5G Standalone (SA) vs Non-Standalone (NSA) Core: Network Slicing, Service-Based Architecture, and UPF Offloading

When mobile network operators initially rolled out 5G services globally, commercial marketing promised revolutionary transformations: sub-millisecond latencies, massive machine-type communication for billions of IoT devices, and dedicated programmatic network slices. Yet early consumers often noticed little practical difference beyond marginally faster download speeds on speed test applications. The technical reason for this disparity lies in the fundamental architectural distinction between 5G Non-Standalone (NSA) and true 5G Standalone (SA).

Non-Standalone 5G was an interim stepping stone that grafted 5G New Radio (NR) base stations onto legacy 4G Evolved Packet Core (EPC) infrastructure. True generational capabilities (such as end-to-end Network Slicing, cloud-native Service-Based Architecture, and edge User Plane Function offloading) are exclusively unlocked by the pure 5G Core (5GC) in Standalone deployments. This architectural guide explores the functional differences between NSA and SA, 3GPP Option 3x vs Option 2 deployments, the Service-Based Architecture (SBA), and dynamic Network Slicing implementation.

1. Foundations: 3GPP Deployment Options (Option 3x vs Option 2)

To enable rapid commercial deployment without forcing telecommunications carriers to replace their multi-billion-dollar core networks overnight, the 3rd Generation Partnership Project (3GPP) defined multiple migration pathways:

5G Non-Standalone (3GPP Option 3x):

In Option 3x, the cellular handset communicates simultaneously with both an LTE eNodeB base station and a 5G gNodeB base station via E-UTRA-NR Dual Connectivity (EN-DC). The critical constraint is that the entire control plane anchor remains tethered to the legacy 4G Evolved Packet Core (EPC). Signaling, mobility handshakes, authentication, and session setup are handled by 4G MME (Mobility Management Entity) hardware. The 5G carrier frequency is utilized merely as a secondary “data booster” pipe.

5G Standalone (3GPP Option 2):

Option 2 completely severs dependencies on legacy 4G infrastructure. The 5G gNodeB base stations connect directly to an all-new, cloud-native 5G Core (5GC). Both control signaling and data transport operate purely on 5G protocols, unlocking microsecond-level synchronization, deterministic Quality of Service (QoS), and network slicing capabilities.

2. Control Plane and User Plane Splitting: Dual Connectivity (EN-DC)

The operational penalty of 5G NSA is rooted in the dual-connectivity radio protocol stack:

• Battery Drain: A mobile phone connected to 5G NSA must power two independent radio transceivers simultaneously: one tracking the 4G LTE anchor carrier and one streaming 5G NR data. This increases handset radio power consumption by 20% to 35%.
• Call Setup and Signaling Latency: Because all connection setup messages must traverse the legacy 4G MME over Diameter signaling protocols, session initialization latency remains constrained to 4G performance bounds (typically 30 to 50 milliseconds). True sub-10ms response times are physically unachievable under NSA.
• Coverage Anchoring: If a user moves into an area with robust 5G coverage but weak 4G LTE signal, the 5G connection drops immediately because the 4G anchor link failed. Standalone 5G operates autonomously, extending coverage ranges significantly.

3. The 5G Core (5GC) Service-Based Architecture (SBA)

The true technological revolution of 5G Standalone is its Service-Based Architecture (SBA). Legacy 4G core networks were built from monolithic hardware appliances connected by rigid point-to-point telecom protocols. The 5G Core is fundamentally an enterprise cloud-native microservices cluster running inside Kubernetes:

• RESTful HTTP/2 and JSON Interfaces: Network functions communicate using standard web APIs (HTTP/2 with JSON payloads) rather than proprietary telecom protocols, enabling automated CI/CD pipeline deployments.
• AMF (Access and Mobility Management Function): Replaces the 4G MME, managing device authentication, connection states, and mobility handovers.
• SMF (Session Management Function): Dynamically creates, modifies, and tears down user data sessions, configuring traffic steering rules on the User Plane Function.
• NRF (Network Repository Function): Functions as a microservices service registry (similar to Consul or Eureka), allowing network functions to discover each other dynamically.
• NSSF (Network Slice Selection Function): Evaluates device subscription parameters to allocate incoming traffic to specific virtual network slices.

4. End-to-End Network Slicing: eMBB, URLLC, and mMTC

Network Slicing is the premier capability of 5G Standalone. Slicing allows a single physical telecommunications network to be partitioned into multiple isolated, independent virtual networks tailored to divergent service requirements:

1. Enhanced Mobile Broadband (eMBB): Optimized for extreme throughput (multi-gigabit downloads, 8K video streaming, immersive virtual reality). Allocates wide radio channels and massive buffer queues.
2. Ultra-Reliable Low-Latency Communication (URLLC): Engineered for autonomous vehicles, industrial robotics, and surgical teleoperation. Guarantees 99.999% packet delivery reliability with sub-1ms radio interface latencies by reserving dedicated mini-slot radio resources and zero-buffer routing.
3. Massive Machine-Type Communication (mMTC): Designed for smart utility meters and agriculture sensors. Supports up to 1,000,000 devices per square kilometer with low data rates and extended battery sleep cycles lasting up to 10 years.

Each network slice operates with its own dedicated Network Slice Instance (NSI), isolated data paths, and tailored SLAs, guaranteeing that heavy consumer smartphone video streaming on an eMBB slice can never degrade latency on a mission-critical hospital URLLC slice.

5. User Plane Function (UPF) Architecture and Edge Offloading

In 5G SA, the data transport pipeline is completely decoupled from control signaling through the User Plane Function (UPF):

The UPF is the single workhorse network function that processes user data packets. Because the control plane (SMF/AMF) is completely separated via the N4 interface (Packet Forwarding Control Protocol, PFCP), operators can deploy lightweight UPF instances anywhere in the network fabric:

• In centralized carrier datacenters for standard public internet traffic.
• In regional metro edge hubs for content delivery caching.
• Directly inside enterprise factory floors for zero-latency local data processing.

6. Comprehensive Technical Comparison: 5G NSA vs 5G SA

Architectural Dimension	5G Non-Standalone (Option 3x)	5G Standalone (Option 2)
Core Network Technology	Legacy 4G Evolved Packet Core (EPC)	Cloud-Native 5G Core (5GC SBA)
End-to-End Latency	25 – 45 ms	< 5 – 10 ms (Sub-1ms radio interface)
Network Slicing Support	Not Supported (Monolithic QCI QoS only)	Full End-to-End Dynamic Slicing (eMBB, URLLC)
Handset Radio Power Draw	High (Simultaneous 4G + 5G radios active)	Low to Moderate (Single 5G radio active)
Voice Call Architecture	VoLTE (Voice over LTE via 4G anchor)	VoNR (Voice over New Radio)
Signaling Protocol Stack	Diameter / GTP-C over SCTP	HTTP/2 REST APIs with JSON serialization

7. Enterprise Transformation: Private 5G Networks and Campus Slicing

The enterprise adoption of 5G is driven entirely by Standalone architecture. Manufacturing facilities, ports, and healthcare campuses deploy private 5G SA networks using dedicated local spectrum (such as CBRS band in the United States or localized N77/N78 bands globally):

• Total Data Sovereignty: Because the UPF sits on-premises, enterprise video streams and production telemetry never leave the corporate physical campus, satisfying stringent cybersecurity and regulatory compliance mandates.
• Deterministic Robotic Control: Replacing volatile industrial Wi-Fi with URLLC slices eliminates packet drops caused by metallic factory interference, enabling continuous automated guided vehicle (AGV) operation.

8. Frequently Asked Questions

Why did telecom carriers deploy 5G NSA before 5G SA?

Speed to market and capital expenditure. Building a cloud-native 5G Core requires completely re-architecting carrier datacenter infrastructure and deploying Kubernetes orchestration. 5G NSA allowed carriers to market “5G” immediately by simply mounting 5G antennas on existing 4G cell towers.

How does a smartphone know which network slice to connect to?

The device operating system and SIM card maintain Network Slice Selection Policy (NSSP) rules. When an enterprise application opens a socket connection, Android or iOS matches the application ID to an S-NSSAI (Single Network Slice Selection Assistance Information) token, requesting that specific slice from the 5G Core.

What is VoNR (Voice over New Radio)?

VoNR is native voice calling over pure 5G Standalone networks. In 5G NSA, when a user makes a phone call, the phone temporarily falls back to 4G VoLTE. In 5G SA, voice calls remain on the 5G carrier with crystal-clear audio codecs and near-instant call connection times.