In hi-tech, impossible just means not yet

Alice, she of Wonderland fame, wouldn’t have made a great physicist. She infamously told the Red Queen “one can’t believe impossible things”. The magenta monarch suggested this was only because Alice hadn’t had much practice.

“Why, sometimes I’ve believed as many as six impossible things before breakfast,” she said.

The exchange came to mind this week when networking equipment leaders Cisco unveiled a revolutionary new silicon chip architecture called Silicon One. It dramatically increases bandwidth and performance, while cutting power consumption by half.

When the requirements for the technology were first laid out, many people in Cisco thought it was an impossible problem to solve, said David Goeckeler, executive vice-president and GM of networking and security at Cisco.

He was speaking from the stage at a press conference in San Francisco. Later, he told Business Times that even when the specs of the final product were circulated, some didn’t believe it could perform at that level. A colleague thought that the number given for power consumption of the chip was an error, because it was so small.

“There’s no way you could do what you’re talking about with so little power — it’s not supposed to work that way,” was the attitude.

“Some of these limitations that we thought about in the past, we just assumed are true,” said Goeckeler.

“What if we don’t assume those are true and we try and build something where we can get past these limitations? That’s what they did with Silicon One. But they didn’t think it might be possible in the beginning. That’s why we didn’t go around talking about this until we could put it in a lab and prove it.”

His colleague, Jonathan Davidson, senior vicepresident and GM of Cisco’s service provider business, said the laws of physics and of business had to be broken.

In particular, Cisco had to do something about Moore’s Law. That’s the principle that the number of transistors on a microchip doubles every two years while its cost halves. It held true for several decades, but in recent years has hit a wall.

“Moore’s Law is stalling,” said Davidson.

“Our current network economics will begin to break as we evolve to operate at massive scale. The physics behind our past achievements are already showing signs of slowing down, while traffic growth continues to accelerate.

“So far, performance increases have helped to reduce the cost of traffic at about the same rate that traffic has increased. $1 in capex today does 11 times the work that it did just a few years ago. However, continuing with the status quo will likely lead to a significant increase in capex unless we reinvent the rules.

“We reinvented from the ground up, the DNA, the performance curve, operations, trust, and even the rules. And these reinventions will allow us to build the future on new architectures — converged, cloud-enhanced, and trustworthy.”

But these won’t be the last impossible problems that need solving. New ones are already lining up. That means the scientists and engineers at Cisco are going to have many more Red Queen moments.

Arthur Goldstuck

‘Network economics will begin to break as we evolve to operate at massive scale’