Cellular Networks and 5G

More than half of web traffic and most app traffic travels over cellular networks at some point. Yet many engineers picture "5G" as simply a faster pipe. In reality a mobile network is two large systems: a radio access network (RAN) that connects devices over the air, and a core network that authenticates subscribers, manages mobility and sessions, and connects to the internet. Their design explains why mobile latency varies so much, why apps sit behind carrier-grade NAT or IPv6-only networks, why keeping the radio awake drains batteries, and what features like network slicing can and cannot do today.

This page explains the architecture of 4G and 5G networks, the parts of the standards that are deployed versus still on roadmaps, and the practical consequences for application developers and network engineers.

TL;DR

Quick Example

Adapting app behavior to the current network on Android — cellular or not, metered or not, and estimated bandwidth:

And on iOS, respecting expensive and Low Data Mode connections:

Core Concepts

RAN and Core

The user plane function can be placed close to users — the basis of multi-access edge computing (MEC), which runs applications near the radio network for lower latency. See Edge Computing.

Deployment Modes and Spectrum

Non-standalone (NSA) 5G adds 5G radio capacity while relying on an LTE anchor and the 4G core; it was the fastest path to launch and remains common. Standalone (SA) 5G uses the 5G Core end to end, enabling lower signaling latency, network slicing, and newer features.

5G spectrum falls into FR1 (roughly 410 MHz to 7.125 GHz, including the widely deployed mid-band around 3.5 GHz) and FR2 (millimeter wave above 24 GHz), which offers very high throughput over short distances and struggles with walls and bodies. What users experience depends far more on spectrum, cell load, and backhaul than on the "5G" label.

Mobility and Radio States

Devices move between cells through handovers while sessions continue. To save power, the radio switches between connected and idle (or inactive) states; waking it costs latency and energy, and after a transfer the radio stays in a high-power state for a "tail" period. Many small, frequent requests therefore drain batteries more than one batched request. See Mobile Background Work and Battery Limits.

Authentication and Identity

Subscribers authenticate with credentials on a SIM or eSIM using mutual authentication protocols (5G-AKA or EAP-AKA'). 5G conceals the permanent subscriber identifier (SUPI) by encrypting it as a SUCI over the air, addressing IMSI-catcher tracking that affected earlier generations. Phone numbers and SIMs remain weak authentication factors for applications because of SIM-swap fraud.

QoS and Network Slicing

The 5G system assigns traffic to QoS flows with characteristics (5QI values) for latency, priority, and packet loss. Network slicing creates logical networks with different characteristics on shared infrastructure — for example for industrial IoT, emergency services, or premium video. Slicing requires standalone 5G and operator offerings; it is not something a public app can turn on.

Service Categories and IoT

Private networks — using licensed, shared (such as CBRS in the United States), or local spectrum — bring cellular coverage and control to factories, ports, and campuses.

What It Means for Applications

Standards Versus Roadmaps

Mobile standards come from 3GPP releases. Release 15 introduced 5G; later releases added features such as RedCap, positioning, and non-terrestrial networks, and Release 18 began 5G-Advanced. 6G is in the study and requirements phase, with the ITU's IMT-2030 framework pointing to commercial systems around 2030. When evaluating claims, ask which release defines a feature, whether your operators have deployed it, and whether devices support it — "6G will enable X" statements are research goals, not deployed capabilities.

Best Practices

Design for the Median Mobile Network

Test with throttled, high-latency, lossy networks and with transitions between Wi-Fi and cellular.

Support IPv6-Only Networks

Use hostnames rather than IPv4 literals, and test on NAT64 networks (Apple requires IPv6-only network compatibility for App Store apps).

Batch and Defer Network Work

Group requests, avoid tight polling, and move large transfers to unmetered networks.

Separate Marketing From Capabilities

Verify features such as slicing, URLLC, or edge computing with your operators before designing around them.

Common Mistakes

Identifying Users by IP Address

Constant Polling

Assuming 5G Means Low Latency Everywhere

FAQ

What is the difference between 4G and 5G?

5G adds a new radio interface with wider channels and more spectrum options, plus a service-based core network that supports features such as network slicing. Real-world gains depend on spectrum, deployment mode, and network load.

What is standalone versus non-standalone 5G?

Non-standalone 5G uses 5G radios with a 4G anchor and core network. Standalone 5G uses the 5G Core end to end and enables features such as network slicing and lower signaling latency.

What is network slicing?

Creating multiple logical networks with different performance and isolation characteristics on shared 5G infrastructure. It requires standalone 5G and is offered by operators, typically to enterprises.

Why do mobile users share IP addresses?

Carriers often use carrier-grade NAT, placing many subscribers behind a smaller pool of public IPv4 addresses, and many networks are IPv6-only with translation for IPv4 destinations.

What is 6G?

The next generation of mobile networks, currently in research and standardization studies, with targets set by the ITU's IMT-2030 framework and commercial deployment expected around 2030.

Related Topics

References