Wu Jian, technical director of CommScope North Asia
With the increasing demand for bandwidth, capacity and lower latency, the migration to faster network speed is promoted. Therefore, every year, it is a test for the adaptability and survivability of large-scale cloud data center. In the past few years, the link speed of data center has increased from 25g / 100g to 100g / 400g. After each further bandwidth leap, there will be a short pause, and then the head of the data center needs to prepare for the next leap. At present, the data center is committed to jumping to 400g. For data center managers, one of the key considerations is which optical communication technology is the most suitable.
Image source: nextplatform 2018
400ge optical fiber transceiver
With optical device manufacturers trying to take the lead in the field of data center, the cost and performance will promote the development of 400g optical communication market. In 2017, cfp8 became the first generation 400ge module for core router and DWDM transmission interface. The size of the module is slightly smaller than that of cfp2, and the optical components support cdaui-16 (16x25g NRZ) or cdaui-8 (8x50g pam4) electrical I / O. Recently, the focus of the industry has shifted to the second generation 400ge modules: qsfp-dd and OSFP. These modules are only thumb sized and are designed for high port density data center switches. They can achieve 12.8 TBPs throughput in a single rack through 32 x 400ge ports, and only support cdaui-8 (8x50g pam4) electrical I / O.
Although cfp8, qsfp-dd and OSFP are hot swappable, this is not true for all 400ge transceiver modules, some of which are directly installed on the host printed circuit board. The PCB routing of these embedded transceivers is very short, which can achieve low power consumption and high port density. Although embedded optical devices have higher bandwidth density and per channel rate, for 400ge, pluggable optical devices are still favored by Ethernet industry because they are easier to maintain and can bring cost-effectiveness of pay on demand.
Long term vision of “starting from the end”
For the industry, the transition to 400g is another milestone in the development of data center. MSA (multi source agreement) working group has carried out 800g research based on eight 100g transceivers. As a member of the 800gpluggable MSA working group, CommScope is working with other IEEE members to develop solutions to support multi-mode fiber based 100g per wavelength server connections. The research is in progress, and the relevant scheme is planned to be put into the market in 2021, and then the 1.6T plan may be launched in 2024.
Although the details involved in the process of moving to a higher speed are daunting, it also helps to examine the process more objectively. With the development of data center services, the speed of storage and server must be improved. In order to support higher speed, appropriate transmission media are needed. When choosing the optical module that can best meet the needs of the network, we need to have a long-term vision of “starting from the end”. Because the more accurate you predict the services you need and deliver, the better the network will be able to support new and future applications.