As I wrote in my previous blog “Ethernet is the Backbone of Modern Cloud Native and Real Time Communication” Ethernet has come a long way since the birth back in 1973. It is the most wide-spread networking technology employed. As I talk about in my previous blog Ethernet is being used in mission critical applications such as industry automation, mobile networks 4G and 5G as well as in truly ruggedized environments. I also touched on TSN that enables Ethernet to become a deterministic networking technology where synchronization of network elements and endpoints allow for prioritization of traffic classes and provides accountable delay and guaranteed bandwidth reservation.
You could argue that Ethernet is amongst one of the most important technologies on the earth today without people realizing this. This is due to Ethernet being so ubiquitous as it has been deployed across the cosmos from the bottom of the oceans all the way into the outer space. Ethernet use cases are still evolving and expanding with use cases for higher speeds over longer distance transport networks, with submarine cables traversing the ocean beds connecting the continents as transport systems between continents.
As we all know by now, more and more applications are developed which are truly bandwidth “hungry”, like high definition 8k video, which requires the underlaying transport technologies to provide higher bandwidth. Hence 400G Ethernet is reality today, and 800G Ethernet will be reality in a near future. With that said you could easily expect 1 terabyte Ethernet to be seen by 2030 as bandwidth consumption is not going down.
The raise of machine learning (ML) and Artificial Intelligence (AI) is demanding more bandwidth on the hyperscalers data center infrastructure with servers moving towards 100G Ethernet to be able to cope with this growth in capacity need. There is an interesting aspect around the growth in bandwidth demand of both hyperscalers data centers, and service providers transport networks as they are both adopting similar technologies.
Within the Ethernet Alliance there is a number, of interesting new developments around higher speeds such as 100G and 400G operation over single-mode fiber at 100G per wavelength. This standard is designed to support cost-effective, more power-efficient single-mode fiber interfaces for 100G and 400G Ethernet using 100G optical technology. A couple of other interesting ones are the 100G operation over Dense Wavelength Division Multiplexing (DWDM) systems. This standard is significant as being the first Ethernet specification of coherent DWDM technology supporting 100G connectivity over lengths of at least 80 kilometers, and then there is the 400G operation over DWDM systems, which extends the Ethernet specification for coherent DWDM technology to 400G. This aspect of Ethernet evolution is of interest for the Communication Service Providers (CSP), as they have been driven higher speed Ethernet for decades due to the multi-service aggregation needs, that continue to grow with support for a variety of connections. Particularly the 5G mobile and edge deployment is driving dramatic increases in both fronthaul and backhaul applications, which continues to push Ethernet requirements for higher rates and longer distances. The increase in X-haul connectivity demands and particularly the long-haul connectivity with massive amount of data traversing cross the cosmos forcing data centers and CSPs to find more flexible, reliable, and cost-effective solutions for transport. The answer to this is the coherent optics that brings light to the smaller footprint transceivers, that incorporates the transport of these massive amount of data across high-bandwidth long-haul fiber connections.
For the first time pluggable coherent optics for both outside of the data center as well as inside the data center match up, being the 400G Ethernet with standards such as 400ZR, as it is designed to work in routers without limiting front panel space with pluggable optics like QSFP-DD. This 400ZR is an industry standard, so in theory this should be interoperable between different equipment vendors, and it for sure will and has changed the way network operators design and build their networks.
IP-over-DWDM finally works thank you to 400ZR, this allows for putting Data Center Interconnects (DCI) optics directly into the router, as Digital Coherent Optics (DCO), which eliminates the transponder layer allowing for the optical layers management to reside directly on switches/routers. At the simplest, coherent optics is a technology utilizing low-cost standard optics transmitting 400G Ethernet over long distances using DWDM and higher modulation schemas such as Quadrature Amplitude Modulation (QAM). Use cases for such could be hyperscalers distributed data centers, allowing for usage by “just put in a switch or router with coherent optics”, that allows for less equipment real estate and less power consumption, this in turn will cause many CSPs to re-evaluate their network build out plans.
Why DCO? - as we move into higher speeds the reality is that it is much harder to transmit over fiber due to the “n2 problem”, hence coherent optics will be the best option. With Digital Coherent Optics you will achieve better economics and sustainability, and it provides a new innovative network architecture based on routed optical network design. It is compatible with standard switches and routers and DWDM systems. The Digital Coherent Optics will demand much more data to be manage compared to traditional optics due to all the software and intelligence residing on the optics modules, hence end goal is to be able to automate as much as possible of the needed configuration, and it must be friendly for Machine-to-Machine communication (M2M). The Digital Coherent Optics management and automation should be considered to be aligned with the industry’s push towards open systems and SDN (Software Defined Networks) architectures.
What is next? - 800ZR project is under way by the OIF (Optical Internetworking Forum) standards body and it is expected to also provide a very long-haul 400G option. Coherent optics within data center might not be feasible from a cost point of view, but once at 1.6T (2x800G) it could make sense.
Ethernet is powering infrastructure everywhere across the cosmos, not only powering the new era of Cloud Native 5G data centers to provide infrastructure for the 5G applications, as it has become the ubiquitous network technology. It is a given that 5G will not be the one and only network technology that can be used to fulfill all the requirements, hence the wired and wireless networks will complement each other and higher speeds on the wired networks will be required to provide the transport for the wide plethora of new applications and services and we will see Ethernet once again reinvent itself as the 50 years old technology. The higher speeds will influence the way Ethernet evolves also on a wider perspective around the area of Information Technology (IT) and Operation Technology (OT) networks as there are still gaps in how these “siloed” networks coexist and cooperate. Traditionally these IT and OT networks have not been very well integrated, hence with the Ethernet evolution there is a need for them to to take advantage of things like automation and other common tools as the end goal here is to have a single-protocol (Ethernet) network, that addresses the needs of a wider scope of infrastructures.