By Casimer DeCusatis
Developments in higher data rates and coherent communication technologies were among the most exciting new developments at OFC 2019. The most important market force driving this technology is the growing need to significantly increase bandwidth between cloud data centers within a campus or metropolitan area network. Such networks require simple topologies (often point-to-point) over distances of 80-100 km (long enough that wavelength multiplexing over single-mode fiber must be employed). To address this need and provide interoperability in coherent dense wavelength division multiplexing (DWDM) systems, the Optical Internet Forum (OIF) has been developing a new standard called 400ZR for the past few years. In this blog, we’ll take a look at the current status of 400ZR and some of the recent innovations announced at OFC.
For newcomers to this field, the alphabet soup of terminology used to describe optical transceivers can be confusing. Most optical transceivers are designated by a group of letters and numbers that specify a form factor and a format for data transmission. For example, QSFP represents a quad small form factor pluggable form factor), an early four channel standard for 10 Gbit/second data rates. The notation typically ends with a distance designation such as SR (short reach, often under 1 km), LR (long reach, 10 km), or ER (extended reach). This notation is part of a multi-source agreement (MSA), which insures that two transceivers from different vendors will interoperate with each other. Of course, this approach isn’t perfect; while short reach connections within data centers and on the client side of telecom networks have demonstrated good multi-vendor interoperability, it’s much less common for equipment from different vendors to work smoothly together over the telecom line side of the network. The emerging OIF standards attempt to bring the same level of multi-vendor interoperability found on short range links into the environment between data centers.
Variations of the standard agreement
Originally, the scope of the 400ZR standard interoperability agreement (IA) targeted 400 Gbit/second Ethernet transmission over data center interconnects between 80-100 km using a single wavelength. It was felt that such an approach would reduce the cost per bit. Over time, the OIF agreed to include support for telecom applications, partially to offset the high development cost of digital signal processing (DSP) chips and optical transceivers by broadening the overall market to include the converged edge network between telecom and cloud computing. As discussed at OFC, there are currently three distinct variations of this standard:
- 400ZR will support 400 Gbit/second Ethernet over 100 km data center interconnects using a pluggable QSFP or OSFP form factor.
- 400ZR+ will support variable multi-haul data rates of 100 to 400 Gbit/second for metropolitan area networks (up to around 120 km) using a CFP2 or OSFP form factor, compatible with the OpenROADM approach. .
- 100ZR will be a low cost variant supporting 100 Gbit/second data rates over 80 km distances at the network edge, designed as a compatible upgrade for 10G WDM links as standardized by the IEEE.
These variations are developed from common building blocks) in order to minimize manufacturing costs as much as possible. There are different timelines for component availability in each case. Currently, the market for 400ZR is fairly well understood but there are concerns that only a relatively small number of cloud service providers would take advantage of this approach. At the OFC 2019 Data Center Summit, both Google and Microsoft discussed their intent to begin deployment of 400ZR in the late 2020 timeframe. Google intends to use 400ZR pluggables for all applications within the designated distance reach. Microsoft’s Azure cloud intends to replace their pioneering ColorZ architecture with a 400ZR equivalent in the same timeframe. Verizon intends to use 400ZR to interconnect regional business hubs with selected peering points in large cities, where the benefit of multi-vendor interoperability combined with higher data rates and DWDM will enable significant bandwidth growth. However, for most other metro applications, Verizon is targeting 400ZR+ coherent solutions capable of multi-haul architectures.
The market for 400ZR+
The market for 400ZR+ is not as clear, but it has the potential to disrupt numerous existing coherent optical network technologies and allow network operators to more easily deploy coherent WDM as plug-in modules for switches and routers. Microsoft also indicated support for 400ZR+ as an upgrade path for its current deployment of CFP2-ACO devices, although this depends strongly on the cost/distance tradeoffs relative to non-pluggable alternatives for 600 – 800 Gbit/second solutions. Overall, about a dozen companies at OFC discussed their plans to invest roughly $500 M to bring variations of 400ZR to market, although only one component company, Neophotonics, has announced coherent optoelectronic components with the 50 GHz bandwidth required to enable future, higher bandwidth links. Several equipment vendors including Ciena, Infinera, Acacia, and Inphi are planning a fully vertically integrated strategy to cover the broadest possible market reach. The OIF has begun hosting education master classes on current and future plans for 400ZR and its variants.
As component developers repurpose their designs to address adjacent markets, more and more companies are beginning to endorse some form of 400ZR in their future roadmaps. Proposed solutions extending data rates to 800 Gbit/second (enabled by current DSPs from Neophotonics) should continue to lower the cost per bit, although they do not address related issues such as a lack of installed fiber in metro areas and spectral efficiency concerns with high data rate coherent communication links. Overall, the emergence of standardized pluggable form factor coherent technology should have a major impact on both cloud and telecom networks, and OFC is a great place to stay informed about all the latest developments in this area.
Does your organization plan to deploy some version of 400ZR, if so when? Drop me a line on Twitter (@Dr_Casimer), and maybe we’ll discuss it in a future version of this blog.
Posted: 21 August 2019 by
Casimer DeCusatis
| with 0 comments