Faster is better. That is true in many areas, whether it’s cars, service, or networking. Specifically in networking, faster communication means more data can be transferred in a set amount of time. It also means that a set amount of data can be transferred in a much shorter period of time. These are both situations we run into when planning data center networks.
Today, we’re seeing the beginning of 100Gbps Ethernet adoption rolling out. Thanks to the acceptance of newer physical interface components it’s much easier to build a 100Gbps link than it was years ago. When the standard links were 10Gbps, we bundled four of them together to get to 40Gbps. We had to bundle eight of them together and make them run a bit faster to get to 100Gbps. But with the advent of 25Gbps links, we can go back to bundling 4 of them together to get to 100Gbps. But what does it mean as we begin to develop past 100Gbps?
Andy Bechtolsheim is a very smart man. On top of all the other great things he’s researched and founded in his lifetime, he is currently the Chief Development Officer at Arista Networks. Mr. Bechtolsheim is constantly looking to the future to figure out where networking is going. And this video highlights that vision perfectly:
This video is talking about something I didn’t think was going to be coming for another few years. 400Gbps Ethernet. Super fast data center connectivity. I think back on my first steps deploying 10Mbps and 100Mbps Ethernet at the beginning of my career and I laugh. I could never have imagined the kinds of speeds that we’re talking about today. Yet, technology marches on and it’s going to march us quite a bit faster by 2019 if Mr. Bechtolsheim is right.
I love the ingenious way that optical developers are building on the work they did with 25Gbps and 50Gbps to use as a stepping stone to get to cheap 100Gbps and 200Gbps. By solving the riddle of making the individual channels faster, they’re helping us bundle them together to get to higher speeds much faster. I we had to sit around and wait for someone to build a 400Gbps physical layer we would still be two years away.
Challenges In the Fast Lane
Another part of the above video is the frank discussion of the challenges that are being faced when creating 100Gbps. Mr. Bechtolsheim isn’t coy when he discusses the IEEE politics that are dictating the physical transceivers the are being mandated for adoption. Politics in standards boards has always pushed technology adoption out too far for very little gain other than the ability to thump one’s chest and claim that a little part of the standard was “their part.”
Likewise, the physical limitations that we’re starting to see are important as well. I think the discussion with Ethan Banks (@ECBanks) at the every end, while short, was important. Mr. Bechtolsheim believes that 400Gbps is the end of the road for the current generation of fiber optics. Due to things like massive power consumption and the rising heat causing laser issues, it does sound like the next evolution in optical networking is going to come from silicon photonics. That’s going to lead to an entirely new way of thinking about data transmissions as well as a problem with backwards compatibility and interoperability with existing networks. No one ever said that creating the future was easy, and this shift will prove it.
Bringing It All Together
It’s not often that you get to hear someone like Andy Bechtolsheim tell you about the future of your business. The discussion at Networking Field Day was just the way it should have been. Mr. Bechtolsheim laid out the advantages as well as the challenges and helped the audience understand why this technology was important. That’s the kind of briefing that you would do well to pay attention to when you see it. The future happens when people make it happen. And when it comes to 400Gbps Ethernet, the future appears to be getting here very soon.