Get best of the week

400g. View from the side of Transmission. ZR / ZR +

image

Want to find out what's new in transmission life in the context of 400Gbit modules? I ask for cat.

Introduction


2020 continues in the trend of "400G to replace 100G interfaces." All major manufacturers report their readiness to ship equipment with support for such interfaces. I want to note that this request now comes mainly from the Hyperscale business. But it turned out that all manufacturers now focus on the Hyperscale business, because there the main money. And such trends will undoubtedly affect everyone. Actually, what are manufacturers ready to offer in terms of optical modules?

QSFP-DD


New Form Factor Module - QSFP-DD (Quad Small Form-factor Double Density)

image

If you are confused with these all SFPs, then roughly, the situation is approximately the following:
  1. SFP - 1Gbit
  2. SFP + - 10Gbit
  3. QSFP + - 40Gbit
  4. QSFP28 - 100Gbit
  5. QSFP56 - 200Gbit
  6. QSFP-DD - 400Gbit


The development of modules is carried out by www.qsfp-dd.com. This is an interface that will gradually replace the QSFP28 (QSFP56 has not become very popular and apparently will not) interfaces on the equipment of all manufacturers, while remaining backward compatible. These modules will be actively used at the level of internal switching of the data center and operator joints, but their small size will not allow them to execute any solution based on a distance of more than 100 km.

OSFP


From the point of view of transmission, we are interested in the OSFP (Octal Small Form Factor Pluggable) stepbrother. Osfpmsa.org/index.html is

image
being developed .

How is it that we get two standard optical modules for 400G ports? Everything is quite simple. In addition to internal high-capacity interconnections between switches, routers and other equipment that is relatively close to each other (from several meters to several kilometers, they are usually called access or intra-DCI), there are also tasks called inter-DCI ( Data Center Interconnect). This is when we need to transfer a large amount of data between data centers at a distance of more than 100km. Or we want to build a Metro network.

DCI
, L2/L3, , , : - , 100.

Some manufacturers (the pioneer and main pusher here Arista) thought about this need and decided to continue to use the positive experience of CFP2 coherent modules (CFP2-ACO and CFP2-DCO) taking into account new trends and the possibility of compatibility with super-popular QSFP. You can read about coherent CFP2 modules here .

But QSFP-DDs do not allow to dissipate as much heat as the coherent optical module generates. Therefore, the OSFP format appeared which is distinguished by the following points:

  • It is slightly larger than the QSFP-DD module, so QSFP-DD is compatible with OSFP through
    adapter
    image
  • It allows you to dissipate about 15W of heat, unlike QSFP-DD due to an external radiator

In sum, such distinctive features make this module a good solution for use in DCI systems. And here there are two types of interfaces that interest us the most:

  • OSFP- ZR
  • OSFP- ZR +

What is their main difference?

Zr


OSFP-ZR was developed by OIF as originally an Interop solution for running DWDM single carrieer modules of various manufacturers in DCI systems (most likely as an alternative to PAM4 100G modules). Its operating range is limited to 120km, which makes it applicable in a somewhat narrow range of solutions, but at the same time they should be very attractive in cost. The range limitation is related to the FEC used there.

ZR +


At a certain point, colleagues from openroadm.org joined in the study and proposed using oFEC (more about it here ), which they had previously developed in close collaboration with ACACIA. The proposed solution will guarantee operation at distances from 500 km and potentially up to 2000 km. Of course, I mean without regeneration, amplification will still be required. This format was called ZR + and it will be of interest to operators and companies that have urban and regional DWDM networks.

Summary table of differences.

image

Conclusion


In general, this news promises to very strongly modify the telecommunications industry in the context of the fact that:

  • More hardware is appearing on merchant silicon with more functionality and lower cost.
  • Hyperscale ( ).
  • optical reach 1000 90% , , :

image

The struggle of formats has not yet ended, and in many ways this year after OFC it will become better to understand where the industry will move (if it is not canceled as MWC ^ _ ^). My personal opinion is that OSPF has good potential and has a future, backed up by an already defined manufacturers alliance led by Arista. In addition, demointerop between different manufacturers is promised this year. There are still not such popular CFP4 / CFP8 for these purposes, but I have a feeling that the CFP format will slowly lose popularity and CFP2 has become a peak of popularity for them.

PS Perhaps my previous articles will help you dive a little deeper into the fascinating world of DWDM:


PPS I specifically did not go into the details of 16QAM modulation and other features of receiving a 400G signal, if it is interesting, write in the comments. There was no revolution here, only evolution.

UPD: spelling corrections.

More articles:


All Articles