Outstanding Questions About 400G Transport Deployment
Carriers worldwide are beginning to outgrow 100G transport networks and are taking the first tentative steps towards 400G backbones. CIR believes these ultra-fast networks will create major new revenue opportunities for optical components firms, silicon chipmakers and equipment companies alike. This new business will occur as 400G transport moves out of today's trial phase and 400G production networks are deployed. This shift is unlikely to happen in a big way for a few more years, but the foundations for success in the 400G space are being established now.
The challenges of 400G transport are made a little less daunting by the fact that 400G is as "second time around." That is to say, 400G is—most probably, anyway—the next phase in the 100G core era. Until 100G transport backbones began to appear a few years back, we had seen well over a decade in which transport networks were built around multiple 10G pipes. The technology that needed to be invented for a shift to a transport network infrastructure operating with 100G pipes can be reused to some extent at 400G.
This is encouraging, but any advantages that 400G transport may inherit from recent 100G efforts don't seem to us to help all that much when it comes to addressing the huge uncertainties that remain in 400G sales/marketing and product design strategies. These uncertainties are of three varieties.
The uncertainties associated with understanding how much the current enthusiasm over 400G transport trials can translate into actual deployment is one variety. Another is the relative absence of agreed standards for 400G public network transport. Yet another—and, as it turns out the most worrisome—is how 400G transport platforms can respond to a rapidly evolving service environment.
Trials and Addressable Markets for 400G Transport Aren't the Same Thing
It is easy to be impressed by the accelerating number of 400G transport trial announcements that get headline treatment in the trade press. What is especially notable about these trials is which service providers aren't involved. They are not just the usual suspects; the great ex-PTTs (such as France Telecom or BT) or AT&T or university research networks, although some of these are involved.
Instead, many of the 400G transport trials are at "Tier Two" carriers, often in smaller markets; Australia, Austria, Canada, Poland and the like.
CIR's long experience with such matters suggests that trials of this kind don't happen without key managers at a service provider being enthusiastic supporters of a particular technology. So we think that these trials are proof that there are already plenty of 400G transport advocates at the service providers who genuinely want to see their networks upgraded to 400G in the next few years.
This is a good thing from the perspective of serious revenue generation from 400G transport in the near future. But to what degree can one extrapolate from the 400G transport trials to existence—and especially the timing—of actual addressable markets?
The motivations and financial arrangements behind trials are always murky. And the trial announcements are typically made with strong involvement of the 400G transport equipment providers; firms with something to sell. So other questions must be asked. How many carriers are really going to expensively upgrade their networks to 400G in the next, say, five years? How many may be able to make do with 100G or 200G?
And when the time for 400G deployments finally rolls around, will the initial substantial expenditures on them come mostly from long-haul and submarine routes; which would be the natural expectation for those of us who have watched the networks evolve over the past 20 years? Or will it be the metro networks that are the first to take up 400G transport?
The Service and Standards Environments Add More Uncertainties for 400G Transport
There is no surefire way to answer these questions—although we attempt to do precisely this in this report! In particular, the issue of which part of the network will first see 400G prove most strategically important is especially hard to answer, because the answer seems to rely on an applications layer environment that is—in CIR's opinion—entering a period dramatic change.
While established telecom hardware and service firms have all been through major transport network upgrades before and have an intuitive grasp of the risks and how to ameliorate them the new applications that 400G transport pipes will have to amalgamate are different in both their bandwidth and performance requirements, placing new demands on equipment and components suppliers and creating new opportunities, too.
The point here is that in the past, the traffic patterns that new transport platforms have had to cope with once deployed, were fairly well understood. This is simply no longer the case and for three main (interrelated) reasons.
The Internet-of-Things—adding unpredictability to the backbone: First, while the network has always had to cope with some degree of distributed intelligence, the level of distribution is about to shift upwards because of inter-related trends that are variously known as the Internet-of-Things (IoT), Machine-to-Machine (M2M), and by a few other names.
The essence of these concepts is that networked intelligence will be everywhere, especially in the form of sensors and actuators of various types. The point here is that the number of nodes in the network will increase by a couple of orders of magnitude, and this cannot help but make a difference at the transport level.
No one can be entirely sure right now what exactly this difference will be. But it is clear that traffic on the backbone will be more unpredictable because each network node will in a sense have a "mind of its own."
Ubiquitous broadband and "big data"—adding bandwidth requirements to the backbone: Much of the traffic from IoT devices would definitely be characterized as narrowband—sensor output and control data is typically not all that bandwidth hungry. However, in aggregate, IoT traffic constitutes what is becoming known as "big data," which may indeed raise bandwidth issues at the backbone level.
All of this suggests a future in which backbones built between bandwidth-stressed points on the network find that bandwidth requirements among these points and even which points are stressed are constantly changing. We note in this context that this isn't entirely a new phenomenon, since broadband data traffic—increasingly represented by millions of video content creators—is already the order of the day.
New content providers take control away from the service providers: All of this means that the service providers are losing control of their networks. Even though the service providers never traditionally had much say over content creation, bandwidth-hungry applications in the past have been few or far between. And while the ability to create low-bandwidth voice traffic was always in the hands of millions of individuals, the service providers were in the position to tightly control how that traffic travelled across public networks.
The new world of IoT and highly distributed content providers threatens to disrupt future backbone facilities, and transport equipment firms—including those that provide 400G transport equipment—must respond to this, as must those companies that create enabling technology in the form of optical components and specialized chips. None of these firms are helped by the fact that at the present time the standards development work for 400G transport—at the OIF and ITU-T—is just getting underway.
Equipment and Components Firms: Together Again at 400G?
All this obviously creates considerable uncertainty—and hence business risk—for equipment makers and components firms that are becoming involved in the ramp up to 400G transport. They are in effect being asked to compete in a market where both the addressable market and the requirements of that market are extremely hard to calculate.
It is clearly hard for these firms to calculate the ROIs on their 400G transport network development projects in these circumstances. At the very least, the risk discounts that are injected into such calculations need to be considerable. In this difficult business environment, a number of strategies suggest themselves.
400G platforms—the old versus the new: Any transport network equipment or component firm that chooses not to have anything to do with 400G is soon going to find itself in the ash heap of telecom history. Service providers aren't going to want to buy from equipment makers that do not have a good 400G story to tell.
At this stage in the game this story may not need to be complete. No telecom equipment firm yet has a platform that is specifically designed with 400G in mind. Most of the more advanced 400G transport platforms in the current generation of trials use technology primarily designed with 100G transport in mind.
By using earlier generations of telecom hardware to carry out 400G trials, equipment manufacturers can reduce their risk in this space significantly. Some of the trials are using platforms that began their life in the 10G era but have been significantly upgraded for these trials.
Providing for the new service environment—a source of competitive advantage: While 400G trials and upgrades are inherently about bandwidth, CIR believes that an important source of competitive advantage in this space for the equipment providers will be the ability of 400G transport platforms that are up to the task of supporting the very demanding service environment that we see emerging over the next decade.
As we discuss later in this report, each equipment vendor is going to have its own idea of how that can best be achieved, and we note that several—arguably most—leading telecom equipment vendors have not really shown their hands in this regard, although some have.
CIR believes that there is an interesting debate that is about to emerge and that will take place in various venues including standards groups about what is really required here. But at one level the answer appears to be as much as possible. This is to say that the 400G platforms that emerge over the next few years will have to provide the maximum amount of flexibility or—putting it another way—make the fewest assumptions about what traffic looks like.
In practice this means that 400G platforms must be able to support any service that the network can throw at it on any wavelength and in any direction. Given that services and applications will arrive from millions of sources over which the service provider has very little control, the traffic/wavelength management and performance management must be of a very high level and must be ubiquitous.
We have a lot more to say about this in the main body of our report. However, we note that what is being asked of 400G platforms is something very close to the agile network concept that was rather popular among telecom analysts in the 1980s, a time in which the idea of a ubiquitously switched optical network was being widely discussed. It is interesting that Alcatel-Lucent has recently reintroduced that term in the context of 100-400G networking.
Optical components and silicon at 400G—the old made new again: Although no one really talks about all-optical switching as a key enabling technology for 400G, our sense is that a lot of other advanced optical technologies that have been theorized about for many years are now becoming essential at 100G and 400G transport. The most notable of these is probably coherent transmission, which is now used in almost all 100G transport networks and will probably always be used at 400G—coherent transmission is an idea that dates back to the 1980s.
Another old optical technology that CIR expects to see a lot more of in a 400G environment is Raman amplification that will come into its own at 400G on long-haul routes. On the silicon side of the house, DSP chips also seem likely to increase their importance, both as way of implementing performance management and the new modulation schemes (e.g. dual-channel 16 QAM) that will accompany the deployment of 400G transport networks. DSP chips will also be used to reduce power consumption and adjust PMD.
This raises the interesting question of who will control component/chip technology going forward. At one time, optical component and (to a large extent) telecom silicon was manufactured by the leading optical equipment providers. Then most of this capability was divested or subcontracted to third parties.
But not surprisingly, the balance of considerations with regard to this issue has changed since then and CIR's sense of the market is that increasingly the networking equipment makers may want to control their own chips and optical components, or, at the very least, emphasize the role that these chips play in the uniqueness of their products.
This means more custom design and possibly more involvement in manufacture of the chips. There are already signs of this among the current providers of transport platforms for the 400G trials. At one time only Infinera was alone in making its core chip a key part of its product/marketing story for its long-haul platform. We note, however, that both Alcatel-Lucent and Ciena are also doing this.
Meanwhile, both Cisco and Huawei have acquired silicon photonics technology for in-house use, although this technology is presumably intended for a much broader range of products than just 400G platforms