The Rise of QSFP – 40GbE to 400GbE and Beyond
The network infrastructure industry is testing 400GbE products and is anticipating the construction of next generation systems. A couple of years ago, there were arguments in the industry about the “correct” pluggable factor of form for 400GbE. But, it seems like that argument has ended, making almost every manufacturer of Ethernet equipment to build solutions based on QSFP-DD.
It is quite vivid, the inspiration for consistently operating based on the form factor based on QSFP. Advantageously, severe market demands are always present to rapidly step up to a single form factor. A considerable price decrease forces a single factor which is critical at high quantities. Light counting estimates of gated content which the volume from initial five years of 400GbE modules is going to be 20 times more than the initial five years of 100GbE because of concurrent adoption in several markets involving Service Providers and Data Centers.
This key market extension with all trying to setup 400GbE at the same time suggests the industry cannot use the following five years to optimize form factors as it has done for 100GbE. Amid each vendor currently developing 400GbE products according to QSFP-DD, 400GbE will assume the leading speed transition that the first form factor is also going to be the high-volume and dense form factor which will support the entire reaches and media.
The 100GbE journey presents some crucial lessons for 400GbE. Even if a dense form factor known as CFP4 is defined and constructed, it is short of drawbacks compatibility about the dense 40GbE QSFP+ module which was ignored as a result. We have no reason to imply a similar thing will not occur in the 400GbE module market if we ignore past lessons.
We need to consider the possibility of extending the QSFP form factor moving from 40GbE to 400GbE to understand the opportunity away from 400GbE correctly.
Amid the goal to make the optical module costs to decrease so fast, it was apparent that a single form factor supporting all reaches was required. The past emphasised this is important to volume adoption in the networking trade. An agreement was also reached that backward compatibility supporting had operational value to a lot and may further speed up the most wanted volume adoption or cost reductions.
With the achievement of QSFP28 for 100GbE, it became apparent that building a compatible solution will be an additional success if achieved. Nevertheless, it won’t come easy and demands a triumph over the impending technical threat. Relaxing the design goals and begin from scratch will be far less intricate. However, the lesson from our innovation experience shows the risk can be managed, and we could adequately brave all challenges.
Within Cisco alone, we developed about 300 unique linecard designs according to QSFP modules. This vital knowledge taught us the possibility of innovating and addressing thermal and high-speed electrical difficulties. From our close association with the major component technology developer, we learned the opportunity to see a path to fitting all the essential components into the existing space, and likewise, it was a tolerable risk. From 3m copper wires to Metr0-reach coherent optical modules, advanced integration was creating solution all the way. However, driving and cooling the modules remain the main concern for the system.
Twenty-four months after, efforts are focused on building and testing QSFP-DD systems and modules designed to drive 400 Gps Ethernet electrical interfaces capable of cooling 20W modules marginally. The 400ZR+ coherent DWDM optical modules are enabled by the 20W target to produce an impressive reach higher than 1000km.
The generally backed QSFP-DD MSA now works to update its specs with these newest thermal abilities. And this is achieved without sacrificing backward compatibility thus imbuing us with the confidence of broader industry adoption and market achievement.
400GbE to 800GbE – The Next Phase
Although 400GbE releases of QSFP-DD are at the beginning of their long exploitation cycle, we’re by now looking ahead to what will follow. By the progress of 100 Gb/s electrical SerDes occurring in IEEE 802.3, we need to look forward to future ASICs to drive 100 Gbps signals along these units.
QSFP modules once more present a clear advantage since we can employ QSFP112 modules pro 400 GbE interfaces (like 400GBASE-DR4) and as well QSFP112-DD for 800GbE capable modules (like dual 400GbE). The swapping of these will be equally essential with the backwards compatibility with the QSFP56-DD already being discussed for 400GbE.
The making of this first generation of QSFP-DD is made feasible for 400GbE by the experience and innovation of painstaking learning of the system, and this is paving the way for more changes and opportunities imbuing us with confidence to crease 800GbE. The lab is already hosting multiple system design methods, and configurations were tested for 100 Gbps electrical SerDes are progressing positively.
The thermal performances advancement and innovation being witnessed for the 400GbE are not the least of the possibilities; further changes and progress are being made. At the same time, having power reduced from 16nm to 7nm processes is indicative of a major power reduction for chips.
When QSFP+ was at its early stages of 40GbE development, over ten years ago, it was difficult for anyone to have thought we would at this point be thinking of the 800GbE options. Nonetheless, innovation is being driven by market success and repeated series of changes to meet ongoing market demands.
It is obvious we’re yet to see the end of this development as QSFP based modules promise a healthy and robust future. Therefore, the networking industry is lucky with abundant of technical, commercial and deployment experience with these units which continues to serve as the foundation to everything being built!