FTTH Conference considers what's next
OCTOBER 5, 2007 By Meghan Fuller -- Dubbed "The Content Revolution: Filling the Pipe," this year's Fiber to the Home Conference, held this week in Orlando, FL, examined futuristic technologies, services, and applications. From the opening keynote address by Bret Swanson, senior fellow, Technology and Democracy Project, at Seattle's Discovery Institute, the conference focused on the need for bandwidth, bandwidth, and yet more bandwidth. And the system vendors talked about the various architectural choices that might support that bandwidth.
Swanson coined the term "Exaflood" to characterize the coming flood of Exabytes over the Internet. (For those unfamiliar with the exact size of an Exabyte, Swanson provided a definition. A Kilobyte is 103 bytes, a Megabyte is 106, a Gigabyte is 109, a Terabyte is 1012, a Petabyte is 1015, and an Exabyte is a whopping 1018 bytes. Put another way, 1 Exabyte = 1 trillion Megabytes.)
One trillion Megabytes may seem hard to fathom, but when he spoke of the Exabyte, Swanson was not referring to a bandwidth benchmark that is decades away. On the contrary, the growing popularity of such applications as peer-to-peer file sharing, online gaming, and video conferencing has placed the Exabyte on our proverbial doorstep.
Take YouTube, a popular topic of discussion at this year's conference. Today, YouTube accounts for about 7% of all U.S. Internet traffic, consuming 50 Petabytes per month or 600 Petabytes a year--which is greater that the entire Internet in the year 2000. What's more incredible is the fact that this content is, for the most part, entirely user-generated. According to Swanson, all professionally produced content, including broadcast and cable TV and radio content, comprises about 100 Petabytes per year. YouTube, by contrast, currently streams that amount of data in two months.
On the gaming front, a single, massively parallel game with 1 million players could generate 100 Petabytes of traffic per month--more than an Exabyte per year or one-tenth the size of today's Internet, said Swanson.
Video conferencing also has the potential to boost Internet traffic by orders of magnitude. Today, Swanson reported, the U.S. generates about 20 Exabytes of voice traffic per year. But the move to videophones would generate about 200 Exabytes at least or about ten times the size of the existing world Internet. "And that's a conservative estimate," he noted.
For his part, Swanson believes that FTTH may be the key industry for the health of the overall U.S. economy in the next five to ten years. The ripple effects of the FTTH industry will be wide and far reaching, he said, as all other industries will turn to fiber to meet their burgeoning bandwidth requirements.
Beyond GPON
In the near term, however, FTTH vendors need to figure out how to support that level of bandwidth. In his presentation, "What's next for PON: Benefits for service providers," David Spence, staff product manager at Tellabs (www.tellabs.com), noted that service providers, by and large, are looking to leverage what they've already deployed (e.g., GPON), but they also want more bandwidth, longer reach, improved economics, and more intelligence. He described two methods whereby service providers could increase bandwidth per subscriber beyond the 75 Mbits/sec that results when GPON's 2.4-Gbit/sec downstream data is shared among 32 subscribers: you can either scale in the time domain or in the frequency domain, he said.
Scaling in the time domain would require boosting GPON's current 2.4-Gbit/sec downstream rate to a higher data rate, with 10 Gbits/sec the most likely option. (It is currently under study within FSAN). In this case, the cost of the high-speed transceiver is spread over the 32 ONTs. That said, each ONT only uses a fraction--1/32nd--of the available bandwidth, admitted Spence.
Scaling in the frequency domain, by contrast, results in a better match between transceiver cost and bandwidth utilization of the ONT. This type of PON, most commonly known as Lambda-PON or WDM-PON, provides a single, dedicated wavelength to each ONT, thereby dramatically increasing the available bandwidth to each home. However, such an architecture requires either a) the deployment of 32 different ONTs, one for each wavelength; or b) the deployment of a tunable laser in each ONT to select the desired wavelength. While component costs make WDM-PON less attractive in the near term, Spence nevertheless confirmed that Tellabs believes "scaling in the frequency domain is the way to go."
The folks at Alcatel-Lucent (www.alcatel-lucent.com) also tackled this subject in a presentation dubbed "Life beyond BPON, GPON, and P2P," delivered by Ronald Herron, director of network technology and strategy. Herron noted that GPON will be sufficient through 2011--if service providers changed split ratios from the currently favored 1 x 32 to 1 x 8.
Beyond 2011, Herron believes network architectures will fall into three camps: A next-generation GPON, still based on TDM; a pure WDM-PON; or a hybrid TDM/WDM PON. From his perspective, the next-generation 10G PON is "a viable future technology," though it will require continued research to make it a reality. And he believes that WDM-PON is "not economical in the foreseeable future" thanks to the high cost of optics per user.
Instead, Herron believes the hybrid approach makes the most sense. This architecture uses a stacked WDM approach to quadruple bandwidth capacity to existing ONTs. Instead of a single wavelength operating in the 20-nm window, the hybrid approach enables four wavelengths to operate in the existing window, which is partitioned into four smaller 5-nm windows. The least expensive of the three approaches, the hybrid architecture is also the least disruptive, said Herron, because there is no need to change the existing ONTs.
What Alcatel-Lucent proposes is to replace the existing 1 x 32 splitter with a 1 x32 hybrid splitter comprising a 1 x 4 wavelength and four ordinary 1 x 8 splitters--all in a single module the same size and shape as an existing 1 x 32 splitter. This architecture would require the network operator to replace the existing splitter with the hybrid splitter, but again all other outside plant equipment--including the ONT--would remain the same. During his discussion, Herron described a bi-mode splitter that initially could be used as a traditional splitter and then as a hybrid 1 x 32 splitter later on, thereby eliminating even the need to change the splitter in the future.
While it remains to be seen which future architecture will become the dominant technology, this week's FTTH Conference underscored one undeniable fact: bandwidth requirements will continue to increase. In fact, Swanson predicted that by 2015, the Exabyte will be old news. Instead, we'll be talking about the Zettabyte--1021.