Tuesday, March 24, 2015

Blueprint: Beyond 100G in Core Networks – Is Coherent Technology Reaching a Plateau?

by Maxim Kuschnerov, Coriant

100G coherent technology has paved a successful path in terrestrial core and submarine networks since its introduction in 2010. In core networks, a figurative marketing war on beyond 100G capacity has been taking place for some time, with vendors announcing 200G, 400G, and IT transport solutions. However, it has been impossible to ignore the fact that beyond 100G tunable line side interfaces do not live up to the general evolutionary trend of 10G to 100G. While the previous per-channel capacity step represented a 10-fold increase without sacrificing reach or channel-count, a similar step seems very unlikely for interfaces beyond 100G.

Flexi-rate Interfaces - Not Quite Yet What you Would Expect

100G connectivity in the core started out with a 4QAM/QPSK modulation scheme, as shown in Figure 1, where binary electrical signals are converted to a format with four constellation points, which is transmitted in two orthogonal polarizations. The applied coherent detection technology is capable of detecting arbitrary multi-level schemes, which can be used to transmit more bits per time slot. Borrowing from a 30 to 40 year old playbook from wireless communications, optics turned its eyes to more flexible modulation formats like 200G 16QAM (see Figure 1) to increase capacity throughput. While the capacity increases by 100% compared to the classic 100G 4QAM, the reach is roughly a quarter. After a limited market introduction in 2014, this year we will see the first wave of commercial 100G/200G 16QAM interfaces. However, 200G 16QAM leaves no room for margin in manufacturing and does not have the wide appeal of a carrier-grade interface due to its inherent performance limitations. While sufficient to cover a portion of demands in long-haul networks, it is not quite the game changer.

When analyzing the actual reach requirements of core network demands it is clear that a more powerful solution to the minimal reach of 16 QAM transmission is required. Fully flexible transceivers should include a middle ground format of 8QAM. With reach of up to 2,000km 8QAM based solutions hits an ideal sweet spot in core networks while still providing an increase in spectral efficiency of 50%. Wide deployment of flexi-rate interfaces in core networks will not come until the adaptation of the 8QAM format is included in the solution, then living up to the promise of a single-spare fully flexible line interface. Figure 1 illustrates the reaches of a flexi-rate interface with multiple modulation schemes on a cumulative distribution of demands in core networks. 8QAM is clearly the undeniable working horse.

Figure 1: Applying flexi-rate interfaces to core networks with Raman amplification (Note: absolute distances of each scheme depends on fiber and amplifier type, span losses, channel counts, end of life margins, and error correction limits.)

Pushing the Limits – What’s Next in Terms of Differentiation

Flexi-rate interfaces for core optical networks are coming close to the theoretical boundaries of maximum channel capacity. It is virtually impossible to deliver a 400G or 1T long-haul interface while keeping channel count and network architecture the same. Digital signal processing offers still some room for improving error correction codes or the holy grail of fiber optic communication – i.e., algorithms for the compensation of fiber nonlinearity – but the return on investment diminishes drastically. Going forward, optical layer technology like Raman amplification or C+L band systems will be quintessential to increase core network capacity without requiring completely new fiber transmission technologies. Figure 2 shows an overview of techniques going forward for fiber capacity increase.
Figure 2: Improvements for core network channel capacities going forward vs. state-of-the art coherent interfaces

It is clear that the majority of innovation for capacity resides in the optical layer and no longer in signal processing chips. Since these integrated circuits become more expensive with each generation, their product definition has to find a good balance between development costs, power consumption, and feature set. Fighting for a 10% reach improvement with a disproportionately high investment cost is unlikely to lead to a profitable business model for many vendors. Moreover, optimization for very high-end optical performance is likely to reduce the viability of these solutions for lower end applications such as metro transport or datacenter interconnects, which are driven by low power and high density requirements. Thus, the trend for 400G/1T interfaces will shift away from pure optical performance improvements which are valuable only in long haul applications and instead focus more on solutions that can leverage the same design across multiple markets spaces, while delivering the most flexible level of architectural feature integration at the lowest power consumption. This reflects a similar experience in consumer market. While processing speed was one of the major buying criteria in the first age of personal computing, the value perception has shifted towards the application or the battery life of the device. The actual processing abilities of a smart phone or a laptop become of much lesser importance if not practically irrelevant.

Evolution of Core Networks in Light of the 100G Metro Surge

100G is on its path conquering the metro market using a mix of coherent and direct detect pluggable interfaces. Coherent technologies designed for the higher end of the metro market (>40-80km) will inevitably take a large share of the long-haul market. This mostly depends on each vendor’s ability to deliver low power flexi-rate interfaces. While these do not require high-end error correction abilities or highest performing optics, the reach and cost position of these interfaces will make the long-haul market an offer that it simply can’t refuse. A true long-haul network equipment manufacturer will not be able to survive without the mix of high-end and low-end coherent interfaces. While the earlier are likely to be proprietary and more power hungry, the lower end segment is on its way towards generic pluggable interconnects. Figure 3 highlights possible market coverage of several competing interface technologies. 

Figure  3 – Technology market segmentation for colored interfaces

In conclusion, the way forward to 1T interfaces is not a naturally outlined evolutionary step. Multi-channel interfaces will be required for both short reach and long-haul. While the war on costs in core networks is far from over, it looks like the industry is about to experience a different innovation pace in the optical performance craze.

With the performance of digital signal processing chips leveling out, the further evolution of photonic components could be key to enabling the next wave of core interconnects.

About the Author

Dr. Maxim Kuschnerov is a Product Manager at Coriant, Munich, responsible for high-speed interfaces and photonic layer technology. Since 2007, he worked for Coriant (former Nokia Siemens Networks) focusing on the development of signal processing concepts and coherent optical transceivers. He has authored and coauthored more than 100 peer-reviewed papers and conference contributions. Moreover, he was a project leader in the advanced research project ModeGap, developing space division multiplexing network technology and pushing the photonic layer vision beyond 2020.

Got an idea for a Blueprint column?  We welcome your ideas on next gen network architecture.
See our guidelines.