So we have all seen the benefits of optical I/O technologies - for example in networking, optical technologies have enabled greater distances, higher throughputs, lower error rates, lower latencies... but often at a higher cost than let's say copper. At the same time we have seen copper-based technologies catching up regularly on key optical characteristics. Copper has also remained the preferred technologies for embedded architectures (e.g. Switch, Router ...) from technical and cost perspectives. While some deeper integration of optical technologies was possible in the last decade, the cost of it was often prohibitive for the general market.
Recent progress of laser technologies has changed the game and the miniaturization of optical components has become far more affordable from the component supply and manufacturing perspectives. However the bridging of electrons-to-photons remained a challenging pricing exercise to popularize the utilization of such products. Most recently we have seen in the industry the release of a new generation of photonics components that enable the development of scalable fabrics that can be embedded in greater architectures, affordably. Such components will handle 10's of Tb/s traffics today, 10x to 100x tomorrow. So don't be surprised if optical technologies penetrate deeply architectures for switches, routers... in Telco infrastructures or in business data centers.
But could optical I/O technologies be leveraged in other compute architectures? Server, storage, converged platforms...? The agnostic profile of optical fabrics makes it interesting to converge traffics and manage them independently. The IT industry at large is currently exploring new avenues for PCIe, Ethernet or InfiniBand . Could these networks be "virtualized" over scalable optical passive fabrics? Can we envisage a full optical end-to-end I/O architecture?
I am definitely interested in your feedback on this topic...