Many industries operate on the principle of “just-in-time inventory”. Car manufacturers receive the parts needed for the day’s production in some cases only during the previous day. Data centers distribute (load balance) network traffic among a pool of servers, and do not add new servers until all exceed a capacity threshold. This just-in-time approach minimizes unused resources and eliminates idle capital.
This is diametrically opposite to the approach taken when building optical networks today. Standard practice is to overbuild optical routes by 3dB to 5dB margin to compensate for future degradation of the fiber or transceiver performance. These margins often translate into about 100G and 200G of lost bandwidth, per route. In effect, up to half the capacity of capital investment in the network is lost to margin!
An analogy is building a highway to handle vehicle speeds of 200mph even though vehicles only travel at 100mph. Then, as the highway degrades over time, it will still be able to handle the 100mph vehicles. However, about half the capacity of the highway is unused at day 1. Intuitively this makes sense for highways, because the alternative of just building to spec (100mph traffic) at lower initial cost, would necessitate expensive repaving every few years.
One can argue that the same logic should apply to optical networks. The modus operandi until now has been to carefully craft static optical routes which remain in operation for decades. Set and forget. So be it if the price to pay was overbuilding them at a higher cost than necessary to meet initial needs.
The difference is that optical networks are now shaking off their cocoons. Unlike highways, they are transforming from handcrafted static configurations into programmatically reconfigurable entities. They aggregate multiple services like Gigabit Ethernet and Fibre Channel effortlessly onto wavelengths. Then they employ advanced continuous modulation and flexible spectrum techniques to create the fastest possible optical highways for each wavelength. Innovative switching technologies allow reconfiguring the paths of these highways in real time. SDN automation brings everything together, orchestrating this symphony of light to produce novel and beneficial optical networking applications.
One emerging application of this programmability is “just in time” capacity provisioning of optical networks. Network operators will be able to provision routes much closer to their feasible limits, maximizing their traffic carrying capacity and associated capital investment. Instead of overbuilding to a 3dB or 100% power margin, routes would rely on a much lower and more reasonable buffer, perhaps about 1dB or 25% power margin.
When the transmission or the route does eventually deteriorate, programmability can maintain the desired line rate in variety of ways. One way is to employ a flexible spectral grid to increase the channel width. For example, opening the channel from say 50GHz to 62.5GHz (a typical increment) gives the signal more room to breathe. Alternatively, it may be desirable to provision and move the signal over to a brand new route, and release the existing facilities for other services. In other words, the network is continually optimizing itself. When no amount of reshuffling can accommodate all desired optical traffic, then the network optimization algorithm, perhaps based on machine learning by this point, simply orders new resources to expand the overall capacity.
Implementing a just-in-time approach to building optical networks will require network operators to adopt a new mindset. However, fortune favors the bold, and as programmable optical networks become more commonplace, we can expect operators whose goals are to improve their ROI to start leading the way.
Jonathan Homa, is Senior Director Portfolio Marketing at ECI for the last few years. Jonathan enjoys French beer in Nice and loves to discuss the 3 hottest topics in networking today: softwarization, virtualization and cloudification.