10+2 Thoughts on the Future of Core Networks
Marius Corici
Deputy Head of the Software-based Network Division at Fraunhofer FOKUS Institute
A core network should be a collection of "programs", not " functions, " running on common hardware. Why, because it is simpler and because it is more flexible, and mostly because web services did already the job to prove it works. Some other considerations underneath:
?? 1. Core Means Essential, NOT Central?- Here we have a semantics problem:? "Core" does not mean immediately centralized. Core networks can be deployed at the edge, inside enterprise headquarters, or on-premises to support localized deployments. Mostly when people refer to moving functionality from the core to the RAN, they refer to an edge-deployed core network functionality added to the RAN due to the here mentioned semantics issue.?
?? 2. Core Networks can be 100% software?- with proper scaling mechanisms SLAs can be achieved without strong requirements on the hardware. No strong requirements on hardware bring the opportunity to run on any infrastructure—cloud, on-premises, bare-metal servers, fridge compute units, etc.??
?? 3. The Current Network Functions/Microservices Model Isn't Sustainable?-?100+ network functions/microservices with strict interfaces in core networks create too much encapsulation-transmit-decapsulation of messages. We need a practical and at the same time backwards-compatible approach. Check Fraunhofer FOKUS Open6GCore (www.open6Gcore.org) for a rethink on network function splits. Ref:?https://ieeexplore.ieee.org/abstract/document/9758088
?? 4. The Control Plane Has No Real Real-Time Pressure?-?Unlike RAN functions that demand ultra-low latency, control plane operations are "near" real-time in Open RAN terms. Network signaling and procedures can take their time in most situations, so strict real-time constraints aren’t critical.?
?? 5. The Data Plane Is a Different Beast - Local breakout, routing, and UPF anchoring define network efficiency. Placing UPFs dynamically is key to optimizing latency, performance, and resource allocation. A cross-layer between transport and session, especially when transport is unreliable (everything is unreliable looking at the KPIs of the HRLLC use cases).?
?? 6. Edge Works Where Edge Naturally Exists?- if there is an edge, it can be used also for core networks:
?? 7. Edge Deployments should be treated with attention?-?Not every deployment needs edge computing. If fiber backhaul is available or the network is highly sparse (like in airplanes), keeping the centralized control plane reduces operational complexity and OPEX (should be obvious).
?? 8. Some Elements Will Stay Central?-?Subscriber data will always have a central component—whether it’s a profile database or a charging function. This does not mean local caches are excluded.?
?? 9. Seamless Communication across administrative domains?-?End-to-end connectivity requires handling micro-white spaces in urban areas, seamless interworking across different administrative domains, and lightweight, dynamic interfaces to bridge multiple administrative policies. Ref:?https://ieeexplore.ieee.org/abstract/document/9609936
??? 10. “Independent" Slices?- an absurd name for running completely isolated core networks for each of the verticals. Independence does not necessarily mean a lack of interworking, however not the tight coupling at the entry in the core network, but instead the seamless core interactions.?
?? 11. The Core is in the best position to reinterpret Near-RIC functionality?-?Subscriber-aware load balancing and traffic engineering is nothing more than session and mobility management. And this has direct implications for the dynamic network dimensioning to fit the network to the requirements in "near real-time". For this, we need proper standardization with conflict mitigation from the design phase. Ref:?https://ieeexplore.ieee.org/document/10073488
?? 12. Area-Based Services: Beyond Subscriber-Centric Thinking - considering that many devices and parts of the network are related to a location not necessarily to a user, a new breed of services can be considered: localized services - public safety, large-scale events, private networks - all can benefit on this type of services. Ref:?https://ieeexplore.ieee.org/abstract/document/10849759
?#5G #6G #CoreNetworks #TelecomInnovation #FutureOfNetworks
Senior Technology Strategy Manager bei Telefonica Germany - o2
1 个月Wunderbar