By Craig Matsumoto, Managing Editor, SDxCentral
Carefully crafted telecom standards have played an important role in building the networks we use today, but in the cloud world, where hardware is considered a throwaway commodity, standards don't carry the same importance.
Moreover, the cloud is in a hurry. Giants including Amazon Web Services (AWS), Google, and Baidu don't want to wait for standards.
An IEEE standard is, intentionally, a three-year endeavor. When it comes to 800 Gb/s, specifically, cloud providers don't want to wait that long. They're not even happy with the wait for a 400 Gb/s standard, which is due to be finalized in 2018 and will have taken four years at that point, Arista co-founder Andy Bechtolsheim explained at the recent Linley Cloud Hardware Conference hosted by The Linley Group.
"We don't have time for this. The design activity for 800 Gb/s has to start this calendar year," he told the Linley conference.
Arista Networks is trying to startup a grass-roots effort to develop standardized 800-Gb/s optics, as reported by SDxCentral in early February. It's a deliberate effort to circumvent the usual standards routine, and it could be a harbinger of the future as webscale giants continue to disrupt the networking industry.
And that's just one example. Network functions virtualization (NFV), an effort mostly touching Layers 2 through 7, is being built on a foundation of open-source projects and ad hoc teamwork, rather than a formal standards process. And Facebook is trying to accelerate the commoditization of hardware through the Open Compute Project, an effort built on open-source cooperation.
Unlike a standards effort, open-source work doesn't seek consensus. Anybody is free to try any idea they like. Competing branches of an open-source idea have to duke it out for community support, which usually comes in the form of developers contributing code.
Open source is new to networking, however, and it's not certain how well the two worlds will get along. Even though carriers such as AT&T have thrown serious weight behind open source efforts, we still haven't watched a full generation's worth of technologies wend their way through the open source process. We don't know what to expect.
And this first go-around has been chaotic. In NFV, the number of overlapping projects in the works led to the creation of OPNFV, an effort to stitch together these open-source pieces into a single reference architecture. (And, yes, OPNFV itself is an open-source project.)
The cloud titans have a disdain for standards formality and an indifference to brand names. If they start to represent a bigger chunk of revenue for networking vendors, it's could lead to some uncomfortable changes in businesses from core routers down to optical transceivers. The success or failure of Arista's 800-Gb/s effort might give us a hint as to whether cloud thinking is strong enough to change the networking industry's routines.
Posted: 13 March 2017 by
Craig Matsumoto, Managing Editor, SDxCentral
| with 0 comments