The journey to the 400/800 g has been open Which route will you choose?(2)
Outlook for 400G Applications
Over the past few years, enterprise data centers and cloud data centers have become very different. The servers and uplink rates in the two environments are at different stages and require different solutions to support the applications. Enterprise data centers primarily deploy multimode fiber with lower bandwidth requirements, while cloud data centers are based on single-mode fiber with higher bandwidth requirements. The migration paths for 400G applications in the two different environments are also different, as shown in the figure below.
Switch to switch links
As mentioned earlier, 400G applications are initially deployed in cloud data centers for switch-to-switch uplinks, either leaf to spine or spine to superspine. as shown in the figure below, this type of link can be deployed using single-mode Base-8 400GBASE DR4, single-mode duplex 400GBASE-FR4/FR8, or multimode Base-8 400GBASE-SR4.2 for deployment.
Large enterprises are beginning to evaluate the migration path to 400G. The primary deployment approach is a fan-out design, where a single 400G switch port fans out to eight 50Gb/s connections or four 100G connections to optimize port utilization and switch density, thereby reducing costs.
Fan-out applications such as 4X100G or 8X50G can use a plug-and-play Base-8 MTP connectivity system. As opposed to traditional designs using MTP to LC module boxes and duplex jumpers, fan-out applications can be implemented with MTP adapters and MTP to LC hybrid fan-out components, as shown in the 400GBASE-DR4 4X100G fan-out application below
In enterprise data centers where server connectivity is just beginning to migrate to 25G, switch-to-server links can be achieved either directly using DAC (direct-attach copper) or AOC (active optical cable) with QSFP-DD to QSFP28 interfaces, or through structured cabling using transceivers and duplex fiber patch cords, as shown in the figure below.
Beyond 400G Applications
While 400G is now a reality, it is only one stop on a long journey. the IEEE Beyond 400Gb/s Ethernet Study Group has begun work on defining 800G goals based on the 400G logical architecture and expects to achieve the following data center-specific goals by 2025.
l Support short-haul transmission of at least 50 meters over 8 pairs of multimode fiber
l Support at least 100 meters of transmission over 8 pairs of multimode fiber
l Support short-haul transmission of at least 500 meters over 8 pairs of single-mode fiber
l Supports at least 2 km on 8 pairs of single-mode fibers
l Supports at least 500 meters over short distances on 4 pairs of single-mode fiber
l Supports at least 2 km over 4 pairs of single-mode fiber
l Enables transmission of at least 2 km at 4 wavelengths in each direction over single-mode fiber
In addition, IEEE is actively defining 1.6T applications with goals that include supporting transmission to 500 meters and 2 kilometers over 8 pairs of single-mode fiber. 8-pair multimode and single-mode application goals could eventually drive demand for base-16 MTP connections.
While no one can know exactly how applications above 400G will evolve, one thing is certain, 800G and above applications will drive further development of MTP and duplex connectivity devices such as small pluggable interfaces. Until 200Gb/s PAM4 is realized (also within the scope of the IEEE Beyond 400Gb/s Ethernet Study Group), these applications will require the use of more fiber, so the performance of the cabling infrastructure and effective cable management will become even more important.