What Is 400ZR and Why Is It Important?
neophotonics.com
By Ferris Lipscomb, Ph.D. on November 3, 2017 | Leave a Comment
What is 400G ZR? 400G ZR is a simple and low cost standard for transmitting 400 gigabit Ethernet over data center interconnection links up to 100 km using DWDM (dense wavelength division multiplexing) and higher order modulation such as 16 QAM. The solution is proposed to be available in a small form factor modules such as OSFP or DD-QSFP.
Why is 400G ZR important? 400G ZR reduces cost and complexity for high bandwidth data center interconnects. Higher order modulation and DWDM unlock high bandwidth, but typically require proprietary coordination at both ends of the fiber. With 400G ZR, you can mix-and-match modules at both ends, something that was previously impossible.
The optics transceiver market is littered with letters and numbers that can be very confusing. Many transceiver designations start with a group of letters that specify a form factor, like CFP, CFP2, QSFP etc, and end with a second group of letters, such as “SR”, “LR”, “ER” etc that specify the way the data is transported. For example CFP2-LR4 means that the physical form factor is CFP2 and the data is carried over the “long reach” distance of 10km on four specific optical wavelengths with specified speeds. These designations arise from the MSA or other standardization specification group and ensure that the transceivers will inter-operate with each other. In other words, any CFP2-LR4 transceiver can communicate with any other CFP2-LR4 over 10 km of fiber, even if the two transceivers come from different manufacturers. This kind of interoperability is common for the short reach connections used inside datacenters or in client side telecom connections where there is only one transceiver pair connected to each fiber.
However such interoperability is much less common over what is called the “line side”, which consists of connections of many kilometers and where many transceivers are connected to a given fiber, each operating on a different DWDM wavelength. Signals can be transmitted for hundreds of kilometers using optical amplifiers. For these applications complex systems have been developed by telecom equipment manufacturers to maximize the number of bits that can be transmitted through a given fiber and to control adding, dropping and switching many channels in the network. Each equipment vendor’s solution is unique and does not operate with any other vendor’s equipment.
Now, however, a new high volume application has emerged that is between these two paradigms. Many datacenters within a campus or a metropolitan area have a growing need to significantly increase bandwidth connections between them. The maximum distance for these data center interconnect (DCI) connections is < 80~100 km. This is a long enough distance that DWDM systems must be used to put many channels on a single fiber. However, the network is relatively simple with mostly point to point connections. And the volumes will be relatively high.
The Optical Internet Forum (OIF) has stepped into this breach and is defining a standard, called 400ZR, which will ensure interoperability in a coherent DWDM system. Originally the OIF 400ZR committee established a scope to transmit 400GE (Ethernet) over data center interconnection links of 80~100 km. By putting 400G on a single wavelength with the same number of optical components (albeit higher bandwidth) that currently transmit 100G, the cost per bit is greatly reduced. However, due to the high development costs of the DSP, optical components, and transceiver modules, the committee agreed to include specifications to support telecom applications, which has the effect of increasing the total market size even further. This means that 400ZR will likely be the work horse approach for both DCI and the converged edge of telecom and cloud connections. The volumes will be very large, and therefore there is a great deal of interest in developing 400ZR products by multiple companies.
It is, of course, not the purpose of this blog post to delineate the specs, and the final OIF implementation agreement has not yet been published. Here we will highlight some of the specifications that impact the interoperation between different vendors. The transmission link in general is composed of a DWDM MUX (40-ch 100GHz-spaced or 64-ch 75GHz-spaced) and a booster EDFA at the transmitting side, and a pre-EDFA and a DWDM DEMUX at the receiver side. It can also be a simple duplex dark fibers without DWDM mux/demux and any EDFAs. While we may casually say that 400ZR operates at 64 Gbaud and 16 QAM, the actual specification must be much more precise. 400ZR actually has a baud rate of 59.84375 Gbaud, which is a result of the combined overhead from FEC, OTN framing, etc. 400ZR also has fixed the modulation format to be dual-polarization 16QAM. Unlike some advanced transponders, there is no option to change baud rates and modulation order on the fly to adjust reach and bandwidth. The transmission link distance should be less than approximately 80km, and the link should provide a link OSNR of about 29dB. A shipped coherent transceiver should meet an OSNR (after transmission) of 26dB, thus leaving a 3dB margin for long-term operation.
The 400ZR implementation agreement does not specify the form factor, but for this application the coherent transceiver is required to be in a small form factor module such as OSFP, or DD-QSFP. These small form factors present a challenge to DSP and component suppliers in terms of both very small component sizes and requirements for low electrical power consumption. Some vendors are even considering the use of the CFP2 form factor in order to accommodate the increased power consumption due to the recently added telecom client-side interfaces.
This presents many challenges to the DSP vendors: low power (e.g., <7W) and small size. It is generally believed that 7nm CMOS node is needed. The power consumption is reduced not only because of the short transmission distance (therefore much less power consumption for chromatic and polarization dispersion compensation), but also much lower sampling rate at the DAC (digital-to-analog converter) and ADC (analog-to-digital converter). In order for the transceivers to be able to interoperate with each other, the DSP must use the same error correction coding and decoding techniques. To this end, the OIF has selected the approach for FEC (forward error correction) that all 400ZR compliant transceivers must use.
For the optical components two miniaturization approaches are considered: (1) COSA (=integrated coherent modulator and receiver)+ stand-alone tunable laser + laser control IC; and (2) all three components (tunable laser + coherent modulator +coherent receiver) integrated in one package. Two material technologies are available: Indium Phosphide (InP) or Silicon Photonics (SiPho), for both approaches. These optical components operate at a 60 Gbaud where InP has proven to be capable, but SiPho still has yet to prove itself.
There are many challenges to making a 400ZR transponder that fits into the small size and power budget of an OSFP or DDQSFP package and also achieving interoperation as well as the necessary cost and volume targets. However, 400ZR may very well be the next big thing in optical communications.
|
