Optical Power Trip
lightreading.com
When Tellium Inc. announced the shipment of its 512-port Aurora optical switch last month (see Tellium Ups the Ante ), observers were impressed by its performance numbers. Big in every way, the unit is also becoming an emblem of one of the networking industry's dirty little secrets -- the growing power consumption and heat dissipation figures for optical networking equipment, which may threaten or stall purchases and deployment schedules for both end users and service providers.
With its four separate 7-foot bays that can consume a maximum of 6,000 watts each, Tellium's Aurora switch has become the poster child for the physics problem that comes with increased bandwidth. Simply put, when you increase the number of networking devices (or the number of circuits in each device), you also increase power demands and heat output.
While there are no reports yet of central offices spontaneously combusting, the problem has already started to generate its own heat. "On a scale of one to 10, it's already a 10," says Jeff Monroe, vice president of design and construction for Metromedia Fiber Network Inc. (Nasdaq: MFNX), of the growing power and heat concerns.
Unfortunately, customers and service providers don't always get adequate data from the equipment vendors to help them figure out how to plan for power and air-conditioning needs. Tellium, whose equipment comes up in just about every conversation about networking power requirements, at least publishes figures on its Website. Other vendors like Cisco Systems Inc. (Nasdaq: CSCO), Avici Systems Inc. (Nasdaq: AVCI; Frankfurt: BVC7), and Corvis Corp. (Nasdaq: CORV) don't publish complete stats for their bigger switches.
"Power is not something that [vendors] like to talk about," said Peter Hunt, product manager for the core networking division at Sycamore Networks Inc. (Nasdaq: SCMR). "But it certainly is on customers' shopping lists to get power consumption as low as possible."
That's mainly because those customers -- and their partners, including providers of colocation facilities, or RBOCs who must colocate equipment -- may be reaching some limits to the amount of gear they can handle in a given location. Don Bennett, telecom test manager for Garwood Laboratories, an independent testing service, said that the older central switching offices are particularly stressed, given that most were built before the advent of the Internet.
"The COs [central offices] are not designed for high-power, high-heat devices," Bennett said. "They're not equipped, HVAC-wise, to handle the new loads. But the prevailing theme is, jam as much [equipment] as you can into the space."
In the past, the central offices were governed by AT&T Corp. (NYSE: T), which instituted a rigorous battery of physical tests for networking equipment. Known as the Network Equipment Building Standards (NEBS), the parameters were developed by AT&T's Bellcore arm to ensure that the phone networks could withstand forces of nature like earthquakes, fire, and electrical interference. The deregulation of the service-provider industry, however, has made the standards somewhat toothless, since independent providers don't have a monopoly parent to which they need to answer.
"The [NEBS] numbers have gotten a bit contorted in the industry," said MFN's Monroe. According to Garwood Labs' Bennett, most networking equipment exceeds some NEBS standards for power consumption and heat output. What vendors are truly concerned with, apparently, is getting more performance out of less space.
"The equipment is packing more bits, but what hasn't gotten smaller is power," MFN's Monroe said. "It follows a more linear curve." Some equipment providers, Monroe said, have even talked about the idea of building water-cooled systems, like the old mainframes and supercomputers.
"At the end of the day, physics is a real thing, and you have to live by its laws," Monroe said.
-- Paul Kapustka, Silicon Valley bureau chief, Light Reading lightreading.com |
