Next-generation data center connectivity for 400G and beyond

[[393969]]

The data center industry is experiencing unprecedented growth and innovation as new players, new business models and new technologies converge. Despite the economic slowdown in 2020, Gartner predicts that global spending on data center infrastructure will reach $200 billion in 2021 and will grow annually through 2024.

Driving this growth will require supporting a faster and seamless transition of data, including technologies such as 5G and the applications it supports, which will require processing, analyzing and storing large amounts of data. This will require not only investments in faster transceiver technology, future-proof physical infrastructure, optical cables and connections, but also the simplification of network architectures that have become increasingly complex over time.

Fiber optic connectors will play an important role in this transformation, and over the past few decades, a variety of connector types have continued to increase, some of which have evolved to support the scalability of data centers as they have evolved to 40G and 100G. From the range of connectors released to the market in the past 30 years, there are two interfaces, the LC duplex interface and the MPO/MTP interface, which have become standard connections in data centers.

However, what is appropriate today can change rapidly as the industry evolves and new architectures emerge. As data center operators look for advanced methods to achieve faster data rates such as 400G, we must consider breakthrough approaches to higher connection density and simplified network design.

In addition to existing interface types, the market has introduced new ultra-small connectors, such as the Senko Nano (SN) and US Conec Mini-Duplex Connector (MDC), to meet this need. So far, equipment manufacturers have not released transceivers for these connector interfaces, but they are expected to announce their availability in the next few years.

MDC – LC Integrated Boot Patch Cords and MDC – MDC Patch Cords

What does this mean for your data center?

From tenants in multi-tenant data centers (MTDCs) who pay by the space, to users who build their own data centers to capacity, achieving greater density and preparing for future needs is increasingly important. Even if you may not need to implement 400G in your data center today, increasing density in the main distribution area (MDA) is always valuable, as future demands from new applications or services may consume the bandwidth that has been available to date.

MDC and SN connectors promise to connect directly from one high-speed transceiver to another, simplifying the ability to plug a single connector into a wide range of switches from 400G to 4x100G. In addition, up to three MDC or SN connectors can be accommodated in the same physical footprint of an LC duplex interface, providing a huge density advantage.

SN – SC Simplex Patch Cord

For operators struggling with reduced data center space, implementing hardware that makes LC duplex connections compatible with LC-MDC patch cords is an effective approach. This not only allows the LC duplex interface to be retained on the transceiver side, but the port density using MDC in the same size module can also be increased by up to 3 times - imagine 432 fibers in a 1U rack unit instead of the current 144 fibers.

Ultra-miniature connectors also help reduce total cost of ownership. Various optical component manufacturers have begun to offer solutions with these connectors, but it is important to find the best cabling infrastructure solution that also allows for the continued use or reuse of existing components. This in turn helps minimize the initial investment while meeting future scalability requirements.

In addition to the SN and MDC, the CS Corning-Senko Duplex (CS) represents another option in the ultra-small connector space. It is important to note that while all three types accommodate two fibers, they have many differences in design and functionality, including the vertical/horizontal dimensions of the fibers. The SN and MDC can also be combined as a 4x2 connector, which is not possible with the CS connector. Given these differences, the CS, MDC, and SN interfaces are not compatible, which has implications for the required optical transceiver interface as well as the passive connectivity components.

Beyond Pluggable Optics

The future of the data center industry will see various technological advancements that will help us achieve higher data rates - from 40G to 100G to 400G all the way to 1.6T.

The use of pluggable optics is likely to continue to play an important role up to 800G, and the connectivity interfaces mentioned above can be expected to develop in the future. However, for 1.6T, the high density and power requirements mean that pluggable optics may not be the best solution.

When it comes to these higher data rates, there is another approach, namely co-packaged solutions. Here, data transmission and data processing are combined with semiconductor components, such as Intel's demonstration of a recently released co-packaged Ethernet switch. By doing so, co-packaged optical devices are expected to increase density, reduce latency, reduce power consumption and shrink switch size.

Achieving these data rates will mean moving beyond the ultra-small connectors mentioned earlier. Further developments are expected in the extended beam area, enabling more applications on connectors such as the USConec MXC connector or the 3M EBO. New developments such as multi-core fiber (MCF) and reduced cladding fiber will also have a tangible impact on the further development of connectors.

Preparation is key

A long-term challenge for data center operators is to ensure that network and structured cabling designs remain flexible in order to minimize cabling infrastructure costs when upgrading to 400G and higher network speeds.

Careful planning and preparation will avoid costly upgrades and further changes and are an important part of increasingly compact connector and fiber management to build 400G networks that can serve large numbers of hyperscale data centers, multi-tenant data centers, and enterprise data centers.