Disaggregation and breaking up of monolithic systems are supposed to bring flexibility, deliver access to inexpensive products and prevent vendor lock-in. This is why network disaggregation is high on the agenda of cloud and network service providers and why openness as the key catalyst is gaining a lot of interest. 

Open interface communities, open protocol initiatives and open source communities are springing up like mushrooms. The trend towards open networking has now also reached optical transmission systems, which consist of transponder shelves interconnected through agile open optical line systems (OLS).  Those line systems are built with optical amplifiers, reconfigurable optical add-drop multiplexers (ROADMs), and optical cross-connects – frequently in combination with an optical service channel. 

Recently, ONF suggested establishing an ecosystem with partners interested in fully disaggregated transport networks, which would handle each element of an optical network as a separate device. This community would develop reference architectures and provide guidance on what level of disaggregation is most suitable for specific use cases.  

It seems to be a good time to look at the benefits of disaggregation, discuss pros and cons, and suggest some direction on how best to separate previously monolithic networks into technology domains.  

Let’s Look Back

Disaggregation as a means to reduce vendor lock-in isn’t a new concept. SDH and OTN are well-standardized transport network technologies and interfaces are highly interoperable. In real networks, single-vendor domains are applied regionally. Real multi-vendor networks create operational complexity, as not all management functions and interfaces are fully standardized. 

OIF and ONF have developed interfaces for separation of the control plane from the data plane. There’s already wide-spread deployment of white box switches in data centers, managed and controlled with open network operating systems and open central controllers. The same principles have been demonstrated with optical transport networks in various showcases. Demos have shown that this technology can be applied in multi-vendor, multi-layer networks, however, up to now, there’s no widespread network deployment.

In summary, disaggregation in public communication networks, despite being theoretically possible, is not widely applied today as not all interfaces for the efficient management and control of large networks are sufficiently standardized.  Operational integration is still a troublesome task.

Levels of Disaggregation in Optical Transmission Systems

As outlined above, the vertical disaggregation of optical transmission systems by separating the data plane from that control plane is already subject to standardization. There are different options for open, programmatic control by means of OpenFlow or Yang-Model based T-API from ONF and TEAS defined at IETF, respectively. Those control interfaces are now complemented by open management interfaces, removing most of the network deployment barriers. 

As the management and control interfaces become standardized, the focus moves to horizontal disaggregation of optical transmission systems. A transponder-less, open optical line system in combination with colored interfaces plugged into switches and routers is perceived to be a commercially very attractive solution. QSFP28 based 100Gbit/s direct detect PAM4 modules can be successfully combined with simple line systems consisting of an intelligent optical amplifier and a monitoring unit. This solution bridges distances of up to 80kms, sufficient for a large number of DCI applications. As most routers and switches support QSFP28 modules, this solution is very cost efficient and already widely deployed today. It’s arguably the most successful open line system so far.

Large optical networks consist of an optical layer with optical amplifiers, optical cross-connects, optical ROADMs, and optical supervisory channels. Today, those solutions are monolithic and provided from a single vendor. A full disaggregation down to the element level of the optical layer as suggested by ONF would require interoperability of the physical interfaces (data plane), as well as the control and management plane. In this context, the work of OpenConfig and OpenROADM is relevant in addition to the above listed open communities. 

So, what needs to be considered on our way to the next level of network efficiency? Let’s have a look at this question in some more details.

What Not to Forget When Disaggregating Your Network

Optical transport is an analogue transmission technology, which needs sophisticated control of power levels and modulation schemes of any single channel to achieve maximum bandwidth efficiency of a fiber link. Fast restoration switching in a large meshed network is no easy task, as optical power levels suffer from transient behavior. A well-designed system architecture is required to add and remove channels without impacting live channels. 

Monolithic optical transmission systems are optimized for highest bandwidth utilization, lowest power consumption and optimum scalability. Disaggregation of such a system requires a set of least-common-denominator interface specifications to be agreed on in order to meet the interests of a wide community of service providers and vendors. In consequence, disaggregated optical systems won’t provide the same level of performance, flexibility and scalability as single-vendor solutions. 

A disaggregated optical line system makes it possible to eliminate transponders by plugging colored interfaces into switches and routers. A good example of this is outlined above. It is, however, limited to distances up to 80km. Could a similar approach also work with large optical networks? 

Suppliers of optical transceivers are in a constant competition for highest bandwidth and smallest footprint. Currently available optical integration technologies don’t allow suppliers to manufacture highest-performance interfaces in packages that can be plugged into installed switches and routers. Today, coherent interfaces with integrated DSP aren’t available in QSFP28 packages. The most compact design for 100Gbit/s is the CFP2 module. However, modules with a larger form factor already achieve 400Gbit/s or even 600Gbit/s. Hence, it will take some time and additional development effort before pluggable colored interfaces in routers and switches will meet the performance requirements of networks that constantly need highest bandwidth.

So, What’s the Best Way to Disaggregate an Optical Transmission System?

A transponder-less, open optical line system in combination with colored interfaces plugged into switches and routers is a very attractive solution. This makes the business case for high-bitrate PAM4-based optical transmission in combination with smart optical amplifiers for distances of up to 80km. It creates an easy network that’s simple to design and implement, especially if automated dispersion compensation is applied.

For medium and large distances and very high bitrates, compact colored interfaces are not available yet. Disaggregated or integrated transponder shelves offer a cost advantage as space requirements of colored interfaces integrated into switches and routers would waste fabric capacity. Open control and management interfaces are a prerequisite of leveraging this advantage.

Full disaggregated optical transmission systems aren’t considered to be a short-term opportunity for highly efficient optical layers. However, there’s significant value in harmonizing control and management interfaces. This enables service providers to simplify integration and operation in multi-layer networks and multi-vendor network domains, while still leaving opportunities for vendor-specific optimization.