In-Network Computing: An Architectural Framework for the 6G Core

Modern telecommunications networks are no longer required to perform only traffic management and connection establishment functions, but are increasingly evolving into general-purpose processing engines. This is facilitated by InNetwork Computing (INC), which has emerged as a key enabling paradigm in this direction. This shift in vision, however, requires a complete redesign of network infrastructures, with a strong emphasis on deep integration between networking and artificial intelligence capabilities. In the wake of these considerations, in this work we introduce an INC framework for the 3GPP 6G architecture to offload processing from the user device to the processing instances of the 6G mobile system, based on service requirements and constraints of individual user sessions, and delve into 3GPP Control Plane (CP) and User Plane (UP) aspects. Furthermore, we provide an experimental evaluation of the advantages of our INC framework in saving processing resources and energy in the user device while meeting the performance requirements of an AI-powered application.