In 1990s fantasies of the all-optical network, electronics played a minimal role. Today it seems that the higher bandwidth demands go the more high-performance silicon is actually needed.
In CIR’s new report “400G, OTN AND NEXT-GENERATION TRANSPORT: A MARKET AND TECHNOLOGY FORECAST” (See: http://cir-inc.com/reports/400-g-otn-and-next-generation-transport-a-market-and-technology-forecast), we discuss (among other topics) how the first trials of 400G transport are reliant on innovative silicon strategies; more so than novel optics.
As CIR sees it, the importance of electronics in high-speed optical networking is growing for two reasons: lagging optical technology and qualitative changes in network traffic.
Optical lags: Despite years of optical networking R&D, both real world fiber and real world laser are behind what physics tells us they could achieve in terms of data rates.
DSP electronics has become essential to high-performance modulation technology. It allows more wavelengths to be packed onto a given fiber. DSP can increase spectral efficiency through wave shaping and greater tolerance to nonlinearities
Traffic and intelligence: Higher levels of network intelligence are now required to support (1) unpredictable traffic patterns from millions of mobile broadband sources and (2) to rapidly provision services at 100 Gbps services and beyond.
Traffic unpredictability will only increase with the advent of the Internet-of-Things (IoT). The IoT, if fully implemented as currently envisioned, has the capability to set off trillions of advanced network events. More network intelligence means more powerful electronics, especially network processors. Optical chips on their own are usually powerful but dumb.
For this reason, multi-functional network processors have found a growing role in high-speed backbone transmission and switching. These processors take care of very critical intelligent functionality including packet or frame forwarding, quality of service (QoS), encryption, access control and TCP offloading.
Nonetheless, it is precisely the growing importance of electronics that may be draining the opportunities from the traditional network silicon suppliers. For example, Cisco, Ciena, Alcatel-Lucent and other equipment vendors are now heavily involved in designing and specifying the chips that go into their boxes and are making proprietary chips a key part of their marketing story.
This could probably not be any other way. If ASICs are as important as they seem to be in the context of (say) 400G transport, “home brewing” chips by the vendors makes it easier to protect IP and make the chipsets fit the particular design needs and functional capabilities favored by an equipment vendor. In addition, designing new chips in-house allows equipment vendors to get platforms faster to the service providers giving service providers faster access to the latest systems and innovations.
Does this mean that there are no opportunities for mass-market ASICs in the high-speed transport space? The answer is almost certainly that there are such opportunities. First, some off-the-shelf ASICs can play peripheral roles in advanced platforms for 100G transport and beyond. Second, even ASIC start-ups that can bring highly innovative designs rapidly to market are going to attract the interest of the important vendors.