Design of 5G Architecture Enhancements for Supporting Edge Split Computing Service Pipeline

Split computing is one of the key service paradigms that future 6G networks are expected to support. In this paradigm, emerging 6G applications, such as Artificial Intelligence (AI) and Machine Learning (ML) inference, or Extended Reality (XR), can be deployed as sequential or parallel sub-service pipelines distributed across multiple Multi-access Edge Computing (MEC) sites. To enable the deployment of split computing, recent studies in 5G and 6G have proposed various solutions that address critical challenges, including optimal task partitioning and the placement or resource allocation of sub-services across MEC nodes. However, the current 5G network architecture requires significant enhancements to implement these split computing orchestration solutions. This article proposes new components and procedures to the 5G Control Plane and User Plane by integrating the Computing-Aware Traffic Steering (CATS) framework and the Segment Routing IPv6 (SRv6)-based Mobile User Plane (MUP). A Proof-of-Concept (PoC) implementation demonstrates that the proposed architecture enhances the capability of existing 5G systems to support split computing pipelines by significantly reducing control plane-user plane configuration overhead as well as underlay network traffic steering path setup and modification latency.