What's the Role of 200 Gbps Optical Connectivity on the Pathway to 400 Gbps?

The ever-burgeoning bandwidth demands on Cloud Data Center infrastructure are intensifying the pressure on optical module providers to enable faster connectivity solutions at volume scales and cost structures. This is fueling tremendous uptake for 100G CWDM4 (4 x 25G) modules and accelerating the ramp to 100G single lambda (PAM-4) modules on the pathway to mainstream adoption of 400G (4 x 100G).

Technology vendors from across the optical networking industry are working hard to drive this progress, leveraging interoperability plugfests among other opportunities to ensure seamless compatibility among a growing ecosystem of components, modules, and switch systems. This activity reflects the urgent need for faster Data Center links, and also underscores the extreme effort and design precision required to achieve coherence among the heterogeneous products coming to market.

With 100G in widescale deployment today and the promise of mainstream 400G deployment seemingly ubiquitous, Cloud Data Centers are eager to take advantage of any and every opportunity to bridge the throughput gap and keep pace with the data deluge. 200G (4 x 50G) optical modules answer this immediate need head on.

ANALOG ADVANTAGES

200G modules provide several key benefits, chief among them the flexibility to leverage a fully analog architecture, the merits of which we assessed in an earlier blog post focused on optical modules for high performance computing (HPC) applications. Though somewhat more difficult to implement than mainstream digital signal processor (DSP) based solutions, fully analog optical interconnects can provide 1,000X lower latency than DSP-based solutions – a crucial attribute for enabling system and network performance at the fastest possible speeds. And while DSPs will remain essential for designing 100G single lambda and 400G modules, DSPs aren’t necessary for 200G module enablement today.

In the absence of DSPs, fully analog 200G optical modules consume much less power and dissipate considerably less heat. Leveraging existing optical components, it’s now possible to enable module-level total power consumption under 22 milliwatts per gigabit. This translates to a 200G optical module for 2km applications with power consumption as low as under 4 watts. A DSP-based module would likely clock in at 2 to 3 watts higher, which doesn’t sound like very much, until you aggregate the resulting power consumption penalty across a Data Center hosting thousands of optical modules. In this context, a 2 to 3 watt power savings per module is hugely advantageous for optimizing OPEX and cooling efficiency.

Low latency and power consumption are important attributes, but not the only performance metrics that matter. Signal integrity is another critical performance criterion given the cascading consequences of transmitting bit errors into the data stream. This poses a particularly daunting challenge as data throughput speeds increase from 100G to 200G and beyond.

The ability to maintain optimal signal integrity performance at 200G in the absence of a DSP is due, in large part, to continued advancements in clock data recovery (CDR) devices and the underlying signal conditioning technology. The newest generation of analog CDRs deployed in fully analog 200G modules have demonstrated the ability to enable a low bit error rate (BER) and better than 1E-8 pre-forward error correction (Pre-FEC), on par with DSP-based 200G modules.

HIGH VALUE, HIGH VOLUME

None of the aforementioned advantages of a fully analog 200G optical module would be worthwhile if the cost structures weren’t approaching comparable alignment with mainstream commercial solutions. But here again, the fully analog 200G module architecture wins against DSP-based 200G modules.

At the device level, the streamlined design of a fully analog 200G module reduces overall component count and sidesteps the costs of DSP development and implementation. At the broader market level, while 100G technology is already mature and component integration is well established, 200G end-to-end interoperable chipsets have just recently hit the market. Looking to the past as our guide, in the short term, 200G modules are expected to emulate cost structures akin to 100G modules when they entered the market a few years ago, and follow a similar downward cost curve as component integration is further standardized and volume shipments accelerate. In due course, 200G modules are expected to achieve a cost structure that’s comparable to today’s 100G modules.

As an intermediate step between 100G and 400G, 200G optical connectivity is a compelling solution for Cloud Data Centers challenged to implement faster optical links at scalable volumes and costs. DSPs will undoubtedly play a pivotal role on the path to 400G, and in the interim, the fully analog 200G module architecture lights the path to faster, cost effective connectivity beyond 100G.

MACOM is committed to leading the evolution of Cloud Data Center interconnects from 100G to 200G and 400G, and at ECOC 2018 we demonstrated a complete, fully analog 200G chipset and TOSA/ROSA subassembly solution that affords optical module providers seamless component interoperability to reduce design complexity and costs. To learn more about MACOM’s optical connectivity solutions for Cloud Data Center infrastructure, visit https://www.macom.com/data-center