Written by Ray Nering
In my previous blog post, I wrote about the new QSFP-DD800 MSA group, which released its Rev.1.0 specification for the QSFP-DD800 form factor back in March. This is just one step in laying the foundation for the inevitable need for higher data rate pluggable optics at some point in the future.
In addition to the new QSFP-DD800 form factor MSA, there has also been considerable effort to standardize 800G signaling. The 25G Ethernet Consortium has re-branded itself as the Ethernet Technology Consortium and has released an 800GE specification. Originally established to develop 25, 50, and 100 Gbps Ethernet specifications, the 25 Gigabit Ethernet Consortium announced on April 6 that it has changed its name to Ethernet Technology Consortium in order to reflect a new focus on higher-speed Ethernet technologies. The goal of the consortium is to enhance the Ethernet specification to operate at new speeds by utilizing specifications that are already developed or in development. This allows the organization to work alongside other industry groups and standards bodies to adapt Ethernet at a pace that aligns with the rapidly evolving needs of the industry. The ETC has more than 45 members with top-level promoter members that include Arista, Broadcom, Cisco, Dell, Google, Mellanox, and Microsoft.
The new ETC standard specifies an 800G implementation based on 8x 106 Gb/s lanes, which is very much in line with the QSFP-DD800 MSA. The 8 lanes use a 2x 400GE PCS layer to connect to a single MAC operating at 800 Gb/s. The interface can also be used as a 2x 400GE interface. In principle, 2x 400G PMDs could be used to form an 800G interface, for instance, 2x400GBASE-DR4 modules. However, for now, the specification does not define an 800GE PMD (Physical Medium Dependency).
The specification leverages many of the industry standards that have been determined or are in progress. For example, it reuses the same FEC algorithm that the IEEE used in its 400G standard, RS(544,514). This makes it easier for next-generation form factors to use the 800GE MAC for 2x 400GE applications. Also, the SERDES will leverage the IEEE’s efforts in the 802.3ck task force, providing 100G per lane. So new modules in the QDFP-DD800 form factor could potentially have eight lanes of 100Gb/s for 2x 400GE modules.
I expect the initial next-generation modules will not be 800GE, but 2x 400GE. Over the next few years, 400GE will become the de-facto standard interface for core network connectivity with 400GBASE FR4 and 400GBASE-LR4. 400GE ports will also provide high-density 100GE connectivity using DR4 and 4x100G modules via breakout cabling. When the industry is ready to start deploying higher capacity ports, 2x 400GE modules will initially hit the market for several reasons. 400GE ports will be very common, and 2x 400GE modules will be a convenient way to provide connectivity. 2x 400G-DR4 will provide the industry’s highest density port for 100GE aggregation.
In the meantime, I believe that 800GE optical interfaces will take time for the industry to develop. Looking back at 400GE, the IEEE started developing those standards in 2013 with the 400GE working group. Today, seven years later, we see modules and platforms finally being deployed.
With 800G speed demands on the near horizon, Cisco optics solutions help our network operator customers keep up with today’s needs and beyond.