Network Slicing Architecture, Isolation Mechanisms, and Challenges in 5G/6G

Introduction Network slicing is not a new concept; in 5G it enables multiple applications with different QoS requirements to share the same physical infrastructure by virtualizing it into separate logical networks, each with customized topology and dedicated resources.

Architecture Framework The system consists of two layers: the Network Layer, which provides user and control‑plane functions for each slice, and the OSS Layer, which hosts design, configuration, and operation assets for slice management. The Network Layer is modular, allowing flexible composition of network functions to build slices for the three main 5G service categories. The OSS Layer offers tools for design, data management, assurance, and orchestration, with orchestration handling slice configuration and runtime operation.

End‑to‑End Isolation To ensure that one slice does not affect another, three levels of isolation are defined: business isolation (traffic of one slice is invisible to others), resource isolation (physical or logical separation of resources), and operation‑maintenance (independent management interfaces for each tenant).

Access‑Network Isolation Isolation in the radio access network can be physical (dedicated spectrum) or logical (resource blocks allocated per slice). At the DU/CU level, isolation is achieved via dedicated hardware, virtual machines, or containers, depending on whether the CU runs on dedicated hardware or a general‑purpose server.

Transport‑Network Isolation Soft isolation uses VLAN tagging to map slice identifiers to VLANs, while hard isolation employs FlexE technology, which inserts a time‑division multiplexing shim at the PCS layer to provide physical time‑slot separation, enabling multiple customers to share the same Ethernet port with strict isolation.

Core‑Network Isolation The 5G core is built on virtualized infrastructure with three layers: resource, network‑function, and management‑orchestration. Isolation can be physical (dedicated hardware) or logical (NFV‑based VM/container isolation, security domains, and tenant‑specific management accounts), with additional protection via encrypted management channels.

Challenges End‑to‑end slice consistency across domains creates operational difficulties: varying readiness of slice features across standards bodies, scalability pressure on OSS due to many concurrent slices, and the need to integrate multi‑vendor solutions while ensuring coordinated operation.