I recently hosted a panel discussion with two industry experts who have been actively contributing to the architecture and evolution of telco clouds since their inception over a decade ago. It was an enlightening discussion, one that clarified for me why 5G and cloud-native architectures represent such a critical inflection point in that evolution. It also underscored what a tremendous opportunity these rapidly evolving architectures represent for Communication Service Providers (CSPs) as they look to transform their businesses in the 5G era. These transformations are already well underway. According to the GSMA, mobile operators will invest $620 billion in their networks between 2022 and 2025, of which 85% will be on 5G.

Joining me on the panel were Paul (PJ) Parker-Johnson, the Chief Analyst & Practice Lead for Cloud & Virtual System Infrastructures at ACG Research, and Bruce Wallis, who is the Senior Director of Product Management - Data Center Switching and Routing at Nokia. PJ recently published a research white paper on the topic of telco cloud architecture called “Creating Agile, Intent-Based Service Delivery Fabrics in CSPs’ Distributed Clouds.” Bruce has been involved in building Nokia’s Data Center Fabric solution since its inception and played a key role in its 2020 launch into the market.

During our discussion, we talked about seven definitive building blocks that represent distinct strands in the DNA of the next generation of telco clouds. When implemented correctly, these building blocks provide the foundation on which CSPs can build their 5G networks and accelerate the transformation of their businesses. I’ve summarized our conversation and included links to video excerpts from the panel that provide additional insights into each of the building blocks.

1. Cloud-native architectures

A cloud-native architecture is the first telco cloud building block. The intersection of 5G and cloud-native architectures provides an automated, software-driven, resilient, flexible, and open framework for the delivery of the next generation of consumer and business mobile services. As well as implementing this framework, CSPs, along with enterprises and cloud providers, are building out a massive network of distributed edge clouds, which also benefits from the cloud-native approach.

“If you look at CSPs moving forward and the extent to which they have a chance to become a part of that value chain to play a meaningful role in the delivery and consumption of those services,” PJ says, “it just opens the door to take part in that transition to a new architectural framework.”

To learn more about the potential of cloud-native architectures, watch an excerpt from our discussion here.
 

2. Open data models, interfaces and standardized protocols

Standardization is another fundamental building block, one that is key to openness and integration. Today’s data center fabric technology has come a long way when it comes to interface and protocol standardization. For example, Google’s general purpose Remote Procedure Call (gRPC) protocol and the management interfaces that have been built on top of it, such as the Google Network Management Interface (gNMI) and the Google Network Operations Interface (gNOI), are being more widely adopted within the industry.

With the standardization and broad adoption of protocols like these, networks and their stacks can be more easily decomposed and integrated across different vendor technology solutions, giving customers more choice and greater value. Some other areas that we discussed which rely on openness and standardization include horizontal and elastic scalability and extensibility.

 “I think the main theme here,” Bruce says, “is that the standardization of the interfaces themselves has never been at the point it's at right now, historically.”

Watch an excerpt from our discussion and learn more about interface and protocol standardization here.
 

3. Open boundaries between the network and the applications

The third building block we discussed on the panel was the merging of the application world with the networking world through automation. Traditionally, networks have been manually stitched together well before the “application” compute stack was built. This worked reasonably well, but with the speed of change within the application world today, this is no longer sustainable.

This creates a problem that can only be solved with deep integration into the cloud management platforms by a network fabric automation solution, one that effectively merges the compute and networking worlds. Up to now, these worlds have been separated within siloes.

As Bruce says, “so you can imagine in this world where you have hundreds, even thousands, of locations, and hundreds of thousands of network functions, or microservices, that all need to be connected and dynamically adjusted across your network. Having a network adapt to this scale of connectivity without automation will be very difficult.”

Watch an excerpt from our discussion on open boundaries between the network and the applications here.
 

4. NetOps automation adoption

NetOps is another key building block we discussed. According to PJ, “NetOps is where its owner thinks of the network as a software-driven entity that functions automatically in line with its operator's intent. Software functions built into a network fabric's management system automatically link the results of preceding steps (like network design) with 'downstream steps' like configuration and network monitoring via built-in logic and controls... much as the DevOps process in application delivery works”

The industry has made various technological leaps to support NetOps and there is broad recognition of its importance. Examples of NetOps include intent-based automation approaches used to dramatically simplify the design (Day 0), the deployment (Day 1), and the operations (Day 2+) of the data center fabric. However, the adoption of NetOps capabilities by CSPs stands at only about 15-20% so far. This must increase to meet the emerging demands of 5G.

Learn more about the NetOps building block by watching an excerpt from our discussion here.
 

5. A digital twin for the network

Implementing a digital twin for the network is ideally embedded into the automation platforms workflows enabling CSPs to test and validate network changes before they are implemented in the network, which makes it another highly valuable building block for the telco cloud. If the digital twin is designed as a virtualized infrastructure and implemented properly, this can save time, money and power, and dramatically reduce network and operational risk over traditional lab-based approaches.

However, there are also some challenges that need to be overcome for this approach to be effective. First, the implementation must not only model the configuration of the fabric, but also the state of the network. Second, the implementation must be built into the workflows of the platform that is hosting it.

“One thing that is very hard to model is the environment, the world in which that network element lives," says Bruce.

“We often refer to this as state, like what routes have been received from neighbors, which interfaces are actually operationally up. And if you can't model these accurately, then you can't really do proper CI/CD for networking.”

Watch an excerpt from our discussion on this topic here.
 

6. Next-generation telemetry

Another important building block for success with telco clouds moving forward is streaming telemetry. It represents a dramatic change from traditional poll-based approaches, which periodically request data from the network in bulk. With streaming telemetry, and in particular “on change” telemetry, asynchronous updates are sent to the operator only as state changes, which is not only much more scalable, it’s also more accurate.

However, streaming telemetry can be hampered by the granularity and accessibility of network data, which is often limited by the network operating system (NOS). A model-driven NOS allows for granular and ubiquitous access to data and acts to further accentuate the value of streaming telemetry. The natural extension of streaming telemetry is within closed-loop automation to feed the analytics platform to help detect anomalies and remediate problems.

When discussing the importance of streaming telemetry in this context, PJ says, “one of the things that I think is really important to take away from the conversation is that this is not yesterday's mode of operation. This is a mode of operation that's for now moving forward and it's different and it’s better.”

Watch an excerpt from our discussion on next-generation telemetry here.
 

7. Edge data centers

The final building block the panel identified is CSPs need to build out edge data centers in close architectural alignment with the design of their overall telco cloud. Data centers are now being distributed across logical hierarchies to capture more regional economies of scale and to move closer to the entity they are serving in an effort to improve performance. The smallest sites can hardly be called data centers at all since the compute requirements are very small and the sites typically only house a few servers. These small sites have unique requirements since they are both cost- and space-constrained yet could be represented by thousands of locations that still need to be managed and automated. The panel discussed how these requirements are addressed, along with the importance of enabling a high degree of local autonomy.

Discussing these requirements, Bruce says “I can see there's kind of more momentum behind standardizing some of the automation for build out, what we typically call day zero, and bootstrapping the fabric, reimagining devices, getting them addressed, interconnected, bringing up the control plane, all those good things that you have to do.”

Watch an excerpt from our discussion on the subject of edge data centers here.
 

Conclusion

The discussion was particularly interesting for me considering the launch of Nokia Adaptive Cloud Networking earlier this year. It was refreshing to hear some of our views from that launch reinforced by the members of the panel.

Clearly CSPs have a lot of work to do as they embark on the once-in-a-generation technological shift that is 5G. However, I strongly believe that with the right technology foundations and the right committed and experienced partners, CSPs have the ability to thrive in this new era and be at the forefront of service innovation as 5G services become commercialized and exciting new mobile services are delivered to the masses.