5g network architecture explained Topical Map Library Entry

NSA uses the existing 4G LTE core (EPC) for control-plane functions and anchors 5G NR radio to LTE, while SA uses a native 5G Core (5GC) that provides cloud‑native control, user-plane separation, and native 5G functions like network slicing and URLLC. In short: NSA is a radio upgrade riding on 4G signaling; SA is a full end‑to‑end 5G architecture.

NSA remains sensible for operators that need rapid time‑to‑market and want to reuse existing LTE investments; it lowers initial CAPEX and simplifies early 5G coverage. However, operators targeting low latency, network slicing, private networks, or advanced enterprise services will need to migrate to SA to unlock those features.

Real-world SA deployments commonly reduce end‑to‑end latency into the single‑digit to low‑teens of milliseconds for optimized user‑plane paths, whereas NSA implementations tied to LTE control planes often exhibit latencies in the ~20–50 ms range depending on transport and core location. Actual improvement depends on UPF placement, transport RTT, and RAN tuning.

True, standards-compliant network slicing (with per‑slice control, orchestration and SLA enforcement) requires 5GC capabilities and is therefore a native SA feature. Some vendors offer limited slicing-like segmentation on NSA via RAN/QoS tricks, but these do not deliver the end‑to‑end isolation, orchestration and billing controls that enterprise slices need.

Operators typically follow phased strategies: (1) initial 5G NR rollouts on NSA to cover hotspots quickly, (2) deploy a cloud‑native 5GC in parallel (greenfield or virtualization of EPC functions), (3) introduce dual‑mode cores (supporting EPC and 5GC interworking), and (4) progressively shift services, UPF placement and orchestration to SA. Many migrations take 12–36 months per market depending on vendor interoperability and regulatory requirements.

SA's cloud‑native and service‑based architecture introduces new threat surfaces (API exposure, microservice vulnerabilities) but also enables stronger security controls like slice‑level isolation, service authentication and more granular policy enforcement. Lawful intercept and regulatory reporting require updated implementations in SA (different interfaces and data models), so operators must validate compliance when migrating cores.

NSA typically reduces near‑term CAPEX by leveraging existing EPC and LTE RAN investments and lowers OPEX through simpler incremental operations, while SA demands higher upfront investment for 5GC, cloud infrastructure and modern orchestration but can reduce long‑term OPEX via automation, containerized scaling and multi‑tenant revenue opportunities. The breakeven point commonly appears once advanced services or slicing generate enterprise ARPU uplift.

Use cases that require deterministic latency, massive device orchestration or per‑slice SLAs—such as industrial URLLC (robotics, motion control), edge cloud gaming, carrier‑grade fixed wireless access with dedicated SLAs, and private campus networks—require SA. Consumer eMBB improvements can be delivered in NSA, but features like guaranteed low latency and true slicing need SA.

NSA typically uses LTE as the anchor carrier for control and may aggregate mid‑band LTE with mmWave NR for throughput, whereas SA allows full 5G control‑plane over NR enabling more flexible spectrum usage (e.g., dynamic spectrum sharing, dual‑connect without LTE anchor) and easier deployment of standalone mmWave or CBRS-based 5G services. Spectrum licensing and guardband considerations still shape aggregation choices.