As I mentioned in my previous blog, pluggable optics (aka optical transceivers) are widely deployed inside the data center, campus, enterprise, or service provider central office. The Cisco Global Cloud Index estimates that total data center traffic (all traffic within or exiting a data center) will reach almost 20 zettabytes per year by 2021, up from 7 zettabytes in 2016. Data center traffic on a global scale will grow at a 25 percent CAGR, with cloud data center traffic growth rate at 27 percent CAGR or 3.3-fold growth from 2016 to 2021.
Network connectivity speeds continue to increase to facilitate this growth in traffic as we see migration from 10G to 40G to 100G and now 400G. Today, 100G pluggable optics are deployed in large scale data center applications, with plans for 400G migration as router/switch ports now support 400G.
The growth in “within data center” traffic accelerates the need for next-generation networking equipment to support higher port density and faster speed transition. This in turn drives the requirements for large scale deployment of high-speed optics to connect the various layers of the networking equipment. As router/switch port speeds have increased, the cost/bit has steadily decreased from advances in silicon (ASICs), which continues to benefit from Moore’s Law. However, while the cost/bit for pluggable optics has also decreased, it has not come down quite as fast as the router/switch port cost.
The result is that as the bit rate increases, pluggable optics represents a larger fraction of the total hardware cost. As an example, at 10G, optics represented about 10% of the total hardware cost of a data center network. As we progress to 400G, optics represent well over 50% of the total hardware cost.
As optics speeds increase, so does the complexity in the design and manufacturing of pluggable optics. Beginning with 100G speeds, this has caused pluggable optics to lag behind the availability of router/switch ASICs, and generally delaying network migration to higher speeds to meet the higher bandwidth demand.
A future trend in the data center relates to the capacity of ASICs driving routers and switches. ASIC performance will continue to increase – from 10Tb to 25Tb to 51Tb and beyond, requiring corresponding increase in the electrical signal rate of each I/O pin (input/output) in the ASIC. This electrical signal rate will advance from 25Gb/s now to 50Gb/s and then to 100Gb/s to enable all of the capacity to enter/exit the ASIC. Each time this rate increases, the challenge in moving the electrical signals across a linecard to the front panel of the router/switch becomes much greater. To solve this challenge, the industry will, at some point, come to rely on optical signals vs. electrical, even to travel just a few inches across a switch linecard! Hence co-packaging of optics and silicon will be a necessity in the future in order to assure continued innovation in routing and switching.
To ensure we are able to deliver solutions that meet our customers’ time to deployment, performance, power and cost expectations, Cisco is making strategic investments in optics technologies (e.g. silicon photonics). Optics, along with silicon and software, is a foundational technology in support of Cisco’s continued innovation in routing and switching.
The optics industry still largely relies on a “discrete” set of components with very cumbersome testing and assembly processes. If we contrast that with the semiconductor industry – where extremely complex functions are integrated into silicon (ASICs) using highly automated processes that have very little to no reliance on human interaction – we see an opportunity to leverage these processes in the processing of “photonic signals” rather than “electrical signals” that semiconductor processes naturally manage. The promise of silicon photonics is to leverage the semiconductor industry processes to more fully automate the process of optics manufacturing.
Cisco acquired Lightwire in 2012. Lightwire was an early innovator in the area of silicon photonics.
Cisco has developed a portfolio of products based on the original Lightwire technology and continues to enhance that portfolio.
The recent acquisition of Luxtera was motivated by the facts that:
- Luxtera has a portfolio of 100G/400G products leveraging silicon photonics technology targeted at data center applications that will help fill out the Cisco portfolio
- Luxtera has developed industry-leading automation of the manufacturing process, delivering on the full promise of silicon photonics to leverage the semiconductor industry’s process and automation technology
- Luxtera has developed mind-share in the industry around embedded optics, which are an important stepping stone on the way to co-packaged silicon and optics
In summary, the data center relies heavily on optics for all but the very short (<10m) applications today. 100G optics are the dominant technology being deployed in data centers today, with an anticipated migration to 400G and beyond. As the costs of the routing/switch ports continues to benefit from silicon improvements, optics becomes a larger portion of the overall spend, and is therefore a key technology for Cisco to develop in concert with silicon to assure we are matching silicon performance improvements, as well as driving improvements in cost and power. Over time, we know that integration of silicon and optics will be required in order to continue expanding the capacity of switches and routers. Cisco will continue to invest in both silicon and optics to assure we are able to deliver those solutions to meet our ever-increasing customer demands for capacity improvements.