Welcome to the hottest market in networking: edge routing * The players * The products * The prospects

April 24, 2002

32 Min Read
Taking Routing to the Edge

Aggregation or aggravation? If you're Cisco Systems Inc. (Nasdaq: CSCO), the industry's preeminent edge router vendor, then you're all too familiar with both.

As the alpha male of the pack – with about 74 percent of the $924 million in edge routers sold in North American in 2001 – Cisco knows the business of IP traffic aggregation better than anyone.

But the 800-pound gorilla of the router industry is rapidly becoming acquainted with the aggravation factor, as well, as an army of companies looks to steal the edge router market out from under it.



Dynamic Table: Edge Router Vendors

Select fields:
Show All Fields
Company NameURLTypeLead VCsTotal FundingProduct NameWAN Interfaces ProtocolsMax. T1s Per ChassisMPLS VPNsIPSec VPNsVOIPFirewall ServicesTraffic ShapingMPLSRPRNEBS Certified

The arrival of so many new players has caused a great deal of confusion about the edge routing market – with seemingly as many opinions about what makes a good edge router as there are vendors making them.

The following report attempts to clear up the confusion by providing overviews of both the edge router market and edge router technology, and then drilling down to examine the following issues in more detail:

  • Who’s playing in this market (and who’s got a shot of making it big)

  • What edge routers are – and what they are not. (Hint: They’re not multiservice switches or MPLS switches)

  • How edge routers save money using new, smaller “high touch” form factors that are optimized to aggregate heterogeneous traffic at the edge of the network

  • How edge routers can convert capacity in the core into hard currency via state-of-the-art technologies such as virtual routing and differentiated services

Feel free to read the article sequentially, or use the links below to jump to the specific information you need.

Hot! Hot! Hot!
Edge routers have remained relatively immune to the capex crisis, and the market continues to grow.

Public Players
Cisco dominates, but Juniper, Riverstone, and Unisphere are coming up fast, and new entrants are cropping up.

The Startup Scene
Analysts differ among themselves, but many feel a shakeout is inevitable in this crowded space.

Complicated Technology
Or, at least, complicated marketing claims make comparison shopping tricky. We're here to help.

What They Ain't
Be they neither multiservice switches nor MPLS switches.

History Lesson
The Edge has always been with us – but it's gone through some changes.

Critical Features List
Some of the key criteria carriers may want to take into consideration when shopping on the edge.

It’s the Services, Stupid!
Where the money is.

— Jim Carr and Stephen Saunders, Founding Editor, Light Reading
http://www.lightreading.com

Jim Carr is an Aptos, Calif., freelance business and technology writer. He can be reached at [email protected]

There’s no disputing the fact that edge routers are hot.

The graphic below, from a recent Light Reading report, shows that edge routers were hot when carriers were building out their networks. They are staying that way during the current period of financial uncertainty – when carriers are looking to optimize their existing networks, rather than grow new ones. What’s more, they will still be hot once we move to the next phase of network development – when next-generation gear will really start to take off in service provider networks.

14346_1.gifThis puts edge routers in a unique position versus other technology categories, the popularity of which tends to fluctuate based on the carrier fiscal environment.

Why is this?

A lot of it has to do with carrier capital expenditure – or how much money service providers are spending with equipment vendors, and on what type of equipment. The networking industry is, after all, a Great Chain of Being – with service providers at the top. When they spend money, it goes to equipment manufacturers, which in turn spend money with component vendors.

The next graphic, which comes from a report written by Light Reading’s paid research division, the Optical Oracle, shows the capex environment up to the end of the last century (or “BBB,” Before the Bubble Burst). It depicts an unsustainable model of investment. Carriers were spending too much money, with too little return:

14346_2.gifCarriers subsequently awoke to the error of their ways, slashed capex spending... and the bubble burst.

The next figure, also from Optical Oracle, shows the situation “post-bubble.” It summarizes capex plans for the 12 leading service providers in the U.S. and shows that they are going to continue slashing capex for the foreseeable future:

14346_3.gifWhat makes edge routers unique is that, while they are not immune to the vicissitudes of carrier capex, they are less susceptible to it than are other equipment categories. And that’s what the next slide shows:

14346_4.gifOn the top, the bad news: Compared to 2000, 2001 was a disastrous year for networking equipment vendors in general, and router folks took it on the chin just like everyone else – with revenues actually dropping in 2001 and only picking up slightly this year. RHK Inc. switching and routing analyst Kelly Dougherty blames the steep decline in 2000 on the fact that service providers "were ratcheting down capital expenses and looking to reduce excess capacity in their networks."

On the bottom, the good news: Sales of IP aggregation routers (edge routers) represent a disproportionate amount of overall router sales. In fact, RHK is forecasting that by 2004, sales of edge routers will have leapfrogged sales of core routers, to account for nearly 60 percent of router vendors’ revenues – compared to 37 percent in the year 2000.

So why is it that edge routers have remained relatively immune to the capex crisis? And why is the edge router market growing, when everything else is contracting?

It turns out that while most networking vendors (and investors) spent the last three years fixated on finding ways to build optical networking products that could add massive amounts of bandwidth to service provider networks, the changing economic climate meant that service providers themselves have been searching – with a sense of increasing desperation – for devices that would allow them to do more with the capacity they already have.

Hence the recent rampant interest and activity in edge routing.

Vendors (and, again, investors) have finally caught on en masse to the need for innovative products designed to help carriers build out the edges of their networks with equipment that costs less to buy and run and simultaneously positions them to deliver a bundle of new revenue-generating services in the future. (This has resulted in some noteworthy strategic about-faces. Redback Networks Inc. [Nasdaq: RBAK], for example, recently refocused its R&D from next-gen optical equipment to edge routing.)

The result is that edge router vendors are now bringing to market a slew of products designed to deliver a bulletproof “double business case”: both saving money for service providers (via smaller, faster, cheaper form factors), and making it for them, through support for revenue generating services that include – but are by no means limited to – application-based quality of service (QOS) and virtual private networks (VPNs) based on IPSec or Multiprotocol Label Switching (MPLS).

As noted, the upshot of all this is that market researchers expect service providers' spending on edge equipment to leapfrog spending on their overbuilt core networks.



Editor's Note: Light Reading is not affiliated with Oracle Corporation.The potential size of the edge market has prompted a mass of companies to develop router products that target it. At least 15 are making significant plays at the edge. Others are probably out there, but still in stealth mode.

These companies are split into two camps: public companies (incumbent equipment manufacturers) and startups.

Among the public companies, Cisco dominates the edge router market, although all three of the market research firms that provided information for this report agree that two other companies – Juniper Networks Inc. (Nasdaq: JNPR), and privately held Unisphere Networks Inc. – are making sizeable inroads into its territory. In fact, both RHK and Synergy Research Inc. have released surprisingly consistent reports (for once) that back up claims by Juniper and Unisphere that they're eking out key wins against Cisco at the edge.

RHK estimates that Cisco Systems owned 74 percent of the North American edge router market in 2001. It says Juniper accounted for 19 percent of the same market, up considerably from the 6 percent it held in 2000, while Unisphere Networks had 7 percent in 2001, up from... well, nothing, basically, the previous year.

Synergy Research, for its part, notes that Cisco had 71.2 percent of the edge router market in 2001 (down from over 90 percent in 2000), with Juniper grabbing 14 percent and Unisphere 13.8 percent. That gives the trio a whopping 99 percent of what Synergy defines as a $1.33 billion 2001 edge router industry.

14346_5.gifFourth-quarter 2001 numbers from a third market research firm, Infonetics Research Inc., roughly tally with those from RHK and Synergy when it comes to Juniper and Unisphere (pegging them at 11 percent market share each). However, they show Cisco's share of the pie as substantially smaller (56 percent) and add a fourth public company to the mix: Riverstone Networks Inc. (Nasdaq: RSTN), which Infonetics credits with 10 percent of the market.

14346_9.gif “Juniper and Unisphere have made headway against Cisco not so much by taking customers away from them, but by selling edge router solutions to new customers," says Erik Keith, an analyst with Current Analysis.

“Behind the Big Three," Keith notes, "there are several established vendors with edge router platforms with compelling density and scaleability propositions and/or multiservice capabilities.”

The main ones are:

Analysts believe some of these “second tier” vendors have a good shot at making serious money in the edge router market. Keith believes Nokia could offer an especially compelling argument. It has what he calls a global "IP strategy" and a "very small box that costs less than [the Cisco, Juniper, and Unisphere products] to buy and scale incrementally."

Both Nokia and Motorola got into the edge router market via acquisitions: Nokia when it bought Amber Networks for $421 million in July 2001 (see Nokia Nabs Amber for $421M) and Motorola when it bought River Delta Networks for $300 million, also in July 2001.

While its stock has fallen out of favor with Wall Street, Redback is being cited as a possible diamond in the rough by another analyst house, Gartner/Dataquest. The company began shipping its SmartEdge 800 edge router last fall, but it already had a presence at the edge with its SMS DSL aggregation and subscriber management product line, points out Jennifer Liscom, the principal switching and routing analyst at Gartner. She says Redback's strong customer base with the SMS product gives it a "good chance" of picking up edge router wins.

In addition to the small number of revenue-generating SmartEdge 800 routers shipped in the fourth quarter of 2001, Redback also claims the product is in trials with 15 service providers. Still, if Redback's last earnings report is anything to go by, it'll be a while before the SmartEdge pays off in any major way (see Redback's SmartEdge Still No Savior).

Meanwhile, like Unisphere, Riverstone says it has made strong gains in the Korean market, where it says it has sold more than a thousand of its RS-series of "metro edge" routers to PowerComm, a subsidiary of KEPCO, South Korea's national electric power company.

Both Redback and Riverstone dispute the market share numbers from RHK and Synergy (not surprising, given that neither vendor registered on these market researchers' edgeometers™).

Redback points out that it has now upgraded its SMS subscriber management platform with IP routing capabilities that allow it to act as an edge router solution. If the SMS were factored into market researchers’ numbers, Redback contends, it would show up as having a much more substantial market share (see Redback Rallies Itself).

So much for the incumbents. On the startup side, tons of venture money is flowing into edge router outfits. VC firms have bequeathed nearly a half-billion dollars ($476 million, to be exact) to seven new players:

14346_6.gifThat's big money in an economy that no longer views high-tech startups through the rose-tinted lorgnettes it was wearing a few years ago.

These outfits claim to have a lead over the market incumbents. They say they're building edge routers that deliver new, denser (and cheaper) architectures that better lend themselves to the requirements at the network edge.

Atop the funding list is Celox Networks, with $155 million in VC money. Still, despite being wealthier than anyone else, Celox has gotten off to a rocky start (see Celox Battens Down the Hatch). Allegro Networks Inc. has collected $84 million; Laurel Networks Inc. $83 million (see Laurel Moves to Phase Three); and Crescent Networks Inc. $55 million.

VCs say investing in the edge is a no-brainer [ed.note: that’s lucky – and apt – given their recent investment record].

"We looked at carrier spending and tried to figure out what sector would have growth. When we looked at the buildout of the backbone of the last several years, we felt [the edge] was a good place to invest because that's where [providers] needed more functionality and capabilities," says Andy “IP” Sessions, a partner with Thomas Weisel Partners, a lead investor in Allegro.

Sessions says his firm invested in Allegro in large part because of its CEO, David House – formerly CEO of Bay Networks (see Allegro Holds a House Party ). "A lot of [startups] have products, but unless you have the experience with carriers, it just doesn't matter." Sessions adds that management experience is also one reason his firm invested in another edge router startup, Celox: Its president and CEO Kent Mathy is a former AT&T Global Network Services executive.

With so many startups playing in this area, some say a shakeout is inevitable. Light Reading recently polled its readership to see how bad they think it will be (see Edge Routers). The largest proportion of the 149 respondents (32 percent) believes that the edge router market can only support three companies. When asked what type of company is most likely to emerge as the leader in edge routing, almost 70 percent believe it will be an incumbent public company – not a startup.

Analyst opinion is divided on this subject.

RHK's Dougherty is bearish [grrrrrrr!]. She says that the fact that sales cycles are slowing helps Cisco maintain its lead: “It makes it difficult for other companies, especially newcomers, to make further market share gains.”

Keith of Current Analysis, on the other hand, is more bullish [mooooooo!] on the startups’ prospects. He believes Cisco could become the "Alcatel" of the edge router market. That is, he figures Cisco will consistently collect 45 percent to 50 percent of annual edge router sales (as Alcatel SA does in the DSLAM market), with the remainder split among three or four others, startups among them. That would be a far cry from Cisco's heyday as the undisputed Supreme Being in the edge router space.

Clearly, edge routing is a hot topic. But from a product technology perspective, it’s also a horribly complicated one.

This complexity is aggravated by the fact that each of the edge router vendors has its own impressive (and persuasive) glossy PowerPoint pitch.

The maddening thing about these presentations is that they invariably use different terms to describe the same technology features [insert sound of author sighing here]; and they put forth wildly contradictory claims about how their USPs (“unique selling points”) measure up against the competition [another sigh, this one deeper].

  • Celox claims to be among the first vendors to market an OC192 interface for its "service creation switch"...

  • Axerra Networks Inc. says its AXN1600 defies conventional networking definition, representing “a unique product category... the multiservice IP concentrator”...

  • Crescent says a single VRX-1000 Service Edge Router supports “more than a thousand virtual routers and a million-plus routes with 120 million packet-per-second throughput”...

  • Laurel claims its ST200 Service Edge Router can map “any type of traffic” and encapsulate it into MPLS for delivery across the core...

  • Net.com boasts a platform that “combines aggregation, subscriber management, IP service creation, routing, multiservice switching...”

  • Then there’s Allegro, which won’t say anything about its product. At all. (Perhaps we should be grateful.)



All this makes it hard to make apples-to-apples comparisons of the different products.

Since vendors can’t make up their minds about what can or should constitute an edge router, it’s critically important that service providers evaluating this market establish their own set of common criteria, which they can then apply to any or all of the edge router products in order to conduct some objective competitive analysis.

That means stepping back from this marketing mess and doing three things:

  • Working out what edge routers are not (bear with us here);

  • Realizing that edge routers are not a new phenomenon (whatever the startup marketeers say), that they have been around for years, and working out how and why edge routing equipment, and the types of applications it is used for, are changing;

  • Taking a step-by-step tour of the latest edge router features, working out how they work, why they are important, and which vendors’ devices support what.

This is pretty much what the rest of this report will do, though, obviously, this is not intended to be a “carrier-class” assessment of these products (to do that you will need a lab, 60 engineers, a box of pocket protectors, three years, an illegal still in the woods, and approximately $14.5 million in cold hard cash).

Before working out precisely what edge routers are, it’s important to realize what they are not.

Specifically, edge routers aren’t the same as another significant equipment category: multiservice switches.

Distinguishing between the two is important but difficult. In fact, it’s a task that even the experts find challenging. "I'm hard-pressed to make the distinction between a router and a multiservice edge switch," shrugs Michael Kennedy, managing partner of Network Strategy Partners LLC, a networking equipment management consultancy.

In simple terms, the main difference between multiservice switches and edge routers is that edge routers are based on packet-switching architectures (using an IP control plane). Conversely, multiservice switches, while they also support Internet Protocol (IP), are usually based on Asynchronous Transfer Mode (ATM) architectures and feature cell-based control planes (see Laurel Moves to Phase Three).

In fact, some edge router vendors question the need for a new term to describe the multiservice switch category at all. “These are ATM switches. Why not just call them that? Anything else is just confusing to customers,” says Troy Dixler, founding engineer at Allegro Networks. “Edge routers support multiple services, too. But do we go around calling our product a ‘multiservice edge router’? No, we do not.”

To make things just that little bit more confusing, the market for multiservice switches is itself divided into two parts: multiservice switches for the edge of the network and multiservice switches for the core. Some vendors' products play in one camp, some in the other, and still others in both.

Here’s how it breaks down:

Currently, the edge multiservice market is dominated by startups like Gotham Networks and WaveSmith Networks Inc. However, publicly held Advanced Fibre Communications Inc. (AFC) (Nasdaq: AFCI) recently made a bold move into this space when it bought startup AccessLAN Communications Inc. last month (March 2002) for $43 million.

Meantime, it is the giant public vendors that currently dominate the core multiservice market: Alcatel, Cisco, Lucent Technologies Inc. (NYSE: LU), Marconi PLC (Nasdaq/London: MONI), and Nortel Networks Corp. (NYSE/Toronto: NT) (see Switch Vendors to Tackle Core Routing). These companies generally claim that their core equipment may also be installed in edge environments.

Then there’s Équipe Communications Corp., a startup, which says that its product is designed to serve only in the core – not the edge. (Équipe recently went through some layoffs before finally closing a third round of funding in March – see Équipe Partners With Ciena, Scores $40M). As a rule of thumb, multiservice switches should hold the most allure to RBOCs that want to migrate an extant ATM infrastructure towards IP, in increments, via cheaper ATM-based hardware. On the other hand, this type of kit is less likely to be attractive for carriers building out IP networks. (Note: Light Reading will be publishing a report on multiservice switches in the first week of June 2002.)

Edge routers also should not be confused with another equipment category: the MPLS switch. This class of device, pioneered by companies like Maple Optical Systems and Tenor Networks Inc., supports neither full IP routing, nor ATM switching. MPLS switches are supposed to act as “toll booths” between metro and core networks, allowing service providers to bill for new services (see Maple Turns Over a New Leaf). They’ve so far proven to be about as popular as a fart at a funeral.

Despite what the startups’ marketing literature may say, edge routing itself isn’t new. Service providers – and enterprises – have needed edge routing capabilities ever since their networks have had edges (you know, back when the earth was without form, and void).

In fact, it’s possible to detect three distinct phases in the development of edge router technology.

In the first, specialized edge router products didn’t exist. Instead, Cisco, which owned the market, outfitted its general-purpose router lines with the appropriate blades to handle edge routing applications.

That’s changed. In fact, Cisco now sells no fewer than four router products into the edge market:

  • 12000 Series Internet Router, outfitted with router interfaces appropriate for edge apps (a high-end product)

  • 10000 Series Internet Router (a mid-range product)

  • 7600 Series Internet Router (formerly the 6500 Catalyst LAN switch, but now updated with some extra nips and tucks; mid-range)

  • 7200 Series Router (a low-end solution)

Phase two occurred about three years ago, when the first edge routers (or specialized devices designed with the edge in mind) appeared. As the diagram below shows, the first edge routers were small devices designed with a few specific applications in mind, notably DSL and leased line aggregation:

14346_7.gifUnisphere was one of the pioneers of this approach with its ERX product, which it acquired when it bought Redstone Communications in 1999. The ERX was a low-end device designed with a specific application in mind: DSL aggregation (or consolidating multiple low-speed DSL traffic streams from a DSLAM into a T1 circuit).

And it was a big hit with service providers, especially in Asia, where government mandates for broadband Internet access have driven router sales. "By the end of last year, 40 percent of the Korean population had broadband access," says Current Analysis’s Keith.

The ERX was designed three years ago. And technology trends within the edge router market have changed since then. DSL aggregation is still a significant feature in some products; Redback, Cisco, Tiara Networks Inc., and Unisphere all support it.

But these days, a third phase of edge routing has commenced. Now the trend is toward building bigger, multiservice edge routers targeting the consolidation of traditional Layer 2 services – ATM, Ethernet, and in some cases Frame Relay – over an IP backbone, usually via MPLS.

14346_8.gifThis trend is being played out in Unisphere’s product portfolio, incidentally. It recently launched the MRX Multi-Service Edge Router, a successor to the ERX (see Unisphere Cutting the Edge). Similarly, Redback now offers the SmartEdge router, in addition to its IP-enabled SMS platform.

Another new trend: The rapid deployment of new metro networks running Gigabit Ethernet, natively, has prompted some companies to focus efforts on developing edge routers that support these interfaces.

Riverstone Networks, in particular, has staked a claim in this area. But its product is an edge anomaly, in some ways. While the vendor has made much noise about how its edge router provides an ideal platform for supporting revenue-generating IP services, it does not support some IP features – such as IPSec VPNs. Despite this, the company has found success, which suggests that what most customers are really buying from Riverstone is Layer 2 Ethernet switching – not edge routing. (Note: Riverstone disputes this analysis, vigorously).

As noted, a service provider evaluating edge router technology needs to make its own list of the features it considers most important, and then hold it up to the products on offer, in order to be able to conduct objective competitive analysis.

Light Reading has had a go at coming up with such a list here. We’re calling it our “Critical Features List” (because we can).

From the top, then:

1) Network Processor Architecture

This feature is crucial to network performance. Traditional routers were based on centralized processor architectures, creating a significant performance drop-off when they were asked to handle advanced functions such as IPSec.

The latest edge router products, in contrast, generally feature variations of distributed-processor schemes and ASICs to deliver value-added IP services without degrading router performance

Network Strategy Partners' Kennedy calls the architectural advances taking place "a bit like the arms race." Buyers have to balance the need to deploy "tried and true products that work well, versus the latest, new technology," which, he notes, generally comes at a premium. That makes introducing new technology "a question of timing. If [a vendor] falls behind, that's big trouble because the rate of innovation is so rapid. That's been a big challenge for big companies like Cisco in that they have difficulty innovating at the rate of startups."

Allegro's so-called "IP multi-router," which will provide multiple router "blades" in a single box, is a prime example of such a new, ASIC-driven product. Allegro's claim to fame is that it dedicates separate processing resources to each I/O blade, forming multiple individual routers rather than creating numerous virtual routers all sharing the same computing power.

The Allegro box's ability to "add processing power as you add more functionality is intriguing," says Chris Liljenstolpe, a senior director for network technology at ISP Cable & Wireless (NYSE: CWP). Liljenstolpe is evaluating Allegro's product for its potential scaleability and is also “looking for edge routers with multiservice edge capabilities, such as encapsulating Layer 2 data services into IP or MPLS."

Unisphere also places ASICs on line cards for its MRX Edge Router. These "manage queues and buffers, count and classify packets, and schedule packets for quality of service between multiple queues at wire speed," says Chris Lawler, vice president and general manager of the company’s IP routing business unit. This would be impossible with a software-based approach to delivering services, he explains.

Steve Byers, Axerra Network's VP of marketing, says the combination of multiple PowerPC processors with ASICs in the company's AXN Multiservice IP Concentrator is at the heart of the product's "service adaptation technology," which encapsulates incoming traffic types for transmission over an IP network. This scheme allows the product to support multiple types of traffic – frame relay, ATM, IP/PPP, leased lines, PCM voice, VOIP – forming what the company calls a single "unified IP" network core.

Cisco and Juniper have reacted to the startups' challenges by developing programmable ASICs for their edge products. Unlike standard ASICs, which are designed much like a PC’s CPU, to handle only preset functions, programmable ASICs, like RAM, are software-upgradeable. This allows field-upgrading of the units when customers want to add features not originally provided by the vendor ASIC, notes Robert Redford, Cisco's VP of marketing for Internet routing.

2) Control Plane Processing

While a product’s network processor architecture dictates its performance, control plane processing dictates its IQ – its ability to handle complex tasks.

Want to know more? So do we. Unfortunately, discussions of control plane architectures are about as easy to understand as the human genome project, and it would not do them justice to précis them here. For more information, check out the following links:

3) Form Factor

Form factor covers two types of physical attribute.

First: physical reliability – things like redundant components and power supplies, hot-swappable interface cards.

Second: port density. The newest edge routers are very compact – in fact, in some cases, the port density in these next-generation products means that a single device, occupying one chassis, can replace scores of conventional routers in a data center.

Product density is a driving force behind the interest at least one potential Allegro customer is showing the company's as-yet unreleased IP Multi-router. Peter Juffernholz, senior manager of IP connectivity services for ISP Teleglobe (NYSE, Toronto: BCE), finds the promise of Allegro's product to support multiple WAN interfaces (which translates into thousands of T1 ports in a single box) compelling. He says that two of Allegro's boxes, if they produce as promised when available, will replace 30 or 40 conventional routers.

That's critical because, he says, "In some locations, we're in a major carrier's 'hotel,' where it gets crowded, so space is essential. If I can reduce from five or 10 racks to one, I'll have room to grow." It's also "cheaper from a buying perspective and from local-loop management, because I can make better use of my high-speed circuits."

Until Allegro's device ships, however, Laurel Networks' ST200 service edge router and Gotham Networks' GN 1600, either of which can support up to 5,376 T1 lines in a single chassis, are the leaders of the pack. In a seven-foot rack, with three chassis per rack, that amounts to 16,128 T1 lines.

Laurel claims its ST200, which is also based on a distributed architecture, is the first edge router to offer channelized OC48/STM16 interfaces that support a variable mix of DS3, OC3c, OC12c, and/or OC48c channels simultaneously.

Crescent Network’s VRX-1000 Service Edge Router also offers a dense package, supporting 12,000 incoming T1s per seven-foot equipment rack.

4) Network Reliability

Network-layer reliability is another feature to look at, especially when evaluating equipment that supports Ethernet-over-WAN interfaces.

A key issue with Ethernet is that it’s not optimized for the WAN.To increase the durability of Ethernet, some vendors are implementing a new specification, called Resilient Packet Ring (RPR), which is being developed by the Institute of Electrical and Electronics Engineers Inc. (IEEE)’s 802.17 committee. The spec calls for developing a new MAC layer that will provide protection for all ring-based transport, including Ethernet, through the use of standard ring-wrapping techniques to switch around a fiber cut.

It’s worth noting that some edge router vendors don’t think RPR is a necessary feature of edge routers – especially now, before the standards are finalized. Although the Resilient Packet Ring Alliance and the IEEE have agreed on a single baseline draft for the standard, it is far from final. Consequently, few of the edge router vendors have incorporated support for a non-standard version of RPR into their products. Juniper says it is waiting for a standards-compliant RPR before supporting it. Aplion Networks Inc., Axerra, Celox, Laurel, Tiara, and Unisphere are among the others who've abstained.

Conversely, Cisco (one of the leading proponents in the RPR standardization effort), Crescent, Redback, and Riverstone all claim to offer some level of RPR support.

5) Software Reliability

Equally as important as physical and network reliability is software reliability. Writing reliable router code takes hundreds of thousands of man-hours – and many millions of dollars. This is one area where incumbents, such as Cisco and Juniper, have a huge advantage over new players, because they have the resources to support this effort on an ongoing basis.

6) Traffic Aggregation

As a recent Light Reading poll shows, many consider the ability to aggregate incoming traffic as the single most important edge router attribute (see Edge Routers). Fortunately, all of the edge router products on the market not only support multiple protocols, but also have the ability to aggregate different types of protocol (voice, Ethernet, ATM, IP, and so on) over a single line, usually using MPLS (see below), and thus reducing service providers’ opex.

7) MPLS

Multiprotocol Label Switching (MPLS) allows a carrier to make more efficient use of its IP network capacity. Basically, it assigns labels to a flow of packets, such as a VOIP stream, so intermediate routers in the network recognize the group without having to inspect individual IP headers. This allows the network to be divvied up for different services and different clients, and enables more rapid packet forwarding.

While lacking several final lower-level standards, MPLS is moving steadily toward standardization. The MPLS Forum says that one of its technical committees has made "significant progress" in developing an MPLS UNI that will standardize the way customer premises equipment connects to service provider networks. In addition, that group also says it has completed a voice-over-MPLS implementation agreement, which provides a standards-based method for service providers to offer voice-over-packet services to interconnect voice media gateways over MPLS.

Virtually all edge router vendors offer MPLS support in their new boxes. Only net.com and Tiara say they don’t. To date, however, the greatest interest for MPLS has been in the core, where it is used for traffic engineering – that is, managing traffic within service providers' internal networks.

RHK's Dougherty is not swept away by all the "vendor hype" about MPLS. "We don't see service providers pounding on the door for MPLS: Most expect to have MPLS in their network at some point, but they're still evaluating it.”

8) QOS

Quality of service is also an important consideration. Vendors use a variety of methods, standard and proprietary, to enforce quality of service over IP networks for time-sensitive or mission-critical apps.

9) OSMINE

Next up is Osmine, a program administered by Telcordia Technologies Inc., which is supposed to ensure that network equipment deployed by service providers can connect to billing, provisioning, and other back-office systems – the operation support systems or OSSs – deployed by major carriers.Getting Osmine certification, unfortunately, costs millions of dollars and requires vendors to dedicate resources and personnel to manage the process and work with Telcordia. Osmine testing is also time-consuming, taking up to 18 months.

The edge router vendors we talked with had differing views on Osmine. Simon Williams, VP of product management at Redback, and Cisco’s Redford both say edge routers don't need Osmine certification – they're required only for transport systems, especially optical devices, Williams claims.

Others disagree, and most are inching toward at least partial Osmine certification. Aplion, Unisphere, and Laurel are among those that say they are moving their products toward Osmine certification; a Juniper spokeswoman says the company plans to work with Telcordia to obtain Osmine certification "in the near future"; and Celox is "evaluating" whether its edge router will be certified.

10) NEBS

Finally, while there are some differences of opinion over the importance of Osmine compliance, there’s certainly no such confusion over NEBS certification – which dictates basic minimums for safety performance and is required for any vendor supplying equipment to RBOCs or CLECs.

Most of the edge router vendors indicate their products are NEBS-certified. Aplion, Axerra, Celox, and Laurel say they are now working toward this.

A lot of the capabilities described in our “Critical Features” list are designed to save service providers money, by reducing either outages, the need for spare inventory, or the amount of rack space occupied in collocation facilities.

But there’s a whole other side to edge routers, and that’s services – more specifically, the ability to support new and revenue generating services.

When edge router vendors talk, it's often all about moving so-called "advanced IP services" to the edge of provider networks. They claim that with cheap bandwidth now an overpriced commodity and several of the long-haul carriers (read: Global Crossing Ltd. and Williams Communications Group) in deep trouble, service providers must deploy edge router platforms that facilitate the delivery of value-added IP services to generate revenue.

The most important edge router-enabled service is the VPN, or virtual private network. This is already a substantial market – and it is positioned to become much, much larger. The Yankee Group estimates that IP VPN service revenues will climb from $913 million last year to $1.6 billion this year, before more than doubling to $4.6 billion in 2004.

For anyone not familiar with VPNs, the idea behind them is to allow carriers to carve up and create “virtual” networks over a shared public IP backbone.

But there are many ways to do that:

  • IPSec-based VPNs, which have the advantage that they incorporate security via cryptographic and authentication features.

  • MPLS-based VPNs, which themselves can be broken down into two different categories:

    • Traditional Layer 3 MPLS VPNs based on the RFC 2547 spec. This type of VPN allows service providers to offer customers private tunnels through public IP networks. The problem is that it only works with IP traffic.

    • Layer 2 VPNs based on either the Martini or Kompella draft Internet Engineering Task Force (IETF) specifications. Both can carry all types of traffic by mapping ATM, Frame Relay, or Ethernet traffic into an IP/MPLS core.



  • GRE VPNs, or Generic Routing Encapsulation VPNs, which provide a pathway, or tunnel, across the shared WAN and encapsulate traffic with new packet headers to ensure delivery to specific destinations. A GRE VPN is private because traffic can enter a tunnel only at an end point. Tunnels do not provide true confidentiality (encryption does) but can carry encrypted traffic.

  • L2TP VPNs, which allow a packet-switched network to tunnel Point-to-Point Protocol (PPP) packets across an intervening network.

  • LAN Interconnect VPNs, which are basically Layer 2 VPNs, without the MPLS.

In addition to VPNs, edge router vendors cite a number of other potentially profitable applications for their products, including firewall security, as well as a related firewall capability: network address translation, or NAT.

Also, when combined, IP QOS and COS (class of service) can be used to allow carriers to charge enterprises a premium for carrying mission-critical data ahead of less important applications.

Whether this all really adds value, or just plain confusion, to the edge router marketplace is another matter, of course. One thing is certain, however: Cisco can only look back and wonder who the next challenger will be to try and take a bite out of its leadership position.

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

You May Also Like