Dr. Patrik Rokyta, VP of Systems Architecture
In June, I spoke at the virtual TelecomsEurope Digital Transformation event. My remarks focused on deployment automation using different technologies in order to achieve the flexibility and agility across the network services that are offered to the residential or enterprise market. I’ve summarized these remarks below.
But before going into the actual network programming, we must first review the refactoring of legacy network elements to cloud native network functions. Several factors are driving this transition. These legacy telecom products are monolithic and incorporate a lot of functionality, which reduces the efforts during the deployment and integration phase. But when it comes to upgrading these network elements or moving them to a different infrastructure, it is a challenge. Any failure in upgrading or operating these monolithic products can significantly degrade the service they enable.
As networks move to 5G, hyperscalers are playing an increasing and important role that goes beyond the delivery of just the cloud native infrastructure. While Communication Service Providers (CSPs) and their suppliers have not historically competed, we are beginning to see increasing competition between the CSPs and the hyperscalers, especially for the enterprise market.
As a result, CSPs must transform their networks from traditional deployment models to cloud native to bring agility into their communication service offerings—which is already happening across the industry. Microservices architectures using container technologies are making deployment of network elements portable across different types of infrastructures, including virtual, cloud and even physical. The significantly smaller size of the deployment artifacts is enabling procedures like continuous delivery, test and deployment with the possibility to detect and rollback a misbehaving function so easy. But the degree of agility and flexibility CSPs have when delivering services to enterprise or residential customers is also largely dependent on how the operator and its suppliers manage to bring automation and programmability into the deployment of the networks and network services.
Important Role of Automation and Programmability
The availability of a network service is influenced by the local redundancy of the service at one location or point of presence, and at the same time, by its delivery from a geographically distributed location. So we usually see at least two data centers that are able to deliver the communication service in question. The service data (seldom the state) is continuously replicated between the locations delivering the service. The service works perfectly fine usually. But how do you deploy such a network service and also modify it at any time?
Well, we already know how to manage the infrastructure specific aspects. We learned it from the network function virtualization providing for the decoupling of the SW from the underlying HW. But there was not much seen in decomposing the monolithic network elements and therefore the management aspects did not change much. With the introduction of cloud native ecosystems and the microservices architecture, the amount of network elements that you need to manage as part of the network’s service has significantly increased. Frankly, this is nothing you want to touch manually.
Fortunately, the cloud native ecosystem has delivered tools that provide for automated deployment, orchestration, and lifecycle management of cloud native services, thus increasing the agility and flexibility of network service implementations. What we now know from cloud native ecosystems is that we can declare the state of the system and then allow the tools to identify any difference between the current state and the defined state, and automatically adjust the system state as needed.
With the smaller deployment units referred to as the containers, there is much better flexibility in testing functional upgrades or fixes. CSPs and suppliers during the integration process are able to test patches, incremental updates, and functional add-ons in a much safer way and roll out canary deployments within minutes or even seconds. A cloud native environment provides for greater flexibility in testing, upgrading or rolling back new functionality.
Network Service Configuration and Management
A CSP cannot touch every piece of software manually in cloud native ecosystems, so they need sophisticated APIs to manage configuration and deployment. As long as the suppliers adhere to the standards for configuration and deployment management, APIs can provide for interoperability across multiple vendors so the integration cost on the operator side can be significantly optimized and the operational tasks unified.
We recently completed a proof of concept (PoC) on how the provisioning management service defined by 3GPP can be used for controlling the configuration and the deployment of a cloud native network service. The results were really impressive. The PoC demonstrated a successful onboarding, instantiation and lifecycle management including blue/green and canary rollout of the network service. The concept is based on a standardized network resource model that allows the operator to control all aspects of a network service deployment at any level of granularity.
There is a lot of discussion going on in the industry about the configuration and management of the network services. The conclusion seems to be that the task can be addressed using different architectural options: via the deployment server delivered by the supplier, via the OSS provisioning management service, or by utilizing the ETSI MANO framework. An important aspect besides the multi-vendor capability is the re-use of the same deployment artifacts in each of the architectural models. This gives operators a lot of choice to address and satisfy the requirements of the communication service by choosing the optimal way to program the network without asking the suppliers to change their packaging and the deployment process.
To learn more about the key learnings from our recent PoC, send me an email at email@example.com.