Automation requires a paradigm shift for service provider network deployments. And it’s more than you think.
Everyone in the industry knows the benefits of 5G network automation. Efficiency skyrockets, the network is less expensive to maintain and resizes itself to fit subscriber demand. The CSP can diversify its business, with a wide variety of new services that are quick to launch. Network reliability increases with a network that looks after itself.
We all also know just how important this topic is. 5G, which is much more than just a radio upgrade, brings with it new network capabilities, but more network complexity as well. For example, Ultra Reliable Low Latency Communications drive network functions to edge clouds, and network slicing creates multiple layers of virtual networks. The plain fact is that without advanced automation, the 5G potential cannot be fully realized.
Past successes are leading to failure
Today’s CSP networks are a result of 100 years of technology advancement, and have been extremely successful in delivering services we use every day. They are reliable and have been optimized for operational costs over many decades. The architecture itself (IMS) is designed to be flexible for delivering IP based communication services.
All of this was achieved by encapsulating functions needed in the network into precisely engineered boxes, connected together with standardized data interfaces. Element Management Systems (EMSs) were added to manage these, followed by automation through scripting of common functions. In addition, network management systems layered atop this structure have their own scripts that apply changes across the network. Services are faster to provision, costs are reduced, and profit is maximized.
Despite all this past success, CSPs have hit a wall. They see the highly effective closed loop automation that webscale businesses like Amazon, Google and Microsoft have created, and, in comparison, the automation in CSP networks seems rudimentary. Despite transitioning to cloud infrastructure, automation benefits that webscale companies enjoy continue to be elusive.
The root cause of this failure is the very approach that provided so much success in the past. Although CSPs are moving to the cloud, they continue to view the infrastructure as a collection of network elements controlled by EMSs. With ETSI’s NFV MANO, the network element paradigm is retained by repackaging network elements as VNFs. The newest approach is to use Cloud-native Network Functions (CNFs), composed of microservices. But this is still not sufficient to solve the automation puzzle for CSPs.
The real paradigm shift
Webscale automation benefits can be realized only by fully adopting webscale practices. It is not enough to create cloudified versions of telecom network elements – whether as virtual machines, VNFs, or even CNFs. The telecom application must be designed from scratch for the webscale cloud environment.
The telecom industry designed the network functions first, then added management elements, and finally layered automation on top of that. This is a natural approach for designing a purpose-built appliance. In contrast, webscale cloud design starts with the compute infrastructure, including operations and automation, as a given. Applications run in a pre-existing standardized, automated environment. The operations processes exist before the application is even created, not layered onto the application after the fact.
The paradigm shift isn’t just that we in the telecom industry need to adopt a cloud-native approach to network element design. Rather it is that we need to adopt an approach that is essentially incompatible with the one we have today. Instead of trying to “convert” network elements to run in cloud, they must be rearchitected altogether to operate natively within a webscale environment with finely granular decomposition into microservices. The 3GPP 5G core network specifications in fact promote this idea.
Operators like Rakuten and DISH (both Nokia customers) have realized this, and launched ambitious projects that build their networks from scratch in this fashion. They have successfully accomplished the paradigm shift.
What’s next?
Moving to a fully cloud-native infrastructure that fully benefits from webscale approaches to automation is going to take years to accomplish. It is the next big technology change that faces the telecom world.
Once mature, both network functions and EMSs will disappear – deconstructed into finely granular microservices all managed by Kubernetes orchestration, all taking advantage of open source plugins, with every telecom vendor using the same set of webscale tooling. Core network functions will live alongside business logic applications, or industrial automation applications. CI/CD update automation, and automated configuration will be built in to the infrastructure.
There will be a minority of CSPs that prefer to buy the whole network stack from a single vendor, but we will also see carriers with no owned infrastructure at all, instead using telecom Network-as-a-Service assets purchased from other CSPs or from telecom network vendors. But all vendors, no matter their approach, stand to gain business benefits at a dramatic rate due to the unleashing of automation using webscale techniques.
If you’d like to know more about what Nokia is doing in this space, visit our Nokia Core Automation page.
