Each new fiber optic transmission technology goes through a similar series of stages due to increased standardization. First, when the technology is brand new, it is implemented on a "line card", a complex electronic circuit board on which each manufacturer can come up with its own unique solution, in contrast to other manufacturers of line cards are not compatible or interoperable. Next, the MSA standard transponder module implements the same transmit, receive and control functions, such as "5 x 7", which is actually 5 inches x 7 inches in size.
These repeater modules are hardwired to the line card and have the specified dimensions and electrical I/O, but leave the internal operations to the manufacturer. Therefore, customers can often use MSA standard transponders from different manufacturers on line cards, but must ensure that transponders from the same manufacturer are "paired" on both ends of the same link.
Optical module
In the next stage, optical module functionality is placed in a "pluggable" form factor ("hot" swappable) that can be inserted into a slot in a running transport system, allowing for increased power without powering down the system additional capacity, thereby disrupting traffic on existing links. This is the stage that 400G coherent transmission is now entering, which we will discuss below. The final stage, which usually returns to the form of a line card, comes when the technology is so mature and the cost is reduced that even the cost of the metal cover is reduced.
When "line-side" optical transmission systems enter the pluggable optical module stage, they typically borrow the MSA form factor from shorter-distance "client" optical modules. Generally, "line side" refers to transmission distances of 80 kilometers or more using DWDM so that each optical module can only use one wavelength channel. "Client" refers to shorter distances, from 50m to 10 km, and only connects one optical module to the fiber, so DWDM is not required.
The line side borrows the pluggable form factor from the client, as the client and data center consume a lot of pluggable optics, and new form factors are constantly being defined for higher density interconnects. The ability to move data in and out of switches and routers often limits data center performance, size, and cost, necessitating continued increases in speed and reduction in size of optical transceivers. With advances in switching electronics, designs are often limited to how many optical modules can actually fit on a 1RU standard rack panel.
Optical module
Therefore, due to the sheer number of client optical modules and, equally importantly, the number of deployed sockets for these optical modules, the line side inevitably reuses these client form factors rather than inventing an entirely new Dimensions. This happened at 10G, where tunable DWDM line-side optics were implemented in XFP and SFP+ form factors, and 100G, where CFP-DCO and CFP2-DCO coherent transponders were implemented in client form factors.
Now it is implemented on 400G. 400G transceivers come in three different form factors, from large to small: CFP8, OSFP, and QSFP-DD. (CFP8 is very similar in size to CFP2, but has a different electrical interface and data flow. For our purposes, we'll treat them as the same.) (There is also an approach called "On-Board Optical" or OBO , it's modular at the board level, but not "hot" swappable, beyond the scope of this article.)
First, these are client-side optical modules for relatively short distances within the data center, such as 50 meters to 2 kilometers. Inside the data center, the QSFP-DD is the preferred form factor because it is the smallest and enables higher quantities in a 1RU panel. However, OSFP has some advantages as it can be extended to 800G, and CFP8 has some early special applications. However, when they are repurposed for line-side applications, the situation may be different.
Unlike the 50 meters to 2 kilometers of optical signals that send just one signal on the fiber, an optical signal that may send 80 such signals on the fiber at a single DWDM wavelength is 80 kilometers to 500 kilometers.
The signal processing required for long-distance coherent transmission is much larger than that required for short-distance PAM4 data center transmission, requiring more complex DSPs and more power. Optical components are also much more complex. Therefore, the industry is now evaluating the performance trade-offs necessary to implement coherent DCO modules in each size, and the performance range of each size can be extended to accommodate coherent transmission. Different form factors may be suitable for different applications.
Two main applications are being considered: 1) 400ZR for data center interconnection networks with distances below 100 km;
2) Multi-haul telecommunications for metro and long-distance networks from 400km to 2000km. (For multi-pass applications, the data rate decreases as the coverage increases.) The final answer is yet to be determined, which is beyond the scope of this article anyway. We'll be content to outline the pros and cons of each form factor.
CFP2: The largest in size and will undoubtedly be the first form to be implemented. It is large enough to use current-generating discrete components to implement coherent optics, and most importantly, it can easily handle the power required for multi-pass telecom applications when 7nm process node DSPs become available. The power specification of the client is 24 W, and the power consumption of the coherent version may be higher. Since all optical component technologies exist today, this form factor is likely to enter service first. Since the CFP2 slot cannot provide 400G of space, special line cards will be required.
OSFP: This form factor is about half the size of CFP2, so it consumes less power. However, 400ZR 80 km links and 500 km to 2000 km multi-haul telecommunication links can be achieved using 7 nm node DSPs that will be available starting in late 2018. The small size also requires integrating coherent optics into a single package, called a coherent optical subassembly (COSA), and significantly reducing power. The client spec is 15 W, but coherent modules are expected to support power levels of 18W and beyond, but for multi-channel applications the power handling is very strict. Therefore, it is feasible, but difficult, to implement multi-range telecom and 400ZR in the form of OSFP.
QSFP-DD: Smallest in size, but also expects the most installed client sockets. The power specification for the client is 12 W, but again, significantly higher levels can be supported in coherent optics, possibly 15w or more. The QSFP-DD form factor will support utilizing 7 nm node DSP and ultra-compact and low-power optical components in COSA form factor.
