Large data centers need a new 100 Gigabit Ethernet optical interface

Data centers need an affordable, low-power 100 Gigabit Ethernet optical interface that can reach beyond 100 meters.

Four consortiums have emerged this year to satisfy data center operator demands for an affordable, low-power 100 Gigabit Ethernet optical interface that can reach beyond 100 meters.

Optical interfaces are notorious for driving up the price of next-generation Ethernet speeds and 100 Gigabit Ethernet (GbE) is the most vulnerable today. An SR10 optical interface comes at a relatively reasonable cost for enterprises, but it is limited to a range of 100 meters. The LR4, on the other hand, can reach to 10 kilometers, but it is so expensive that only telecommunications service providers can afford one.

"The focus is to address the market requirements for these large data centers that are growing very fast," said Dr. Mario Paniccia, Intel fellow and general manager of Intel's Silicon Photonics Group, during a conference call with the media to announce one of these new consortiums, the CLR4 100G Industry Alliance.

"People are building data centers with millions of square feet, multiple football fields in size," said Andy Bechtolsheim, founder and chief development officer of data center switch vendor Arista Networks, during the same conference call. "The reach between switches [in these data centers] is greater than 100 meters. In many cases, it's 500 meters and sometimes up to a kilometer or two."

In short, today's 100 GbE optics are too expensive, consume too much power, produce too much heat, and don't have the range to satisfy the needs of large data centers.

"Heat and power are a huge issue in the 100 Gb market, for sure," said Nick Buraglio, network engineer for a global research network. "Many [colocation] facilities charge not only for rack space, but by power draw, and may not be prepared for handling the needs of a slew of new 100 Gb tenants."

Four proposals for a new 100 Gigabit Ethernet optical interface

In response, several industry alliances have formed to develop multi-source agreements (MSAs), or de facto standards, to build a new 100 GbE optical interface that is low-cost, low-power and can reach up to 2 kilometers.

The CLR4 100G alliance is designing an affordable, low-power optical interface for a QFSP transceiver. It will support a range of 2 kilometers over a duplex, multi-mode fiber with 4 lanes of 25 Gbps light paths. Today's standard optics support 10 lanes of 10 Gbps, which leads to thicker, more expensive cables.

Last month Mellanox Technologies launched its own group, the OpenOptics MSA, to target the same problems.

A third collective, CWDM4-MSA, backed by optical systems manufacturers Avago Technologies, Finisar Corp. and Oclaro Inc., along with cable maker JDSU, announced its own MSA on the same day Intel and Arista announced theirs. Both initiatives are similar, but the Avago group is focusing on single-mode rather than dual-mode fiber. Single-mode has less attenuation, or loss, over longer distances.

Avago and its partners also backed yet another group, the PSM4 MSA, in January, which aims to produce an affordable 100 GbE optical module that can reach at least 500 meters.

IEEE failed to act on 100 Gigabit Ethernet optical interface problem

The field is crowded in part because last summer the IEEE's task force on high speed Ethernet was unable to reach a consensus on a standard for medium-range 100 GbE optical interface.

"When the IEEE vote failed last July, there was no single proposed low-cost 100 Gb optic," Bechtolsheim said. "It was predictable that the industry participants, who wanted to see these optics come to fruition, would form MSAs. The key thing is we can't wait another nine months for [an optical specification]."

One area most people seem to agree on is the need to build affordable optics for QFSP interfaces, rather than CFP interfaces that are reliant on fiber with 10 lanes of 10 Gbps.

More on next-generation Ethernet

Will we ever need 400 Gigabit Ethernet?

Best practices for migrating to 40 Gigabit Ethernet

"We've been stuck sitting with CFP-based interfaces that are simply, by their physical nature, really expensive," said Eric Hanselman, chief analyst at 451 Research. "If we can get to 25 Gb lanes in some agreed-upon fashion and with a lower-cost connector infrastructure, that would really start to move [the industry] to 100 GbE."

Hanselman said the technology underlying these proposals is solid. The challenge is reaching a consensus so vendors can start building and selling optics in volumes high enough to push prices down.

"You need a lot of people on the bandwagon when this thing is ready to go," he said.

Let us know what you think about the story; email: Shamus McGillicuddy, news director, or follow him on Twitter @ShamusTT.

Dig deeper on Data Center Network Infrastructure

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

-ADS BY GOOGLE

SearchSDN

SearchEnterpriseWAN

SearchUnifiedCommunications

SearchMobileComputing

SearchDataCenter

SearchITChannel

Close