Building the 21st century central office

For more than a century, the central office is where the telecommunications industry has hung its hat. But the job for which these unassuming but ubiquitous brick houses were built, and the technology within them, are changing dramatically. In fact, it's tough to say what the CO will look like 10 years from now. But by analyzing some current business and technology trends and extrapolating them, some industry experts offer a theoretical glimpse of the 21st-century central office.

In general terms, it's not hard to predict where COs are headed. The slow migration from circuit-switched to packet-based networks has COs looking more like data centers every year, and this evolution will continue. Over time, said Pieter Poll, chief technology officer for Qwest Communications, “the central office becomes a data center. … It's inevitable. If you don't do that, you become choked in your ability to move extreme amounts of bandwidth to customers.”

But COs — widely distributed and sometimes dusty — weren't designed to be data centers and thus exhibit key characteristics that will make such a transformation a tough trick to pull off.

“The typical CO that we've created the Bell system out of was never intended to be the data center world that we're finding ourselves moving toward today,” Poll said. “We don't have the raised-floor, traditional, UPS power, cooled environment you'd find in a Qwest hosting center. We have [Network Equipment Building Systems] Level 3 requirements in place around things like power and density. I think it's fair to say those requirements are simply at odds with modern technology.

“We need to create environments in there that allow us to safely and economically operate equipment that has far higher power densities, heat dissipation requirements and power feeds than we ever built those locations for in the past,” he added.

SPACE RACE

One of the ever-present pressures on the central office is the scarcity of space and power — or more accurately, the imperative to use space and power as efficiently as possible. In the future, keeping up with escalating broadband traffic volumes will require carriers to continually inject ever more capacity into the central office, which implies increasing space and power needs over time. And the launch of new services, requiring more new equipment, would seem to crowd the CO ever more.

But looking forward, space is also being created in the CO by a number of factors, such as the increasingly dense, more-integrated technology produced by vendors. Take last year's packet optical networking platforms, for example.

“Rows and rows of digital cross-connects, DWDM gear and Sonet gear have all been replaced with one box,” said Stuart Elby, vice president of network architecture for Verizon. “It's all integrated at the system-on-a-chip level, the interface level. The density is just tremendous. A few racks of that replaces bays and bays of traditional OC-48 Sonet gear and digital cross-connects.”

Moreover, much of the space and power consumed in COs today is devoted to translating legacy traffic into packets and vice versa. The further carriers get in their circuit-to-packet migrations, the more that middleman equipment goes away, taking a significant power drain with it. And as some functions in today's CO move to more-centralized locations, some believe the CO could shrink over time, while others think the opposing forces affecting CO size will offset each other.

“In general, the service resources are going to become more centralized over time,” said Tom Nolle, president of CIMI, a consultancy. “That will free up enough space that I don't see physical size of the CO or its overall power requirements changing very much. It's just that the nature of the equipment there is going to change a lot.”

With regard to some equipment, though, increases in density don't always free up as much space and power as expected because they create another issue: greater cooling requirements. For example, when Verizon deploys some very dense routers today, it needs to surround each router-filled rack with two or three empty racks just to make cooling manageable.

“You have to scratch your head and say, ‘Why did they design such a dense box if I can't put two of them next to each other anyway?’” Elby said. “We drove them to design those sorts of devices. But maybe we've gone beyond what's really needed in a rack. ”

In a few years, Elby imagines a new generation of routers taking a step back from the leading edge of density in favor of an approach that takes heat and power into greater account, achieving a new level of power efficiency inspired by advances made in consumer electronics.

“The power management in a laptop is probably more sophisticated than in a Class 5 switch,” Elby said.

Verizon, which estimates that its COs consume nearly 3.5 billion kilowatt hours each year (not far short of the output of the Hoover Dam), vows to lean harder on suppliers in this regard. This month the carrier announced a set of internally developed standards for a wide range of networking gear to be purchased next year, with a goal of cutting CO power consumption by 20%. The move was inspired by EnergyStar, the consumer electronics standards created by the federal government in 1992, which the Environmental Protection Agency estimates saved $14 billion in energy costs in 2006 alone.

“[We're] looking at it holistically and saying, ‘How do we rethink how you design boards, cards and switches to be energy-efficient?’” Elby said. “Unless some disruptive technology comes along, it's just going to be a lot of hard work, getting whatever the CO equivalent of EnergyStar is.”

SERVERS OF THE FUTURE

Over the course of the next decade, the drive to increase power efficiency in the CO will focus on one device more than any other.

“If I look forward five or 10 years, the big power consumer, the big cost in the CO is going to be on that server front,” Elby said. “That's becoming the main component going into the CO today.”

As recently as the late 1990s, a typical customer-facing CO was filled with rows and rows of ATM and frame relay equipment, Class 5 switches, and Sonet and DWDM gear. Routers and servers were rarities. While much of the same gear remains today, that ATM and frame relay gear is most likely sitting on a forklift on its way out the door. You'd be hard-pressed to find a CO without a router, and rack after rack is filled with servers.

As telephony increasingly becomes a server-based offering, Verizon servers handle a variety of functions, including service portals for enhanced data communication services, video content storage and proxy servers for IP multimedia subsystem (IMS) services. Today they are proprietary hardware running softswitches, but in a few years they will be more generic servers hosting software-based IMS services.

But as many servers as COs accumulate, telcos are unlikely to build the riverside mega data centers for which Google and its kind are famous. Because telcos are already heavily invested in CO real estate, it wouldn't make sense to copy Google's centralized model, most sources said.

To the extent that COs resemble Google's data houses, they'll be miniaturized versions, both because COs are more distributed and because they'll serve different functions. For starters, telecom applications are much less computing-intensive than search engines and therefore require much less power per rack than Google does (power economy being the single greatest criterion driving Google's design). Even if voice is eventually relegated to discrete IMS or SIP servers and not integrated into other gear, the attendant power requirements will be a fraction of that required by Class 5 switches today, Nolle said.

“IP transformation, which is one of the causes of CO condensation, allows for a more regional-based architecture relative to where services and applications are hosted,” said Tom Anderson, director of Alcatel-Lucent's North American Chief Technology Office. “The CO becomes a much smaller, compact hub, but the place where real service delivery is managed relative to not only voice but other applications is really someplace else.”

Telcos will benefit from the previous efforts of Google and large enterprises to pound more efficiencies out of server designs, Elby said. As carrier capacity requirements drive their equipment from 40 Gb/s to 100 Gb/s interfaces, they will match those common in the data center sector, and the technologies of the two worlds will interbreed more, driving economies of scale and blurring the line between CO and data center. But COs, with their characteristic distribution, will have unique requirements for server technology, Elby added, so telcos will have to do plenty of their own pounding on suppliers.

Data centers will likely always be the first recipients of new improvements in 100 Gb/s technology, Elby said. But that technology then will have to evolve to suit the telco model. Connecting servers that are sitting atop one another is a different task than connecting devices across a metro area, which means that telcos will have to deal with latency issues not experienced by Google and its peers. Telcos also will have to deal with personnel issues not seen in data centers.

“I don't have a sophisticated IT force in every CO, and I can't envision doing that,” Elby said. “That's just too much retraining to say, ‘OK, you're not going to work on a cross-connect anymore, you'll manage a Sun server.’ We need very sophisticated remote management capabilities and tools onto these server farms. That hasn't been a focus [of the server vendor community] because usually those large farms are well-manned by IT forces in other industries.”

CENTRALIZING THE CO

Personnel is another aspect in which the future CO could be subject to change. Operations is such a dominant cost in carrier economics that economizing operations will be a major priority for carriers, and spreading a skilled work force out among central offices is inherently inefficient, Nolle said. Meanwhile, carriers are increasingly automating the central office with reconfigurable optical add/drop multiplexers (ROADMs), automated fiber patch panels and control planes that provision optical networks.

As early as next year, Verizon Business hopes to achieve automated cross-country provisioning of optical networks so that network operators needing a path between, say, New York and Los Angeles simply input that requirement and let the network do the rest, finding the best path end-to-end. That level of automation would free up a lot of hands that normally would be needed between New York and L.A. When asked how many, one Verizon executive said only “many,” declining to guess at the numbers because such labor issues can be sensitive subjects.

But according to Nolle, carriers are not headed toward drastic reductions in CO staffing, even with all the technological change on the way. “Most operators will tell you: The largest headcounts are associated with touch at the end of the loop,” he said. “As long as you have loops, you'll have touch at the end of the loop.”

While it's true that replacing copper loops with fiber and passive equipment reduces maintenance needs, only a minority of American homes will enjoy fiber in the foreseeable future, Nolle said.

In the future, COs will be exploited most for what distinguished them the most from data centers: their distribution. “What today are COs are extremely valuable real estate because they're at the crossroads of how we get to customers,” Qwest's Poll said.

To make the most of that distribution, telcos must strip out all the functions that are better centralized. For example, Elby said all the provisioning work in a CO will be pulled back into a limited number of network operations centers, designed with disaster recovery as a priority. “You could envision three or four around the country could cover it,” he said.

Most service technologies could be centralized to a small number of metro locales, Nolle said. In the future CO, aside from voice, “I don't believe any other type of service is likely to be significantly hosted in the CO,” he said. “There's not a huge advantage to forward-placing other service assets in today's world. The only value would be if we assume metro bandwidth costs were significant. But for the operators in particular, that's not likely to be true.”

RISE OF THE NOC

As vendors put more built-in operations, administration and maintenance functionality in their service-managing gear, more provisioning and management reverts to the network operations center (NOC), which becomes increasingly more influential, Nolle said. “There's going to be relatively little service technology that stands between the NOC and the dispatch process,” he said. “If I can't fix something in the NOC, I'll roll a truck.”

Some services might be left to COs, especially large COs serving dense areas. “For example, it would make sense to have whatever voice calling facilities supporting Manhattan at the CO level because there would be so much intra-CO calling,” Nolle said.

And some content is best centralized — Verizon's nationwide HBO feed, for example, comes from two locations. But other content types are best-seated as close to users as possible, Elby said, including long-tail content, locally inserted content and user-generated content. Where there are specific communities of interest for a given piece of content, it makes no sense to tie up the network by dragging that content across the country.

“We're just on the cusp of user-generated content becoming a major driver,” Elby said. “It goes beyond YouTube and MySpace. More and more content is getting uploaded. Most people today that take digital photos use a free service like Google or Snapfish. Those are free today, but at some point someone will figure out how to monetize those. And that will be more distributed as well.”

Network-based personal video recording content (for managed TiVolike services) is also likely to be stored close to customers in COs.

Another way to make the most of the distributed CO network is with peer-to-peer (P2P) technology, which Verizon and others are increasingly investigating through efforts such as the recently formed P4P Working Group.

Although P2P is often thought of for content distribution, it could be used by telcos for signaling distribution. P2P signaling technology could be used to locate content and establish a session between content and those who want it.

“If we can do that, we can distribute the signaling functions, which significantly facilitates the provisioning of reliable services without having these huge [server] farms sitting in COs,” Nolle said, predicting another function for the 21st century CO: “At the service level, the CO becomes a P2P supernode.”