Unlocking the full benefits of cloud-native computing and containers for telcos

[ServicesExperts]

By Tomotake Koike, Senior Chief Solution Architect, Cloud Native Computing Practice, HPE Advisory & Professional Services

Cloud native computing changes enterprise architecture completely. It provides a new level of efficiency and innovation for digital transformation, and it applies to many industries.

The telecommunication industry is no exception. How do cloud native computing technologies, especially Kubernetes, benefit the telecommunication business? The answer lies in the rapid evolution of container orchestration technology and innovations in the cloud native computing community. (Read: What are containers?)

The challenges of Kubernetes for telcos

Originally, it was the common understanding that Kubernetes was not a good fit for large, complex data transfers because of its basic characteristics. The ‘upstream’ development project in the Kubernetes community focuses on simplicity as its platform concept. It is the same for Kubernetes network architecture, which gives a single network path, and data transmit goes through container network interface (CNI). Due to the limitations of aligning to these basic concepts, telecommunication organizations felt that Kubernetes did not match the requirements of their network functions, since the user plane (U-Plane) requires massive throughput.

The basic concept of Kubernetes architecture operations is simplified by limiting its development to the minimum required functions, so that the network can be easily managed and controlled (although the actual iptables control, etc., is very complicated). In this way, with the Kubernetes cluster in the upstream vanilla state (no plugin etc. installed), it can be maintained with the CNI. There is only one network, and in order to hide the topology inside the cluster, communication is basically done with network address translation (NAT) to outside. However, even though it is control-plane (C-Plane), this becomes a big obstacle in network functions that handle user-plane (U-Plane) packets that require very large throughput and transaction processing system (TPS).  Orchestration of these types of cloud native network functions (CNF) has been very challenging.

You may be asking: But what about virtualized network functions (VNF)?  Even with VNF, in a complete virtual environment, when connecting from the VM Host, you would deploy NAT or a bridge of the virtual switch on the host and connect.  However, this configuration cannot sufficiently handle the high throughput and TPS packet communications required by the U-Plane.

Fortunately, VNF can leverage technologies such as single-root input/output virtualization (SR-IOV) and the data plane development kit (DPDK) that make the resources of physical devices available directly from VMs in order to accelerate throughput.

A breakthrough for the U-Plane

CNI also has multiple plugins available that make the resources of physical network devices directly available from the container (or pod), just like VMs. By using one of these, it becomes possible to meet the difficult communication band requirements of the U-Plane. However, as mentioned above, the Kubernetes cluster can manage only one CNI, and if this is entwined with physical network resources or Kubernetes orchestrators (such as IPAM and Service Discovery linked to pods by Kubernetes), the benefits of the function will not be achieved, and manageability will be impaired.

To overcome this challenge, new functionality has become available in the cloud native computing community: Multus-CNI. This allows you to attach multiple different types of CNI to pods. First, for the network managed by the Kubernetes cluster by default, deploy it with a general CNI (such as Calico etc.) managed by Kubernetes as a full-service. On top of that, for CNF pods that need to handle U-Planes, add a network interface to off-road to physical network devices. By doing this, CNF on Kubernetes enables U-Plane packet processing, filling that gap in Kubernetes for telecommunication service providers.

HPE can help you get the most from cloud-native computing

HPE has a long history of providing IT consultation services to telecommunication service providers to drive their cloud native computing-based network function transformation. What we have learned is that their top priority is the operability of Kubernetes and platforms, without demanding excessive communication performance. As we see more evolution of Kubernetes functionalities and plugins, it is clear that Kubernetes will become the most common platform for telecommunication service providers.

HPE Advisory and Professional Container Adoption Services can help your team to design its adoption roadmap and architect, build, and run your container platforms. Whether it is on-premises or off-premises, we will build in a truly scalable, standard, repeatable, and reliable manner for automated integration of your unique characteristics into your production environment.

To learn more, see our HPE Container Adoption Solution Brief.

Learn more about advisory and professional services from HPE Pointnext Services.

Tomotake Koike has a keen interest in new, comprehensive technologies, especially in cloud-native areas. He joined HPE in 2006 as a solution architect working in delivery for telcos, having previously developed his skills in this area with a network equipment vendor. Tomotake joined his current HPE team in 2018 to work on cloud-native computing.

Services Experts
Hewlett Packard Enterprise

twitter.com/HPE_Pointnext
linkedin.com/showcase/hpe-pointnext-services/
hpe.com/pointnext