In 5G, the control and user (data) planes are split (Control & User Plane Separation – CUPS) and deployed in a new Service Based Architecture (SBA). This provides the flexibility to deliver user plane functionality at the edge as well as the network core. And the UPF can be co-located with local and central data centers at both locations.
This enables multi-access edge computing (MEC), which delivers resources at the edge to support new low-latency, ultra-reliable, and mass-volume 5G applications. It also enables the application-specific, end-to-end assembly of connectivity and computing resources across a shared, multi-tenant 5G infrastructure (network slicing). In other words, the UPF is the prime enabler of new 5G applications and customer-specific edge services.
The 5G UPF is responsible for:
It must deliver these services under extreme throughput conditions, and it has to be 100% stable to meet ultra-reliable connectivity requirements. It also has to be as close to 100% accurate in application identification as possible for effective network slicing and orchestration of network functions.
However, application classification is a highly complex and resource-intensive process. It has become even more complicated with the increased use of encryption. In addition, the protocol signatures used in application classification must be continuously updated, making it difficult to maintain accuracy in application awareness for the UPF.
A logical solution is to leverage 5G’s service-based architecture to offload application classification from the UPF. However, if application identification is offloaded from the UPF controller, application ID (AppID) propagation is difficult.
Enea overcomes this obstacle with the Qosmos ixEngine®, a traffic analytics and Deep Packet Inspection (DPI) engine. Application classification and AppID generation can be offloaded to Qosmos ixEngine, which is deployed as a containerized (or cloud-native) network function (CNF), fitting seamlessly into a cloud-native 5G SBA. This solution protects UPF stability and performance while delivering application identification with a level of accuracy and throughput that can only be provided by a commercial-grade DPI engine.