contact contact
Tags: Cloud and Data Center
April 21, 2016

Breaking Down the DCI Market

Stu Elby 1Stu Elby

Sr. Vice President, Data Center Business Group

As data center interconnect (DCI) bandwidth and investment continue their rapid growth trajectory, driven by the growth of cloud-based applications, content and services, we are seeing some clear trends in DCI market segmentation.  Analysts and industry observers have recognized that “DCI” is actually a range of applications that benefit from a variety of technology solutions, from short-reach pluggable optics all the way to ultra-long-haul dense wavelength division multiplexing (DWDM) systems, and from compact appliance-like devices with plug-and-play simplicity to a variety of more complex approaches.

While data center interconnections have been around almost as long as data centers, DCI technology has evolved rapidly in recent years.  When Infinera launched the Cloud Xpress in late 2014, it created a new category of DCI solutions.  Market research analysts started tracking this category, calling it variously “small form factor (SFF)” or “purpose-built” DCI, and projected that it would be the fastest growing category of DCI optical systems.  ACG Research and Ovum both project growth of this new category at over 60 percent per year for the next four to five years.

Figure 1: DCI Optical Systems Forecast by Type (Source: ACG Research, 2H-2015 Worldwide Optical Data Center Interconnect Forecast, April 2016)
Figure 1: DCI Optical Systems Forecast by Type (Source: ACG Research, 2H-2015 Worldwide Optical Data Center Interconnect Forecast, April 2016)

Distance has long been recognized as another important segmentation variable.  DCI links can be any length, from a few hundred meters to thousands of kilometers (km).  In market forecasts to date, analysts have broken down optical DCI into only two categories, metro (generally within about 600 km) and long haul (anything longer).  As metro DCI has evolved and multiple DCI technologies have been deployed for varying applications, it has become clear that we need a more granular segmentation based on distance.

ACG’s new service provider survey, mentioned in a recent blog, provides valuable new data about the expected distribution of DCI links now and in 2019.  As shown in the top of Figure 2, about 30 percent of links are less than 30 km and the rest are distributed across a range of distances. 

Figure 2: DCI Distance Distribution (Source: ACG Research), with Infinera Summary of Technology Alignment

 

The ACG blog also mentions an important insight about how DCI technologies address different distances.  ACG’s key observation, which aligns well with our conversations with customers, is that the shorter DCI links are often built using short-reach pluggable optics.  A majority of the DCI links using pluggable optics are not captured in the market forecasts for optical DCI systems built on coherent DWDM technology.  We’ve indicated that point graphically with the arrows at the bottom of Figure 2.

This is an important point in understanding how an embryonic 100G DCI technology, PAM-4, fits into the market. At the OFC conference in March, Inphi announced a new pluggable module with claims of up to 80 km reach.  We believe that may be a challenging target given the complexity of the technology and the variable quality of existing fiber, and that real-world deployments of PAM-4 will mostly be seen in the under 40 km range, at least in the near term.  That means PAM-4 will primarily be an alternative to existing short-reach pluggable optics, and will overlap only slightly with applications using coherent DWDM.

In those distance ranges where either PAM-4 and coherent DWDM technology could be used, we agree with ACG that a few other factors will limit PAM-4 acceptance.  In particular:

  • Capacity: PAM-4 is limited to 4 terabits per second (4T) per fiber pair, while coherent 16 QAM systems will achieve around 27T.  Anywhere capacity demand is high and fiber is scarce or expensive, coherent technology will be preferred
  • Operational complexity: At any distance over about 7 km, PAM-4 is not a plug-and-play solution.  External amplifiers and dispersion compensation modules must be deployed and tuned, leading to complicated link-by-link design and commissioning.  Fixed wavelength modules create complexity in sparing and capacity upgrades.  New generations of technology will not be deployable on the same fiber with the first generation, leading to forklift upgrades

In our experience, most DCI users are not optical network experts and they value the extreme plug-and-play simplicity of an optimized DCI appliance such as Cloud Xpress.  Even sophisticated DCI users will not tolerate the PAM-4 solution complexity in many of their applications, for example, when using DCI to connect to peering partners in neutral Internet Exchange facilities where there is no equipment available to host pluggable modules, or when deploying DCI in international markets where there is no operations staff with optical networking expertise.

Combining all of these factors, we believe that PAM-4 is appropriate for a narrow set of applications where the links are short, the fiber is cheap, high quality and plentiful, and the operator has the right type of networking equipment and advanced optical engineering and operational expertise.  Just like existing pluggable optics solutions, PAM-4 adds a new pluggable option that is largely complementary to coherent DWDM systems.

In sum, we’re happy to see that analysts and industry observers are beginning to further segment the DCI market and educate the industry about the complementary nature of DCI technologies.  We continue to see a bright future for Cloud Xpress product family and our other DCI solutions.

For more information on Infinera’s Cloud Xpress, please contact us.

Related links:

Legal Disclaimer