QSFP-DD helps speed up 400GbE deployment

Historically, high-speed Ethernet interfaces were initially driven by the density and spectral efficiency requirements of service providers. Optical modules start out big - usually one card at a time. Over several generations, each velocity has been merged into one of two packages: SFP and QSFP. And as the demand increases substantially, the cost and power consumption also decrease gradually.

The five-year growth rate of the 400GbE without copper cable and AOC is also predicted to be 20 times faster than that of the 100GbE (see figure below: comparison of the first five years of the 100GbE and the 400GbE). This opportunity is driving unprecedented industry investment by established companies and startups.

400GbE on the fast lane

The 10GbE took ten years to evolve from XENPAK to SFP+. The 100GbE deployed CFP, CPAK, and CFP2, and then moved to QSFP28 within 5 years. The 100GbE also suffered from CFP4's missteps, but also learned a key lesson that optics must be synchronized with other goals for platforms, networks and businesses. If not, it's a waste of time and money. The size of CFP4 is attractive, but it breaks backward compatibility and doesn't meet all three goals.

Achieving high capacity and low cost for the 400GbE is critical for network operators, chipmakers, optical vendors, router and switch vendors, and many other members of the optical ecosystem. Notably, one of the reasons for the successful rollout of the 400GbE was the convergence of rate and PMD among IEEE, OIF, and multi-source protocol (MSA) members. Unfortunately, the industry initially had two pluggable packaging models, leading to redevelopment and manufacturing. This duplication may reduce the market's ability to extend common solutions.

Disadvantages of the two transceiver encapsulation modes

Key factors for the successful launch of the 400GbE include cost, investment protection and supply chain. When cost is critical, it is important not to over-build functionality. The original 100GbE standard was based on single-mode fiber 10 km. Later, the industry developed short distances to optimize power and reduce costs. The 400GbE will benefit from a faster full range of availability and smaller sizes - from the initial 1m to 10km in 2019 to 100km in 2020.

Another cost driver is achieving economies of scale. Unfortunately, the existence of both module encapsulation formats prevents the market from taking full advantage of the other benefits of consistency. Lean manufacturing is essential for economies of scale, as demand drives output and thus costs. Thus, Shared production line is the key, and large ecosystems with dozens of companies will benefit from standardization. The ecosystem includes developers of manufacturing equipment, test equipment, software design tools, connectors and cages, thermal solutions, compliance and certification equipment. Given the early expectations and rapid growth of the 400GbE, it is important to achieve commonality quickly.

Why is backward compatibility so important?

Over the past decade, the cost ratio between mainframe platforms and optical devices has shifted significantly toward optical devices, a trend that will only accelerate with the arrival of the 400GbE. Compatibility between generations helps offset this trend. By the end of 2019, more than 24 million QSFP modules will be deployed, with an investment of more than $8 billion. Even with the introduction of the 400GbE, demand for the 100GbE QSFP will continue to grow strongly, thanks to the emergence of 100GbE servers and increased bandwidth across the network for enterprises and service providers.

It is not enough to add new devices and run faster on the same network; multiple aspects of backward compatibility must be considered, including the reuse of existing modules and the ongoing investment in the 100GbE. Therefore, only new ports that support existing modules can be implemented. Second, the cost, power, and footprint advantages of deploying the latest routers and switches predate the need for 400GbE functionality. This allows operators to prepare for future growth and benefit from new hardware without having to immediately invest in the first generation 400GbE optics. Finally, there is the need to protect investments in installed routers and switches that are compatible with the cooling architecture (for example, from top to bottom, or side to side). QSFP-DD solves this problem by separating modules and radiators, enabling the host system to be customized as needed.

Where possible, the new generation should try to maintain backward compatibility. Extending third-or even fourth-generation compatibility is technically demanding, but also valuable. The decision to balance investment protection against new requirements is never straightforward. At issue in 2017 and 2018 is the need for encapsulated form conversions at 400GbE, 800GbE and even higher speeds. It is generally accepted that encapsulated form conversions should be made when absolutely necessary due to technical problems or cost. Achieving investment protection, high density, and a full range of capabilities requires risk, but QSFP-DD can address all of this, enabling the industry to move forward with economies of scale.

QSFP-DD challenges and solutions

Backward compatibility with QSFP-DD requires addressing a variety of challenges, including component size and layout, module and system cooling, and electrical connectors that support four and eight channels using 56G SerDes. These factors are interrelated and require consideration of other system components, such as high-power ASIC. Of course, these mechanical challenges are easier to solve for new types of modules that break compatibility with previous generations.

One of the most obvious technical challenges is cooling. The initial 400GbE PMD is expected to require 12W, while the QSFP28 only supports about 4W, so it is understandable why some find it difficult to make the leap. The success of the original goal drove even greater ambition. The 400ZR/ZR+ coherent module, which is planned for 2020, may require 20W. Continuous innovation, including systems and Cage design, has shown that this is possible, and the standards organization will soon approve these solutions for QSFP-DD. The final step in supporting 20W is achieved by adding a radiator to the front of the module. As shown in the figure below.

Another thermal consideration is that optical modules cannot be considered a closed system; they must operate in the overall design of a router, switch, or server. Compared with OSFP, an important feature of QSFP modules is that their smaller footprint allows for greater air intake. This factor benefits the rest of the system, as can be clearly seen in the platform offering both options.

There are many other areas where achieving investment protection in QSFP-DD requires large-scale industry-wide collaboration. Every step in the journey from 40G to 400G represents significant technological progress, many of which were once thought impossible. At this stage, people are starting to address these challenges for future Ethernet speeds, so we should be skeptical of the initial argument that repetitive QSFP has reached its limit.

Managing the supply chain has become a key differentiating factor for the success of hardware suppliers and network operators. Because the deployment of large-scale scalable data centers is so large, vendor diversity is critical, and if the two modular encapsulation patterns persist, each vendor may have to split up supply chain management.

The best market option is to supply in bulk at acceptable cost. Once the module encapsulation pattern is successfully integrated, the 400GbE rollout will benefit from the optimization of all contributors and multiple competitors in the vendor. We cannot repeat the lessons of CFP4 just to pursue an unwarranted short-term risk reduction.

Conclusion

The launch of the 400GbE starts in 2019 and will climb rapidly. The debate over the form of optical module packaging is largely over, and whether the system vendor has chosen QSFP-DD or both, the focus has now shifted to future speeds. In the long run, the 400GbE will mainly choose QSFP-DD. As the industry continues to integrate QSFP-DD, economies of scale will emerge and the 400GbE will reach its full potential.