ColorChip Joins Intel, Arista in 100G CLR4 Alliance for mid-reach data center links

Apr. 1, 2014  | LightWave Online

http://www.lightwaveonline.com/articles/2014/04/intel-arista-launch-100g-clr4-alliance-for-mid-reach-data-center-links.html

Intel, Arista launch 100G CLR4 Alliance for mid-reach data center links

April 1, 2014

The number of new multisource agreements (MSAs) vying to address 100 Gigabit Ethernet at reaches between 500 m and 2 km grew to four today with the announcement of the 100G CLR4 Alliance. Backed by Intel, Arista Networks, and several chip vendors, systems houses, and optical component and transceiver vendors, the 100G CLR4 Alliance will target 4×25-Gbps via duplex singlemode fiber in a QSFP optical module form factor.

In a teleconference for the press held yesterday, Intel Fellow Mario Paniccia and Arista Networks Founder, Chairman, and Chief Development Officer Andy Bechtolsheim said that the increase in traffic within data centers has outstripped the ability of 10 and 40 Gigabit Ethernet interfaces to keep up, both in terms of data rate and line card density. Meanwhile, the IEEE’s 100 Gigabit Ethernet 100GBase-LR4 PMD is too expensive and “telco oriented” at its target reach of 10 km, while the 100-m 100GBase-SR4 specification doesn’t meet the necessary link lengths. A new specification is therefore required, Paniccia and Bechtolsheim told those on the call.

The CLR4 specifications – which Paniccia and Bechtolsheim predicted would be locked down by this May, after preliminary specifications are released publicly this month – would create an interface that would solve these issues, the alliance members expect. Use of the QSFP form factor would result in significant faceplate density (36 ports in 1 RU) while pulling only 3.5 W per transceiver.

A wide variety of technology vendors and users have joined the alliance, including eBay. Other members besides the three already mentioned include 3ality Technica, Altera, Aurrion, Brocade, Dell, Ciena, ColorChip, Hewlett-Packard, Fabrinet, Fujitsu, Kaiam, MACOM, NeoPhotonics, Netronome, Oclaro, Oplink, SAE Magnetics, Semtech, and Skorpios. While Intel, Aurrion, and Skorpios have announced silicon photonics technology for data center applications, Paniccia said in response to a question that the CLR4 specifications will not specifically target that technology.

However, it will target what has become a very crowded MSA space. The 100G CLR4 group is the third in the last month to announce a 4x25G approach using singlemode fiber and WDM for mid-reach data center links, joining the OpenOptics MSAthat debuted at OFC 2014 and the CWDM4 MSA that was unveiled mere hours before Paniccia and Bechtolsheim addressed the media. (The PSM4 MSA announced this past January will use a four-fiber, parallel approach to 100 Gigabit Ethernet requirements.) Paniccia said that the 100G CLR4 Alliance aimed to create consensus around a single specification, and that organizers had “socialized our spec to everybody” with the idea of creating a specification from the end user perspective rather than the optics perspective.

Bechtolsheim, meanwhile, said 100G CLR4 specifications will not copy the OpenOptics MSA’s use of 1550-nm optics (which led him to call OpenOptics “the dense WDM MSA”). However, he had favorable words for the CWDM4 MSA’s direction.

“The CWDM one to me makes the most sense, given that there’s broad industry support for that and it’s also the one we believe will get to market most quickly,” Bechtolsheim said. “But if I could add one thing here, which is between last July [when the IEEE next generation high-speed Ethernet task force declined to create a new PMD to address the mid-reach requirement] and now, nine months have passed. And nine months ago it wasn’t important enough to come up with a spec, people disbanded – and suddenly you have multiple people and proposals. The key thing here is that we cannot wait another nine months for the spec to come to fruition.”

Bechtolsheim pointed out that Arista has announced a line card designed to accommodate 100 Gigabit Ethernet QSFP interfaces “and we need those optics now.”