The “second half of the chessboard” parable is a fitting metaphor for the industry to which I have dedicated my career. If you’re unfamiliar with the parable, here’s the gist of it: The game of chess was introduced to a king. The king was so delighted, he offered the inventor any reward she desired. The inventor asked for a single grain of rice to be placed on the first square of the chessboard and to double the amount of rice each week, for each square on the chessboard. The king foolishly agreed, and by the time they crossed the first half of the chessboard, the kingdom was depleted of rice. If they had reached the last square, the rice would have amounted to the size of Mount Everest!
Why does this matter to networking and why share this now? After over thirty years of exponential traffic growth, networks are now crossing the second half of the chessboard. Vigorous growth continues with no end in sight, and the size and scale of the largest networks today dwarf those from just five years ago.
Consequently, new (and bigger) problems have emerged. For example, how can networks grow to colossal sizes and simultaneously increase agility? How can the economics of networking be redefined to lower the cost of ownership? Today, for every dollar spent on networking equipment, operators are spending five dollars or more every year, on operational costs. Power efficiency, climate impact, and greenhouse gas emissions have risen to the top. And projections show, that by 2025, the Internet may consume 20% of the world’s power. If it was its own country, the Internet would be the fifth-largest energy consumer in the world.
The chess-enamored king was short-sighted. Technology leaders must continuously anticipate future problems and commit to innovations years in advance. Cisco Silicon One is one of those innovations. Designed specifically to address the challenges that emerge in the second half of the chessboard, Cisco Silicon One is the engine for our Cisco 8000 series routers. These platforms are already over twice as dense (12.8Tbps per NPU with 19.2Tbps forthcoming in platforms soon) and over three times more power-efficient (2.25 Watts/100G) than any other system available today.
Yet, being king of the mountain over the rest of the industry isn’t the most exciting part. We are most excited about the velocity and acceleration that Cisco Silicon One architecture enables for the future. Historically, new silicon has been released every 3-5 years. Different platforms typically used different silicon depending on size, role, and functionality. Additional time was needed for the operating software to eventually provide feature parity and quality across platforms. Ironically, innovations always seemed to come with delays. That is changing.
In the two years since Cisco Silicon One was introduced, we have released 11 new devices that range from 3.2Tbps to 25.6Tbps. Customers can simplify operations, speed time to value, and reduce costs using a single silicon strategy that spans network platforms from a 3.2Tbps top-of-rack switch (TOR) to a 259.2Tbps modular core router.
And we are now shipping these technologies in volume. Already, some of the world’s largest, future-ready network operators are embracing the value. Meta, Microsoft, and Deutsche Telekom are just some of the customers who have publicly shared their enthusiasm.
Recently, another vendor made some noise about a head-to-head performance test with our Cisco 8000 series router. It was a bit like a soccer match where the host team made the rules, hired the referees, put a Cisco uniform on a mannequin, and then declared themselves victorious. It was particularly cringe-worthy because I believe most technical decision-makers would easily see through the misdirection. Unfortunately, you can’t fix engineering issues with marketing. Customers know right away when tests are designed to make the vendor look good instead of doing the customer well.
To be honest, I get a little bemused when other vendors focus on legacy benchmarks that have little relevance to future deployments. For example, very few instances identify four million routes is vital. When customers do need it, Cisco has other platforms designed to support ten million routes. And hitting mythical line rates of sustained 64-byte packets across an entire system? Everyone already knows this never happens in a real network, because video (large payloads) is the dominant traffic. Trading off density or power efficiency to meet some fantasy metrics has negative implications and results in real costs to customers.
New times. New problems. New solutions. Cisco is building the Internet for the Future. Check.
Massive scale, power efficiency, agility, and longevity. In the future, these are the metrics that matter most because they directly impact the economics of networking. Checkmate.
Great write. The key sentence for me is “Unfortunately, you can’t fix engineering issues with marketing. Customers know right away when tests are designed to make the vendor look good instead of doing the customer well.”
Keeping focus on the customer is and should be the main motivation any company. That’s how it should be in life as well, trying to make the world a better place and not focusing on how to be better then your neighbour.