Fig. 2: NSA deployment options 3/3a/3x
While the differences between the various variations of option-3 are subtle, we feel option-3x is the best bet for operators looking to support 5G services alongside their existing 4G offerings. Here, the voice services (VoLTE) continue to be with 4G radio, whereas the 5G data is over 5G radio connected directly with EPC. The signalling from the UE towards EPC is over 4G radio.
3. Enhancements to EPC to support 5G
First, the most pressing need in EPC to support 5G PDU sessions is to enhance the user data path to support high-speed UE sessions. This is achieved through high-speed packet-handling technologies like FD.io/VPP using DPDK.
Second, 4G-CUPS allows independent scaling of user plane and data plane in EPC. While this is not essential for 5G deployment, it helps in future migrations in a gradual manner. That is, upgrade the data plane (UPF) followed by the control plane or vice versa.
Third, the charging function in 5G (ChF) supports both online and offline charging over the same interface. To support this, enhancements are required to mimic the ChF behavior over Gy (P-GE interface towards Online Charging Server).
Finally, to make sure that the heavy user data load induced by the introduction of 5G users does not impact the quality of experience of the existing 4G users, there must be a mechanism for resource monitoring and selective throttling of data rates.
4. Deployment and Testing considerations
Typical EPC deployments are appliance based – an integrated control plane and data plane running on custom hardware. The architectural enhancements in 5G provide an option to have cloud-native deployment, with standard hardware and cloud-orchestration tools on which network functions may be realized. In other words, the deployment model transitions from a collection of custom appliances to a collection of applications (Network Functions) in the operators’ data centre cloud.
While the operators would like to continue the existing EPC as is, to support the existing user base, the capacity expansion is also thought of as a means to migrate to a new 5G core. Hence, some of the architectural artefacts of the 5G core are introduced into EPC. The first of these is the introduction of Control Plane and User Plane Separation (CUPS) into EPC. The EPC user plane (SGW-U + PGW-U), after separation from the control plane is functionally equivalent of 5G UPF. The EPC control plane blocks (MME, SGW-C, and PGW-C) can be realized as virtual appliances. These can be running in operators’ data centres with standard hardware and cloud orchestration tools. The same data centre may eventually migrate to run 5G core network functions.
The network functions may be independently tested as standalone, by having tester functions to exercise the services provided by the network function. This simplifies automation. A set of automated scripts may be executed against network functions for their acceptance prior to testing the collection of network functions for an end-to-end deployment test.
The improved speeds/ bandwidth with the new 5G RAT allows services like eMBB. NSA option 3x provides an immediate opportunity for a network operator to support some of the 5G use cases like eMBB that can be monetized, while at the same time, supporting the existing 4G deployment. This improves the time to market and provides a longer return on the investments in EPC. This also provides the much-needed time for networking vendors to implement a fully cloud-native 5G core from the ground up. While the eventual goal is to get to a stand-alone 5G core deployment, the EPC is here to stay in many NSA options in the years to come.