Cisco’s single-chip ASIC-based architecture dubbed Silicon One breaks so many well-established precepts, it can help to approach some of the biggest ones first. In fact, once you can accept that a single silicon design, can work well for both routing and switching, other breakthrough claims become easier to swallow as the hyperbole begins to make sense.
Distinctions between routing and switching often feel superfluous to the casual observer. Hardware differences are justified by the unique role or place in the network. A network, by definition, requires coordination between elements built from different architecture. Network operators dedicate many resources to refining the orchestration these elements require to squeeze maximum results as they optimize both speed and efficiency. Operators have had to spend a lot of time testing features across the network as a result. Silicon One offers the first single silicon architecture serving different market segments and different functions in its bid to sharply cut operator OpEx.
Now, I prefer my marketing promises as more of an amuse-bouche. Something tasty and promising upfront immediately followed with a profound entree of technical proof points. At first pass, the sheer number of marketing promises from this launch threatens to overwhelm the palate.
This podcast is the inevitably incomplete back story for Cisco Silicon One. This audio will more than likely trigger you in multiple ways. I only ask that you listen closely and do your best to push past any cognitive bias this reveals.
Research shows that when people are resistant to change, it helps to reinforce what will stay the same. Visions for change are more compelling when they include visions of continuity.
- From the book, ‘Think Again,’ by Adam Grant.
Listen and learn as we explore:* What it means for Cisco to be a fabless semiconductor company. * How research showing diminished returns for the entire industry drove Cisco to re-think what was possible. * Improbable first meetings with the creators of the Jericho DNX line of chips. * The audacity to consider erasing well-worn boundaries between routing and switching. * How architecture level changes can ripple into completely unexpected efficiency gains. * Striking the balance of engineering trade-offs inherent to different design choices. * How to read the sub-text inherent to the glowing network data sheets we spend on for better decision making.
00:00 - The Launch
05:44 - Part 1: Fabless
15:04 - Erasing Hard Boundaries
24:07 - Value Proposition
29:35 - Part 2: Making Memory
41:10 - Part 3: Router comparisons are not easy
47:51 - It’s Amazing