Google spent much of 2010 beating the drums for an alternative to the IEEE’s 100-Gigabit Ethernet (100GbE) 100GBase-LR4 specifications. We want something that would produce modules that would prove cheaper and more likely to be developed quickly enough to meet our requirements, Google representatives said during speaking opportunities at the major industry tradeshows.
The company raised eyebrows in December 2010 when Google decision makers apparently decided that if they wanted something done right, they would have to do it themselves. Google sponsored its own transceiver multi-source agreement (MSA), in collaboration with Brocade (NASDAQ:BRCD), JDSU (NASDAQ:JDSU), and Santur Corp. (see "Google joins tech vendors in 10x10G 100 Gbps optical transceiver multi-source agreement").
Dubbed the 10x10 MSA – to highlight the fact that it targeted a module that combined 10 wavelengths of 10 Gbps, rather than the 4x25-Gbps approach outlined in 100GBase-LR4 – the group worked quickly to release its initial set of specifications in the first week of March 2011, then used OFC/NFOEC the following week to make as large a splash as possible.
Yet, beneath the whoopla, doubts were raised about the MSA’s overall influence on the market. Key to these doubts was uncertainty regarding both ends of the user/supplier equation: How much of an opportunity did Google and its ilk represent, and how many suppliers would chase it?
10x10 MSA: Fast and efficient
The speed with which Google and its fellow MSA members moved from launching the effort to releasing specifications both established Google as a major market influencer and reinforced the “keep it simple and straightforward” message it had been preaching. The problem with 100GBase-LR4 devices, Google representatives had asserted during their trade show tour, was that they would be expensive and power hungry because of the use of 25-Gbps components; they also would take longer to develop and optimize than Google’s requirements would allow.
What the industry really needed, Google said, was an interface that would supply 10X the data rate of current 10-Gbps modules but at only 4X the cost and power dissipation. Thus, the 10x10 MSA developed specifications that called for a CFP form factor module that would transmit and receive 10 wavelengths of 10 Gbps over 2 km of singlemode fiber (http://www.10x10msa.org/documents/MSA%20Technical%20-Rev1_2%20_Mar3.pdf). The approach would leverage current 10-Gbps optical components and electronics – which obviously would result initially in 10X the cost and power dissipation of current 10-Gbps technology, but offered a better chance than 4x25-Gbps approaches of meeting the 4X cost/dissipation target in the near term, the MSA members claimed in a white paper. In fact, the MSA members have stated their approach will achieve such savings as soon as 2012. Meanwhile, they say 100GBase-LR4 technology will start at 100X the cost and 17X the power dissipation of SFP+ 10-Gbps transceivers, giving that approach a much longer road to travel.
Meanwhile, the MSA members have their eyes on the next generation of module as well. The 10x10 HD specifications would lead to an alternative to future 100GBase-LR4 CFP2 modules that the MSA says could be even cheaper than 10GBase-LR SFP+ modules by 2013.
10x10 MSA pros and cons
The 10x10 MSA module would appear to have two things going for it: a built-in ecosystems of component and module vendors, systems suppliers who might incorporate the modules into their hardware, and three users who might be interested in deploying it in their networks; and a set of specifications that likely would lead to a cheaper module than any based on 100GBase-LR4. However, it also has two things going against it: the fact that it isn’t within the official IEEE 802.3 Ethernet cocoon and a certain amount of uncertainty regarding how much that matters to potential customers.
Clearly, Google could care less about the IEEE stamp of approval. It could quite rightly point to a number of successful transceiver MSAs that weren’t developed under the watchful eye of a major standards organization. The two other potential users within the MSA membership – Facebook and AMS-IX, an Internet exchange based in Amsterdam – obviously are at least willing to entertain the idea of being mavericks as well.
However, enterprise applications in general have a reputation for being very conservative and standards-driven, particularly in areas where the IEEE operates. Both the Ethernet Alliance and the IEEE have decided to at least discuss the merits of the 10x10 MSA during 2011, sources indicate. But until such talk leads to action, the lack of IEEE approval presents a quandary to transceiver vendors. Just how many companies have data centers that require 100-Gbps speeds and reaches that exceed the 100 m that 100GBase-SR10 supports (and would rather singlemode fiber than SR10’s ribbon multimode cabling) and don’t worship at the IEEE’s altar? And if the answer to this question is “Not many,” does Google (and maybe Facebook and AMS-IX, plus a few others) represent a market worth pursing – particularly if you’re not certain you can dominate it?
Even module suppliers within the MSA appear to have been slowed by this question. At OFC/NFOEC, the MSA announced that JDSU, Santur (who had jointly demonstrated a prototype 10x10 MSA device the previous fall at ECOC 2010), and newly minted MSA member Oplink would supply modules based on the new specifications. Santur, in fact, had already begun to ship 10x10 modules to companies such as Juniper Networks. However, neither JDSU nor Oplink have announced 10x10 modules as this article is being written in May 2011. In a conversation at OFC/NFOEC, a source at JDSU said he believed there was a demand for 10x10 MSA devices beyond Google, but declined to discuss when JDSU would attempt to meet it. Other potential module vendors within the MSA membership – Source Photonics, China’s Innolight, and Effdon, a startup that hasn’t publicly unveiled a product of any sort yet – have remained similarly quiet.
This doesn’t mean that the Google’s efforts will prove to have been in vain. Scott Kipp of Brocade, who is the MSA chair, told Lightwave at OFC/NFOEC 2011 that three module vendors would probably be sufficient to support the requirement, if the Fibre Channel space is any indication.
So the real 10x10 conundrum may not be whether Google will find vendors to meet its requirements – but whether those vendors wind up happy they did.