I joined Infinera on May 2 after multiple rounds of interviews. They sized me up and, of course, I sized up Infinera. After over 20 years in the industry, I had to admit that I was really impressed by a team with this level of passion, intellect, and tenacity. What was most impressive was when I really understood that team was able to do what most in the industry said was impossible: develop the photonic integrated circuit (PIC). This consisted of taking 60+ discrete optical components and integrating them into two fingernail-sized indium phosphide transmit and receive chips. The transmit chip takes in 100G of electrical signals and outputs 10 x 10G streams of light of a tunable color; the receive chip receives the light and coverts it back to 100G of electrical interfaces. Inputs and outputs can be 1GE, 10GE, 40GE, 100GE and many lower speed choices too. This allows the cost of converting from optical to electrical to optical (OEO) to become very reasonable by leveraging indium phosphide that follows a Moore’s Law curve, creating a fundamental shift in approach.
The industry had always said that the cheapest course of action was to keep a photon a photon because OEO was too expensive. To this point, the fundamental direction of the industry was to develop a separate OTN switching layer to pack wavelengths more efficiently and to develop a separate optical ROADM layer to enable the flexibility to switch wavelengths. Since it was expensive to do an optical-to-electrical conversion to get to the OTN electrical switch, the idea behind the ROADM was to keep things optical as long as possible and provide only a coarse level of switching at the optical wavelength level. For example, you could switch all 10G of an optical wave from north to west but nothing more granular. If you did need something more granular, you would switch the wavelength into a transponder that would convert it to a 10G electrical signal and then in turn get connected to a switch. The right amount of traffic would be dropped by the switch and then the remainder of the 10G signal sent back through the transponder, converted to optical, and out to the ROADM. That is a lot of layers, boxes and a lot of connections, phew.
What the PIC delivers is incredible density, scale and cost-effective OEO and this enables a new approach–the collapse of these two layers into a single integrated platform called the DTN. The Infinera DTN is often called a Digital ROADM, and when installed in a network with other nodes it is referred to as the Digital Optical Network (DON). Unlike the analog ROADM approach, with a DON a 10 or 40 or 100G wave could come through the Digital ROADM (based on the PIC) and a 2.5 Gbps or 10 Gbps service could be dropped and multiplexed into another wave going west or dropped as an electrical service right at that location while the original wave continued to the north or west (or any direction for that matter). This approach enables very granular management, which results in extreme flexibility to create and deliver services while at the same time very efficiently packing wavelengths for network efficiency. In other words, it helps create new revenue while saving costs.
“Pretty cool!” I said when I finally understood it. Anyway, this description just scratches the surface of course. There’s a lot more to learn about PICs. Tune into our next blog where Geoff Bennett, our resident product evangelist, will explain even more.