Due to the open structural space, the prospect of on-board optical interconnection is bright. The iteration of technology allows miniaturized and low-power pluggable optics to eventually be replaced by Consortium for On-Board Optics (COBO) or Co-Package Optics (CPO). In the future, on-board optical interconnects will dominate the major markets in the data center.
The 50G PAM4 can be widely applied, but not the 100G PAM4 (with DSP) currently. It is expected that a perfect system will be designed to make the 100G PAM4 widely applied, however, the specific requirements of the 100G PAM4 technology for random noise and channel have forced it to choose a trusted and limited application area.
About 100G PAM4, there are only two applications that can be guessed at present: to be used for short-distance interconnections with Silicon Photonics optics; to be used for medium-long distance interconnections (2-10km) with optimized EML lasers.
The data center must leap from 200G to 400G. Currently, only pluggable optis based on Silicon Photonics technology can realize the internal interconnection of 400G data center. We all hope that the 400G pluggable optics based on mature EML technology can quickly replace the application of 100G optics. However, the technical difficulties of 400G Silicon Photonics pluggable optics (such as 400G DR4 modules) are very huge. The 400G technical difficulties mentioned here include random noise and signal integrity, as well as heat and EMC brought by the container model of the data center. The next generation of 400G data centers have to consider more complex technical interference. These technical factors exist both in the macro and in the micro.
For long-distance data center interconnections, we suspect that 100G PAM4 should be feasible, and at least 100G Single Lambda 10km should be successful. However, it is still unclear whether this application is commercially convenient and cost-effective.
New technologies are no longer applicable to the popular mainstream architecture and major products, because any new technology has its own room for reproduction. It is foreseeable that the data center and 5G will move towards a completely different space in the application of new technologies.
For the new technology, the best case will be 100G PAM4 and Silicon Photonics combine to achieve a data center based on on-board optical interconnect. Perhaps, starting with the 400G, the data center can directly enter the on-board era. There is nothing wrong with this. After all, after getting smaller and smaller, it must be the opposite of the object, from the singularity to the open space. The future of the data center is here.
For optical module manufacturers, it is necessary to integrate into the systematic design of the data center, bid farewell to the era of plug and play and listen to the call of new technology. For modular companies with sufficient investment and proven skills in the field of nonlinear optics, on-board data centers are an excellent example of design capabilities.
Mature data centers will still experience 200G applications for a long time. Relying on the various product and architecture designs of the 400G, there is currently a lack of inspiration and guidance, so it will take a long time and a huge cost.
Obviously, the industry's limitations and short history of data centers still lack of precipitation for the development and application of new technologies. After experiencing the subversion of optical devices, according to the current logic, the data center should give way to 5G to continue to lead the development of optical communication. In fact, as long as the data center continues 200G layout, the development and contribution of the data center to optical communication will not be lost to 5G. After all, the current maximum rate of 5G backhaul is only 200G.
The data center needs to calm down and complete the learning curve like any other industry, and then plan a new leap forward. Pluggable optics are used in local applications because on-board optics will go into history, but this will not hinder our confidence in pluggable continuous development and market applications. Companies with the ability to do so must hold two swords: pluggable and on-board.