As 400G optical transceivers have been put into use, data center connectivities are moving to 800G Ethernet gradually. Recently, the OFC 2020 conference took place in San Diego focused on the 800G and more enabling technologies for 800G links as well as future fiber optics, making the 800G Ethernet become another hotspot after the 400G networks. However, 800G optics is on the stage of standards perfection and testing, there is still a long way to go for its deployment on a large scale.
Compared with 200G and 400G Ethernet, 800G is relatively brand new. But as the need for higher bandwidth networking equipment and connections in cloud expansion and hyperscale data centers, 800G optical modules and transmissions are the inevitable trends in the next three to five years. Currently, there are several workgroups that have put forward 800G Ethernet specifications or hardware specifications. Let’s run down the main important ones and get the points of them.
Till now, the IEEE organization hasn't worked on the 800G Ethernet standard. But the 25 Gigabit Ethernet Consortium, now renamed as Ethernet Technology Consortium (ETC), has announced the 800GBASE-R specification for 800 Gigabit Ethernet (GbE) in April 2020. As the ETC claims, the goal of this 800GBASE-R specification is to repurpose the standard 400GbE logic as much as possible to create an 800 GbE MAC (media access control) and PCS (Physical Coding Sublayer) specification and to reduce the cost for users implementing multi-rate Ethernet ports. It can be regarded as a wider version of 400G Ethernet, and introduces new MAC and PCS. Still based on the existing 106.25G lanes which were pioneered in 400G Ethernet, this 800GBASER specification aims to double the number of total lanes from 4 to 8 in PCS. Though it seems to be a simple change in concept, it will require a great amount of work and tech skills to bond together the additionally added lanes in this fashion.
The 800GBASE-R standard defines the basic layer of 800G networks, and it is looking to double performance once again, to feed ever-hungrier datacenters. Besides the layer, the specification of the 800G optical transceiver module is another key point that needs to be concerned. The 800G Pluggable MSA (Multi sources Agreement) group released an 800G specification targeted PAM4-based aggregate 800 Gps Ethernet transmission for data center applications in September 2019.
800G-SR8: a low-cost 8x100G module for SR applications, covering the sweet spot of 60-100 meters based on SMF solution.
800G-FR4: new FEC is needed for this 4x200G module application.
Besides, this specification also put the possible pluggable optical modules, potentially in QSFP112-DD and OSFP 32 form factors, as the interface mechanism, and believes that pluggable form still is the ideal interface for data center operators to require 800G.
Another transceiver MSA group consisting of Cisco, Broadcom, Juniper, Intel, etc. also think the pluggable form factor is an ideal choice for data center connectivity. This group recently brought the first release version of 800G QSFP Double Density (QSFP-DD800) pluggable optical transceiver connector and case system standards, defining the 800G optical transceiver form factor - QSFP-DD800. It's said that the QSFP-DD800 supports 8 high-speed electrical interfaces connecting to the host, each lane is 100G/s. And it's able to be compatible with the previous QSFP-DD or QSFP modules including QSFP+, QSFP28, QSFP56, and 400G QSFP-DD, providing network operators tremendous commercial and operational advantages during the 800G network deployment.
Though the necessary specifications of 800G Ethernet and optical modules are relatively completed, it's likely that they need to be revised or perfected later in accordance with practical applications. And this situation won't affect the increasing development of fiber optic modules in high-speed data center connectivity. The availability of high-volume, low-cost optics will remain the enabler of all speed transitions. As the network speed increases beyond 800 Gbps, the pluggable optics will hit density and power issues. When this occurs, alternative technologies like co-packaged optics (CPO) may be needed.
But currently, the 800G pluggable optical modules won't be produced and put into use quickly for "client-side" applications due to the complexity of the structure. The needs to move to a higher speed while keeping the high-performance stress great challenges for optics components. The 800G Pluggable MSA group claims that the several subcomponents targeted in the MSA are already being prototyped and the first 800G modules expected to sample in 2021. And recently the Inphi announces a new 800G 7nm PAM4 DSP (digital signal processor) that is the first 800Gbps or 8x 100Gbps PAM4 DSP to enable 800G optical transceiver modules in QSFP-DD800 or OSFP form factors, making possible for the production of 800G pluggable transceivers.
However, for the "line-side" 800G networks, the 800G optics are not strange for brand network operators. And its main function has been achieved on "line cards" which are complex electronic circuit boards, on which optics manufacturers can develop their own network solution with no compatibility or interoperability with other manufacturer’s line cards. For example, Ciena WaveLogic 5 Extreme (WL5e) delivers 800G of capacity over a single wavelength with tunable capacity from 200G. And in DCI (data center interconnect) applications, it's able to support 600Gb/s across 1,000km links, 400Gb/s across ultra-long-haul routes, and 200Gb/s across Trans-Pacific compensated cables. And Huawei also launches its 800G tunable ultra-high-speed optical module, claiming to guarantee bandwidth in first-rate 5G networks, and supports line rates tuned from 200 Gbit/s to 800 Gbit/s to adaptable to various applications.
But no matter the WL5e or Huawei's 800G optical modules, both of them need to be used together with their portfolios if users need to apply the 800G speed for backbone transmission, metro transmission, and data center interconnections. And the cost will definitely be higher than that of the current 100G, 200G, and 400G networks.
The growing traffic and bandwidth in cloud explosion and data centers not only enhance the growth of 800G optical transceivers but also put forward higher requirements for switching capacity of network switches and routers. Therefore, the choice made in the 800G specifications will take into account both the transceiver form factors and the technology on the switch/router side. Considering the 8x100G and 4x200G solutions and the QSFP-DD800 specification also come out, it's sure that the 800G optics will also be launched quickly to keep up with the current high-speed applications.