Open or Proprietary: The DCI in a Connected Universe

The cloud is quickly transitioning from primarily a back-up storage resource and a tool for emergency bursting of unpredictable workloads to more of an operational platform for mission-critical workloads. This means the data center interconnect (DCI) is taking on new responsibilities as well.

As with much of the rest of wide area networking (WAN) infrastructure, the DCI is being tasked with providing more LAN-like connectivity, which poses a challenge for an infrastructure that is most commonly built for long-haul data movement.

According to Market Research Future, the DCI market is on pace to hit nearly $50 billion in value by 2023, a 42 percent compound growth rate. The field represents a broad array of systems, everything from network connectivity to power distribution units, electrical switches, UPS and cooling. But the trend is unmistakable: the more the enterprise comes to rely on both the cloud and even newer data processing facilities on the IoT edge, DCI technology will be necessary in order to circumvent the bottlenecks that would otherwise bring services to a halt.

It doesn’t take a genius to figure out that if someone were to corner the DCI market — perhaps by leveraging an existing dominance in data center networking — then that company would own the means by which every other company on earth needs to connect with an increasingly digital-reliant world. Fortunately, some very powerful forces are working to prevent this from happening. Facebook recently unveiled its Transponder Abstraction Interface, which allows anyone to write code for the Voyager DWDM transponder. This, in turn, means that organizations can build their own DCIs using virtually anyone’s networking hardware, essentially abstracting the physical network connecting disparate data resources.

But while disaggregation on the DCI level may solve some problems, says Viavi Solutions’ Amie Cox, IT executives should be aware of the new challenges it creates. For one thing, the lack of a single vendor means the enterprise will have to assume design and validation responsibilities for the network, which requires a fairly specialized skillset. As well, the enterprise will have to implement its own live monitoring and troubleshooting regimen. Depending on the workloads, this might require much deeper visibility into traffic patterns and even packet loads than is customary today. This isn’t to say that disaggregation is a bad thing, but that you should know what you’re getting into before you start down that path.

Indeed, many of today’s vendor solutions appear to stress the need for adaptability in an increasingly complex networking environment. Dell EMC and Pluribus Networks, for instance, have teamed up to create the Adaptive Cloud Fabric architecture that provides a single, federated network software environment without the need for an external controller. Based on the Netvisor OS, the system provides Layer 2 and 3 networking with VXLAN capability and embedded performance monitoring and telemetry. While the fabric can function internally, it can also support DCI services over a wide area network or dark fiber, with seamless overlay capabilities across multiple hardware configurations as long as each data center has IP connectivity.

An expanding market tends to produce a symbiotic relationship with innovation. The more people want something, the more incentive there is to make it better and cheaper. The DCI is at this stage right now, which means we are likely to see a steady stream of new solutions going forward.

As long as there is constantly swelling demand for more data and new services, the motivation is strong for new and better approaches to networking.

Arthur Cole is a freelance journalist with more than 25 years’ experience covering enterprise IT, telecommunications and other high-tech industries.

Latest Articles

Follow Us On Social Media

Explore More