Network Evolution The Ties That Bind
New platforms and protocols have emerged to aid carriers and operators in their drive toward a converged, unified network, capable of weaving
together a tangle of legacy and emerging services
In a perfect world, every service and application a carrier or network operator offered would run over the same network, creating new levels of efficiencies and an overall platform poised to snare revenue from a variety of existing and emerging, yet nascent, services.
Well, as everybody knows, the world is less than Utopian. However, some things appear to be mutating and evolving for the good, especially in the metropolitan area network (MAN), which today is dominated primarily by Sonet (synchronous optical network) and Ethernet platforms. New protocols and network infrastructures, including proprietary and standardized flavors, are now aiming to create a unified network for typical voice and data services, as well as newer, more bandwidth-intensive applications still a bit further out on the horizon.
For now, though, network operators and carriers are relying on distinctly separate networks, or "silos," for each type of service. One network might be built for video traffic, another for data, and maybe even a third outfitted for low-latency, carrier-class voice services. Although three separate networks can operate and feed services to hungry customers, managing and provisioning them all provides the foundation for a cumbersome set of operational and technological challenges.
Typically, Sonet, with its inbred QoS and resiliency, has been used for voice. Then, as high-speed data services started to emerge, Ethernet technologies started moving to the forefront, as well.
At the same time, a new class of advanced infrastructures, protocols and transport methods have surfaced and continue to gain momentum. Topping that list: next-generation Sonet, resilient packet ring (RPR), optical Ethernet and dense wave division multiplexing (DWDM).
Cable operators "are looking for ways to improve the efficiency of the network as they add more and more traffic on it," says Paul Connolly, vice president of marketing and network architectures for S-A's transmission sector. "It becomes a nightmare if you have to provision equipment for every type of traffic being carried."
RPR is gaining some attention at the MSO level because the technology appears to be well-suited to combine the inherent needs of carrier-class voice and high-speed data services at favorable costs. In addition to its inherent advantages, RPR, a Layer 2 MAC (media access control) technology, is designed to co-exist with Sonet and optical Ethernet networks.
The market for next-gen Sonet equipment could reach $12 billion in 2006, up from today's estimated $2.7 billion, forecasts the Yankee Group.
When next-gen Sonet is pooled with DWDM and optical Ethernet, the market forecast balloons to $23.6 billion in 2005, up from $13 billion last year, according to Cahners In-Stat, a sister company to CED magazine.RPR: The best of both worlds?
Though RPR is designed to take the best and brightest from Sonet and Ethernet, and is billed as the answer to the future, it actually complements, not replaces, existing technology.
"Sonet, next-generation Sonet and packet ring are mutually compatible," adds Steve Garrison, director of corporate marketing at Riverstone Networks. "You can still put RPR on top of that."
Though there's no single, dominant right answer for video services, the attraction of RPR is that it can send packet-based traffic, but can also handle circuit-switched requirements, notes Connolly.
In addition to its carrier-class capabilities, packet ring technology "gives twice the bandwidth on the same pipe and the same resiliency as you get with Sonet," says Garrison.
The presence of RPR offers a lower cost method for aggregating real QoS data signals out of the cable modem termination system, and bringing it back to the Internet, says Mannix O'Connor, director of business development for Lantern Networks, one of the key contributors to the evolving RPR standard.RPR standards still a year away
Though RPR looks good on paper, pushing that paper through the requisite approvals is a time-consuming, politically-charged process.
Because a final RPR standard isn't expected until the first quarter of 2003, the question today is whether carriers and operators will decide, in the meantime, to move ahead with their next-gen network plans.
"It's not exactly an option to do it later," offers Yankee Group Senior Analyst Marian Stasney. "Standards always follow implementation."
"It would be great if I could wait until all of these protocols and everything goes through, and everything's set, but in the meantime, I've invested billions in my network," says Paul Gemme, senior vice president of networking engineering for Time Warner Cable. "I've got to make some of that money back. I can't wait."
But moving ahead with non-standard equipment comes with its own set of risks.
"I always worry about getting tied into pre-standard products, because everybody puts their little spin on it," Gemme says. "So what you can ultimately end up with is a proprietary system. All we can hope is that we make the correct judgment calls so that we're not going to have to abandon large amounts of invested capital."
One MSO that has already moved down the pre-standards pipeline is Cox Communications, which is installing Scientific-Atlanta Inc.'s Prisma IP platform in Lafayette, La. The goal there is to build one network for the company's slate of voice, video and data services.
S-A repackages and markets the equipment via a partnership with Luminous Networks, another key cog in the RPR standards machine.
The RPR Alliance reached a significant milestone in late January when it announced consensus on a baseline draft for the standard. Significant not only because it was one step closer to reaching a standard, but that factions in the group mustered some common ground on key RPR protection and bandwidth management elements (see sidebar, page 44).
The baseline "Darwin" RPR document includes everything adopted at a meeting last November, including the MAC and protection schemes, save for one that covers topology discovery.
The next move involves posting the first RPR draft to the IEEE 802.17 Web site for member evaluation and analysis. From there, the group will create a comment database, which will be discussed at a meeting this month. Two similar rounds of discussion are expected to take place, with the final one set for July.
"The expectation is that by the time we've gone through three rounds of this, we will be down to the fine points," says RPR Alliance Chairman Bob Love.
Before RPR is finalized, however, packet ring vendors are expected to spin hardware based on the unfrozen standard.
Though many changes can be done in software, some changes in the hardware, including silicon, might be required before equipment fully adheres to RPR's final standard, concedes Gady Rosenfeld, director of strategic marketing for Silicon Valley start-up Corrigent Systems.
Despite those potential challenges, Lantern Networks is confident that it will be able to generate plenty of business for its pre-standards equipment. "We will have our first revenues this quarter. When we talk about trials and shipments to major customers, it's not just hype, but revenue shipments," O'Connor says.
Raj Sharma, director of technology strategy for Luminous Networks, agrees. RPR "is designed for the future, but with the present in mind."
Others see standards-based RPR as a longer-term play. "As with any disruptive technology, it's easily a five- to 10-year cycle," Riverstone's Garrison estimates.Mixing it up
Though RPR, next-gen Sonet and optical Ethernet are getting plenty of attention, technologies such as DWDM and MPLS (multi-protocol label switching) are starting to get their due, as well.
DWDM, which runs independent of Sonet or RPR, is well known for its bandwidth capabilities. "DWDM is like Ethernet on steroids," says Agnes Imregh, vice president of marketing for LuxN, a Sunnyvale, Calif.-based optical company.
DWDM, says Jeff Baher, director of marketing for Cisco's metro IP access unit, has two roles: to handle storage and data and TDM (time division multiplexed services), and to multiply fiber capacity. Which role it takes depends on where a network operator resides on the metro food chain, he says.
Because cable operators are looking at network convergence, "right now, they're finding ways to be cost-effective using DWDM," says Patrick Harshman, president of Harmonic Inc.'s broadband access network division. "DWDM does pretty well serving that requirement for a common platform, which later can go on top of something like RPR."
Also keeping a close eye on DWDM developments is Toronto-based start-up Photonami, which has developed a proprietary platform, dubbed iDWDM.
"Sonet is really designed for voice. It has been used for data, but it's rather inefficient and needs to be over-provisioned to handle bursts," says Photonami Vice President of Business Development Richard Pearce. "DWDM is about getting scale into the metro."
Photonami's plan is to tap DWDM to offer scalable bandwidth via "optical burst switching," a proprietary technique that allocates a wavelength to data traffic on an as-needed basis.
Because Photonami's goal is to offer a system with carrier-class capabilities via the use of "hot standbys," the iDWDM platform can co-exist with, but could potentially replace RPR, Pearce says.
Internet Photonics, another DWDM start-up, is touting a platform that combines carrier-grade optical gigabit Ethernet with 10G metro DWDM transport.
On the cable front, "VOD will be the big thing for us," predicts company Vice President of Solutions Marketing Gary Southwell, noting that Internet Photonics' gear, as an alternative to RPR, forges return paths for VOD and iTV channels.
Moving further down the chain, MPLS "provides Layer 2 switching capability for packet traffic, but all the configuration and control is done like Layer 3 services today," says Erik Metz, senior staff engineer for Motorola Broadband's Transmission Network Systems unit. "It's irrelevant to the architecture."
MPLS, handled at the network's core, creates a "pseudo wire," similar to ATM (asynchronous transfer mode), but it handles it in an Ethernet environment. "All of those things add up to efficiency and lower cost points," Riverstone's Garrison says.Making the decision
For MSOs, adding RPR, next-gen Sonet and a mix of DWDM and MPLS will largely depend on what services they plan to offer.
Cost will also factor in. Though RPR-like equipment remains proprietary for now, standardized equipment will drop in cost because the equipment will use common components and silicon and become available from a phalanx of competitive vendors.
However, gauging those potential cost savings will prove difficult until equipment is actually standardized. "We can only check RPR on a technical level for now, so there is not enough cost information on RPR yet," Metz says.
In the meantime, Gemme says he's leaning toward the Gigabit Ethernet platform because of its bandwidth potential and ability to handle lots of video streams. "RPR and MPLS are more protocol issues," he says. "We think [GigE] will allow us to analyze other protocols as they come about."
Gemme says MPLS will be a benefit when there's a heavy load on the network. "I don't think it's essential today," he adds. "You can work around it today. I would use it more for voice-over-IP, data and business traffic than I would for video-on-demand."
Though every decision will not always be cut-and-dried, network flexibility and capex management will always play key roles.
"Operators need that [flexibility] in order to handle services offered today, but need to be prepared to use the network for services coming along two to five years from now," says Dave Lively, senior manager of optical strategy at Cisco. "They can't afford to forklift the infrastructure and replace it with something completely new."