Modern Era Networking
How do you know you are talking to a person who knows very little about modern networking? When they tell you that they think next-generation networking is about building a spine/leaf networks with legacy protocols so they can have some OS portability like the webscale companies. If you have been reading this blog for five years, I would first like to say thank you and second I am certain that you know the answer is not reinventing the past. I have been writing about the emerging Modern Era of networking for a few years.
Here are a few curated choices for your reading pleasure:
Feb 2016: Four Years Later
August 2015: Networking Will Change
April 2015: A Piece of Wanton and Profligate Ostentation
September 2014: Beginning with a Different Perspective
September 2014: Sales Call Gone Wrong and the Innovator’s Dilemma Moment
June 2014: Do It Again Part 3: It is Easy to Believe in Self-Referential Data Sets
June 2014: Do It Again Part 2: My Thoughts on the State of Networking June 2014
October 2013: Future Generations Riding on the Highways that We Built
September 2013: Just to tell our souls we’re still the young lions…
June 2013: Fluidity of Network Capacity, Commoditization, DIY Openness and the Demise of the Network Engineer
May 2013: The Bigger Picture – Beyond Incrementalism
May 2013: SDN, It’s Free just like a Puppy!
May 2013: Dr. Strangelove: Or How I Learned to Stop Worrying and Love SDN
May 2013: Self-Similar Nature of Ethernet Traffic
Let’s not take my word as I only spend about 60-80 hours a week selling actual networking products designed for the Modern Era of networking. Let’s take someone else’s word, maybe someone who builds at large scale. Here is a video link. Please take the 22 minutes to watch the video and then come back are read the rest of the post.
I hope you thought that was worth the time. Here is what I heard:
- The era of free performance upgrades is over, which means Network Engineers are going to have to design systems that produce quantifiable results. Random results from random devices self-discovering their roles is a legacy design process.
- Networks are the key enabler to bring servers together in groups of 10s, 100s and 1000s.
- Networking is about storage and storage is about the network
- In the Protocol Era no device in the network knew their role (e.g. most random), they self-discovery, which was great to build and scale the internet, but that is not the problem we are solving for today.
- Latency matters
- 3:15 mark…we could not buy a network that could provide the capabilities we needed.
- 5:00 mark…Scale, speed and availability…that is what Google is designing for.
- Design for predictability so that the worse case is close to the average case (This is not possible in the Protocol Era of networking).
- 6:15 mark…Central logical control with algorithms for a very large fabric that makes the system simpler in the end, compared to a fully de-centralized scheme.
- At Plexxi we want to build a simpler network. We want to elevate the perspective from devices to systems. Networking in the Modern Era is done at the system level – not at the device level. In the Modern Era, network engineers design systems that have linkage to the logic of the business goals.
Plexxi is focused on the basics: great products, less complexity, a higher level of abstraction and a network that is not about speeds, feeds, buffers, protocols and layers. At Plexxi we are trying to simplify the consumption and deployment of network infrastructure. For example, our first three hardware switch platforms were Plexxi designs, but our last three hardware switch platforms (one is still unannounced) are by other companies and one of those companies is technically a competitor in the design of networks for the Protocol Era. We have the ability to repurpose a Protocol Era hardware design with a Plexxi OS that enables it to be part of a Modern Era using Plexxi Control and Plexxi Connect.
Our product development focus is not about bigger buffers, but rather about better integration and control to make switches operate as a system. Our solution is not about boxes; it is about the properties and quality of the system experience. It is about centralized logical control, using a variety of hardware options to produce a large-scale fabric that allows for the end-user to design for predictability around the workloads that are most important to their business. In other words we think we got the architecture correct for the broader market and that is why this happened. In full disclosure, Google does not use any Plexxi technology in their network.