MSAs vs. standards for the data center
Technology development and assessment for enterprise networks – and, by extension, the data center – used to be pretty straightforward.
By Stephen Hardy
Technology development and assessment for enterprise networks -- and, by extension, the data center -- used to be pretty straightforward. Standards bodies, particularly the IEEE, would create consensus around a set of specifications for important applications, and the industry and user communities would adopt these specifications as their marching orders.
It became common wisdom that if a product didn't conform to an IEEE specification, network managers wouldn't deploy it. However, such common wisdom and the paradigm on which it rests appear to be breaking apart as attention turns to 100 Gigabit Ethernet (100GbE).
The cracks in the standards foundation emanate from the perceived need for 100GbE modules targeted at reaches from 500 m to 2--3 km. The Google-led 10x10 MSA targeted this application a few years ago; the fact that the MSA hasn't gained traction in many, if any, data centers apparently hasn't dimmed the belief that the IEEE missed the mark not once but twice. It didn't address this application when it developed the original 100GbE specifications, nor last year within its current Next Generation High Speed Ethernet work, critics of the 100GbE PMD roster assert.
Since the beginning of this year, we've seen four multisource agreements appear, all targeting 100GbE over singlemode fiber for this "mid-range" application. They include:
- PSM4 (Parallel Single Mode 4-lane) MSA
- OpenOptics MSA
- CWDM4 MSA
- 100G CLR4 Alliance
The perceived failure to directly address the mid-reach requirement (and the fact that a nascent proposal for a 25 Gigabit Ethernet specification failed to even reach a vote at the most recent IEEE plenary in Beijing) has led some to wonder whether the IEEE has lost touch with the needs of the market. I don't think this is the case.
There's plenty of evidence that the IEEE has registered the desire to treat the mid-range directly. While the current next gen Ethernet task force won't create a PMD specifically for this emerging requirement, it's not because it ignored the demand. The requirement was addressed in the task force's deliberations -- the standards builders just couldn't agree on a technological approach, according to my sources. And we see that the task force charged with construction of the 400 Gigabit Ethernet specifications has included specifications specifically for 500 m and 2 km on its “to do” list.
What I think the current MSA proliferation illustrates is that the requirements of the mega data center -- and how to address them -- remain something of a mystery. That's not only true for standards bodies but for the vendor community that supplies most of the people who participate in the standards-making process. The Googles and Facebooks of the world are becoming more visible, but figuring out whether they should get their own standards (and products designed specifically for them) is something the community is still trying to determine.
The 10×10 MSA effort didn't offer much enlightenment. Perhaps this new spate of MSAs will prove more useful.
Archived Lightwave Issues