over a Cat 5 cable. Redundant fiber connectivity from the SIU to the Telseon secure core yields high availability and up to 1000 Mbit/s inaggregate bandwidth.

Telseon¹s service allows a customer to keep their existing IP Addressing Scheme; Telseon IP Service forwards packets based upon the Ethernet MAC address. Telseon IP Service connects locations via ³logical wire² connections. A logical wire is a virtual connection between Service Interfaces in a Telseon metropolitan network. Once network access is established, customers are ³on net² and can add or change logical wires and increase or decrease bandwidth on demand via our Web-based provisioning system. Logical wires have two important characteristics:1) packets are delivered transparently from source to destination, just as if the two Service Interfaces were connected by a physical wire, and 2) logical wires are leak-proof; no packets will leak into, or out of, a logical wire. Once connected, customers can use a web-based interface to create new logical wires or change bandwidth allocation in minutes, prioritize traffic with different classes of service, view service configurations and make changes,and obtain reports on performance, utilization and network events.

XO Communications

In its new incarnation XO Communications (formerly Nextlink/Concentric) is talking gigabit Ethernet and IP. XO has begun the first phase of an expansive migration to packet-based switching technology, which is expected to deliver the full range of traditional and enhanced local and long-distance services and to increase the company's ability to efficiently handle an increasing amount of ISP traffic. That plan includes shoring up its backbone with Ciena DWDM gear, fitting that with gigabit Ethernet interfaces and management systems, and installing softswitches from Sonus Networks Inc. (Nasdaq: SONS). XO is an “EtherLEC,” as defined in this report, but it is focusing significant energy on delivering gigabit Ethernet transport solutions, which may migrate to a more fully realized metro Ethernet offering in the future.

Yipes Communications

Yipes got this market started and put a brand on the concept of metro Ethernet services. It has created a national network by building out metro fiber networks with Metromedia Fiber Network Inc. (MFN) (Nasdaq: MFNX) fiber, then leasing capacity on the Level 3 Communications Inc. (Nasdaq: LVLT) and Qwest Communications International Corp. (NYSE:Q) backbones to create a true national backbone.

Yipes services include Yipes MAN (LAN-to-LAN service between business locations) and Yipes NET (high-speed Internet service), both scaleable from 1 Mbit/s to 1 Gbit/s, in 1 Mbit/s increments. Businesses can also connect sites across the country at customer-selected speeds using Yipes WAN, a wide-area network service. The Yipes network employs a combination of Juniper Networks Inc. (Nasdaq: JNPR) core routers, Extreme Networks Inc. (Nasdaq: EXTR) Summit switches, and NetScreen WAN security.

Yipes Communications Inc. has announced service in 15 cities: Atlanta; Boston; Chicago; Denver; Ft. Collins-Longmont, Colo.; Ft. Lauderdale, Fla.; Houston; Miami; Palo Alto, Calif.; Philadelphia; Riverside, Calif.; San Diego; San Francisco; Washington, D.C.; and Worcester, Mass. By the end of this year, Yipes will serve 20 cities coast to coast.

(More about Yipes on The Big Idea,Next Gen Carriers Cash In,Yipes/Syndesis Integrate Provisioning Plans, and Yipes Bets Big on Gig.)

Appian Communications

Appian Communications is walking the line between next-gen Sonet and metro optical Ethernet. Appian’s Optical Services Activation Platform (OSAP) 4800 system has a strong distributed packet switch foundation, which delivers "packet private line" services that emulate TDM private lines but with the flexibility of packet switching.

Subscribers connected to a service provider via Appian's OSAP 4800 can obtain Guaranteed Bit Rates (GBR) and Maximum Burst Rates (MBR) that provide specific bandwidth guarantees instead of a "best efforts" service. The GBR and MBR parameters are software definable and can be changed at any time by the operator or subscriber, based on the individual SLA. Within each SLA, the Appian solution will also support IP Class of Service (COS) based on MPLS, IP TOS (including DiffServ), or 802.1p standards, for support of real-time applications such as IP telephony on the same interface as the one carrying IP data.

Appian has also been promoting its own protection scheme, called Optical Data Protection (ODP). ODP is the first protection mechanism that allows sharing of optical paths among multiple data services on a ring with sub-50 millisecond switchover at Layer 1.

Appian currently has an ongoing beta trial in Everest Broadband Networks’ building networks and betas in at least two other carrier networks.

(More about Appian on Appian Peps Up Provisioning andAppian Teeters On the Edge .)


Atrica Inc. is a pure-play Ethernet startup to watch. As a kind of spinoff from 3Com Corp. (Nasdaq: COMS), Atrica is building a metro optical Ethernet product based on three main tenets: optical switching, Ethernet technology, and traffic management. It is too early to say just how they will bring these together, but Atrica has a sound management team with plenty of experience in working with Ethernet. See their contribution to this marketplace at

(More about Atrica on Out of Atrica and Benhamou Steps Aside at 3Com.)

Extreme Networks

Extreme Networks Inc. (Nasdaq: EXTR) was one of the first gigabit Ethernet switch companies to speak directly to the needs of metro area Ethernet networking. Because of that, it was chosen by Yipes to be the foundation of its switched Ethernet access network.

The Summit switches aggregate multiple 10/100 Ethernet lines in a building basement and uplink them over a fiber connection to an access POP in the network. These haven’t been designed to operate in a ring just yet, but the new Alpine line of switches from Extreme are being “ruggedized” for the metro, with latency and QOS support. Extreme is doing well in ISP POPs and the enterprise, and is now trying to get in on the metro opportunity. Yipes has been its first big win.

(More about Extreme on Extreme Launches A Sonet Killer.)

Force10 Networks

Force 10 Networks Inc., a startup built by a group of network processor and IP QOS wizards, was formed to create carrier-class 10-gigabit Ethernet solutions. From the looks of their early presentations, this company is building not a metro optical Ethernet solution, but a Layer3 switch with 10-gigabit Ethernet capabilities. They may see more action at the core of the network, where that kind of scale can be used to aggregate and transport gigabit Ethernet from POP to POP.

(More about Force 10 on Force 10 Raises $50 Million.)

Foundry Networks

Foundry Networks Inc. (Nasdaq: FDRY), part of the Layer 3 switch crowd, has recently been upgrading its switch line to support more MAN-optimized networking capabilities. A recent win is Telseon, which will use Foundry’s BigIron Layer 3 gigabit IP switches to increase metro network performance. The switch supports up to 120 gigabit Ethernet ports and handles 178 million packets of data per second. It supports up to 480 Gbit/s of total switching capacity, with wire-speed IP switching in all ports. Foundry's Internet optical networking solutions include long-haul gigabit Ethernet up to 150km, OSPF (open shortest path first), BGP4 (border gateway protocol 4) routing protocol implementation, super-VLAN aggregation over the metro-area backbone, and rate limiting and traffic shaping features.

(More about Foundry on Foundry Enters WAN Ethernet Fray andJudgment Day for Foundry Core Router.)

Lantern Communications

Lantern Communications was one of the first startups to talk about metro optical Ethernet networks as equivalent in architecture and functionality to Sonet networks. Many Layer 2/3 switch makers have asserted their MAN capabilities, but Lantern is designing its product from day one to be deployed in a survivable ring architecture, mimicking as closely as possible the way in which existing Sonet access networks are deployed.

Lantern’s solution supports deterministic delivery of QOS, to support SLAs defined for each customer of the network. Lantern has developed a technology called “virtual flow queuing,” which controls delay through the network. The distributed switch architecture allows granular control of individual IP flows or aggregated flows in the network. Flows can be configured for committed and peak bandwidth, burst, and delay.

The Lantern solution, like those of Luminous, Nortel, and Cisco, employs some version of the emerging Resilient Packet Ring specification for packet-layer survivability in a metro ring architecture. Lantern’s solution is based on 10-gigabit Ethernet as a transport mechanism, with each ring supporting a total throughput of 40 Gbit/s over 200km.

(More about Lantern on Lantern Lights Path to Gig Ethernet.)

Luminous Networks

Luminous Networks Inc. is in the ever-widening camp of metro Ethernet players that are optimizing their systems for fiber ring architectures. The Luminous solution is an IP-optimized, resilient packet transport (RPT) solution that provides differentiated services and preserves the legacy circuit services. This solution integrates IP, Ethernet, and DWDM functionality for advanced capabilities at a fundamentally lower cost point.

Luminous achieves sub-50 millisecond protection switching in a packet ring that can scale to 254 nodes. Its RPT solution allows both protected and unprotected services on the ring, and it is the only pure-packet company claiming to provide synchronous TDM circuits such as T1/E1. In other words, they can stratum-time the ring from an external source and provide a DS1 (1.5 Mbit/s) that is synchronous at both ends over 128 nodes.

Luminous has developed sophisticated QOS capabilities for data, including packet classification, limiting/policing, and shaping. The boxes can set up Ethernet "wire services" that are configurable in increments of 64 kbit/s and that can be assigned EF, AF, or BE Diffserv classes. The performance of these "virtual wires" can be monitored and verified as SLAs, and billing info can be sent to external billing systems. Like a Sonet ring, the system can monitor the performance of the network, and deliver detailed alarms and stats to the Luminous Management System for fault detection and isolation, and/or to external OSSs.

Luminous can also perform routing on traffic ingress, with cut-through forwarding to egress on the system, such that the entire PacketWave network of any number of nodes appears as a single router to external systems.

(More about Luminous on Luminous PacketWave Completes Beta ,Luminous Launches Optical Access Switch,Luminous Raises $60M, and Scientific-Atlanta Invests $10M in Luminous.)

Native Networks

Native Networks is a recently minted startup, funded by a variety of venture capitalists in Israel and Soros Equity Partners. Native Networks is foundedby former members of Orkit, and Tadiran (now ECI Telecom), and has developed a distribution and access solution called EtherMux-M. This device is billed as an ³Add/Drop statistical Multiplexer (ADsM)² for the access network. Like other metro optical Ethernet vendors, Native is adding intelligent QoS mechanisms to Ethernet to make it carrier grade and multiservice capable. EtherMux-M features granular provisioning, MPLS-based traffic engineering, survivability., and integrated network management capabilities. In addition to Ethernet services, EtherMux-M enables provisioning of native T1/E1 circuit-switched services over the access network. EtherMux-M providesmultipoint-to-multipoint SM-TLS (shared media transparent LAN service). Using Native Networks Virtual Trunks, sites can be assigned different SLA and QoS levels while maintaining full isolation between different end-user communities.

Nortel Networks

Nortel Networks Corp. (NYSE/Toronto: NT) is bringing Ethernet to the metro under the wings of its Local Internet initiative. Today this initiative is basically a bundling of Nortel products into a complete solution, creating a next-gen Sonet-based metro Ethernet network that can support Ethernet services in granularity of 1 Mbit/s.

The products used to build the network today include the OPTera Packet Edge System (gigabit Ethernet interfaces), OPTera Metro 3000 Multiservice Platform, Passport 8600 Routing Switch, Business Policy Switches, Shasta 5000 Broadband Access Node, and Preside NMS. If a carrier were to just use the Passport switches and the Shasta 5000 they could create true Ethernet over fiber, but rumor has it Nortel has a new equipment line in store for pure-play metro optical Ethernet.

Nortel is also big on 10-gigabit Ethernet and has already sold 10-gigabit Ethernet equipment to two carriers, Bell Nexxia in Canada and Utfors in Scandinavia. Nortel has been involved in 10-gigabit Ethernet for a while now and is well positioned to address that market as effectively as the startups.

(More about Nortel’s metro Ethernet developments on Shasta La Vista and Gigabit Ethernet Goes Carrier Class and on Nortel’s Web site, at

Riverstone Networks

Riverstone Networks is supporting metro optical Ethernet through its RS 3000 and RS 8000 Switch Routers, which are already carrying live Ethernet traffic in Telseon networks. These are true routers and differ from many of the metro Ethernet players, which are Layer 2 devices only, leaving routing to the Ciscos and Junipers in the network.

Riverstone, like many of the other metro Ethernet players, has built in technology that allows metro service providers to offer such services as customer self-provisioning; bandwidth that can be dynamically controlled for specific customers, applications, or users; and server load balancing. Riverstone's RapidOS operating system makes it possible for providers to capture and precisely bill for bandwidth and application usage.

Riverstone’s metro customers include Telseon, Urban Media Inc., and IntelliSpace.

(More about Riverstone on Cabletron Floats Riverstone andRiverstone Flows Toward Gigabit Ethernet.)

Tropic Networks

Formed in May 2000 by a team of optical and networking engineers and executives from Nortel Networks and Newbridge Networks Corp. (NYSE: NN; Toronto: NNC), Tropic Networks Inc., is an optical networking startup based in Ottawa and Boston, developing a next-generation optical IP services platform that integrates packet and photonic capabilities, enabling metro carriers to rapidly deploy high-bandwidth Ethernet- and IP-enabled services. Tropic’s product, slated for release in late 2001, is designed on a single, packet-based network model that collapses the layers and integrates data services and optical transport.

(More about Tropic on Tropic Gets $10 Million First Round.)

WorldWide Packets

World Wide Packets Inc. has been developing gigabit Ethernet solutions and is currently involved in taking that to 10-gigabit Ethernet. For the MAN market, WorldWide Packets is about to unveil its Ethernet-to-the-Subscriber solution, which is part of its LightningEdge infrastructure product line.

This line includes: 1) the LightningEdge Access Portal, providing a Gigabit of dedicated bandwidth for simultaneous transmission of voice, video, and data; 2) the LightningEdge Access Distributor, allowing service providers to connect with subscribers over standard gigabit Ethernet or 10-gigabit Ethernet links; and 3) the LightningEdge Network Supervisor, management and provisioning software allowing voice, video, and data to become provisioning options from the service provider to their subscriber base, rather than individual applications from separate providers.

(More about WorldWide Packets on Bernard Daines,World Wide Packets Product in Trials, andWorld Wide Packets Debuts First Product.)Metro optical Ethernet is more than just Ethernet-based access rings. In my report for Pioneer Consulting, "Optical Edge Networks," I forecast the market for “metro optical IP” to start slowly but reach $1 billion in 2004, still well below that of the competing next-generation Sonet market (see Report Forecasts Metro Winners, Losers ). At the same time, I indicated it may well be the “sleeping giant” of the MAN market. Why? Because metro optical Ethernet is more than just a new flavor of access rings. It is a complete network substitution technology and has many more applications than simply collecting traffic from the basements of buildings.

As we’ve described in this report, metro Ethernet can be deployed into metro inter-POP rings, cable TV data network backbones, private enterprise campus networks, storage area networking backbones, and private MPEG video distribution networks. In fact, the inter-POP market is likely to be an early and substantial driver of the optical Ethernet market because of the elegance of that kind of implementation. As 10-gigabit Ethernet rolls out, there will be a substantial market for gigabit Ethernet aggregation and transport between metro and core POPs in carrier IP backbones.

As a guide, it is worth considering how rapidly Internet data centers are being constructed. Today the major data center operators, AT&T Corp. (NYSE: T), Digex Inc. (Nasdaq: DIGX), Exodus Communications Inc. (Nasdaq: EXDS), GlobalCenter Inc., PSInet Inc. (Nasdaq: PSIX), and Qwest, are rapidly building and expanding their facilities — to upwards of 60 million square feet by the end of 2001. Add to that the plethora of colocation companies like and LayerOne Inc. building metro hubs that all service providers can share. Level 3 is doing much the same, though over their entire network, and has already made it clear they would love a 10-gigabit Ethernet solution that allowed them to transport data between their hubs at the lowest cost possible.

The result? Gigabit and 10-gigabit Ethernet as a transport mechanism could ultimately create a $5 billion equipment market in the metro area, providing the common interface, aggregation, and transport technology for high-bandwidth IP services. The service market opportunity is even larger, with an addressable market of roughly $5 billion today for fiber-based private lines and over $20 billion in 2004.

The key to this market’s expansion, obviously, is dark fiber. Metro optical Ethernet networks are most easily created using dark fiber, and the current emphasis on deploying dark fiber into metros in both North America and Europe is setting the stage for a long-lasting metro Ethernet opportunity. This will not be limited to the more nimble CLECs of today — with access to wholesale dark, even the most traditional LECs may deploy metro optical Ethernet systems in a tactical fashion to fend off encroaching competitors.

The end result? Ethernet is here to stay, and will be showing up in every kind of network over the next few years, even in the home. We haven’t said much about home networking here, but managed Ethernet services may start arriving at your doorstep sooner than you think. It may be just the thing to kickstart the oft-delayed fiber-to-the-home market. With gigabit Ethernet everywhere else, why not bring it home?

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like