31 5G Network Slicing for IoT
- Network Slice: A logically isolated end-to-end virtual network over shared physical 5G infrastructure, with guaranteed QoS parameters (bandwidth, latency, reliability)
- S-NSSAI (Single Network Slice Selection Assistance Information): 32-bit identifier composed of SST (Slice/Service Type, 8-bit) and SD (Slice Differentiator, 24-bit) used by UEs to request specific slices
- SST (Slice/Service Type): Standard 8-bit type code: 1=eMBB, 2=URLLC, 3=MIoT (Massive IoT), 4=V2X; operator-specific types use 128–254
- NSSF (Network Slice Selection Function): 5G Core NF that selects appropriate network slices for incoming UE connections based on subscription, local policy, and capacity
- Slice Isolation: Property ensuring that traffic, resources, and failures in one slice do not affect other slices; critical for co-hosting public safety and commercial IoT on the same infrastructure
- SLA (Service Level Agreement): Contract between operator and enterprise customer specifying guaranteed slice parameters: minimum bandwidth, maximum latency, availability (e.g., 99.999%), and geographic coverage
- RAN Slicing: Extending network slicing to the Radio Access Network with dedicated PRB (Physical Resource Block) allocation per slice; more complex than core slicing
- Slice Lifecycle Management: Operations for creating, modifying, scaling, and deleting slices dynamically via the 5G Management and Orchestration (MANO) framework
32 5G Network Slicing: Virtual Networks for Diverse IoT
By the end of this chapter, you will be able to:
- Explain network slicing concepts and the 5G Service-Based Architecture that enables them
- Design network slices for different IoT service types (eMBB, URLLC, mMTC)
- Compare private 5G deployment models for enterprise IoT
- Evaluate spectrum options for private 5G networks
- Configure QoS parameters and 5QI levels for IoT network slices
- Justify the selection of isolation levels (soft, hard, physical) based on application requirements
32.1 Prerequisites
Before diving into this chapter, you should be familiar with:
- 5G Device Categories: Device category selection
- 5G Advanced Overview: 5G evolution timeline
- Cellular IoT Fundamentals: Basic cellular IoT concepts
5G Deep Dives:
- 5G Advanced Overview - Evolution timeline
- 5G Device Categories - NB-IoT to 5G NR
- 5G URLLC and Future - Critical IoT and 6G
Enterprise Deployment:
- Private 5G Networks - Deployment guide
- Cellular IoT Applications - Use cases
In one sentence: Network slicing enables multiple virtual networks with different SLAs to coexist on shared 5G infrastructure, allowing one deployment to serve eMBB video cameras, URLLC industrial control, and mMTC sensors simultaneously.
Remember this: Each slice is like a dedicated highway lane - eMBB is the express lane (high throughput), URLLC is the emergency lane (guaranteed priority), and mMTC is the carpool lane (high vehicle count, moderate speed).
32.2 For Beginners: Understanding Network Slicing
The Problem: Different IoT applications have vastly different requirements: - Factory robots need ultra-low latency (< 1 ms) - Video cameras need high bandwidth (50+ Mbps) - Sensors need high density support (10,000+ per km²)
Traditional Solution: Build separate networks for each use case. Expensive!
5G Solution: Network Slicing - create multiple “virtual networks” on one physical infrastructure.
Analogy: Think of it like a highway system: - Physical infrastructure = The actual roads, bridges, tunnels - Network slices = Different lane types on the same highway
| Lane Type | Network Slice | Optimized For |
|---|---|---|
| Express Lane | eMBB | High speed, high throughput |
| Emergency Lane | URLLC | Guaranteed access, no delays |
| Carpool Lane | mMTC | Many vehicles, efficient capacity |
Key Benefit: Each IoT application gets exactly the performance it needs without over-provisioning.
“Network slicing is like having different lanes on a highway, all on the same road!” Sammy the Sensor explained. “One lane is reserved for emergency vehicles that need to go fast with no delays – that is URLLC for factory robots. Another lane is for big trucks carrying heavy loads – that is eMBB for video cameras. And a third lane is for lots of small cars traveling together – that is mMTC for millions of sensors like me!”
“The magic is that they all share the same physical infrastructure,” Lila the LED added. “One set of cell towers and cables supports all these different virtual networks. It is like an apartment building where each tenant has their own space with their own rules, but they all share the same building structure.”
Max the Microcontroller nodded. “Each slice gets guaranteed resources. My factory robot slice gets sub-millisecond latency no matter how busy the network is. The video surveillance slice gets high bandwidth. And the sensor slice supports huge numbers of connections. Nobody can steal resources from another slice.”
“Private 5G takes this even further,” Bella the Battery said. “A factory can have its own private 5G network with custom slices. One for robot arms, one for cameras, and one for environmental sensors. All the data stays on-site, and the factory controls everything. It is like having your own private highway instead of sharing a public one!”
32.3 Network Slicing Architecture
32.3.1 What is Network Slicing?
Network slicing creates virtual, isolated networks within a shared 5G infrastructure:
32.3.2 Slice Types for IoT
| Slice Type | SLA Guarantee | IoT Use Case |
|---|---|---|
| eMBB | Throughput (100+ Mbps) | Video surveillance, AR/VR |
| URLLC | Latency (<1 ms), Reliability (99.999%) | Factory automation, autonomous vehicles |
| mMTC | Density (1M devices/km²) | Smart meters, agriculture sensors |
| Custom | Application-specific | Private IoT networks |
32.3.3 5G Core Architecture for Slicing
32.3.4 Key Network Functions
| Function | Role | Slicing Impact |
|---|---|---|
| NSSF | Network Slice Selection Function | Chooses appropriate slice for device |
| AMF | Access and Mobility Management | Manages device registration per slice |
| SMF | Session Management Function | Configures QoS per slice |
| UPF | User Plane Function | Routes traffic through slice |
32.4 Private 5G Networks
32.4.1 Why Private 5G?
| Benefit | Description |
|---|---|
| Dedicated Capacity | No sharing with public users |
| Custom Coverage | Optimized for specific site |
| Data Sovereignty | Traffic stays on-premises |
| Low Latency | Local core network |
| Security | Isolated from public network |
| Control | Enterprise manages policies |
32.4.2 Deployment Models
32.4.3 Comparing Deployment Models
| Factor | Standalone Private | Hybrid | Network Slice |
|---|---|---|---|
| CAPEX | High ($200K-1M+) | Medium ($100K-500K) | Low ($0-50K) |
| OPEX | Medium (self-managed) | Low-Medium | Per-device fees |
| Control | Full | Partial | Limited |
| Latency | Lowest (all local) | Low | Medium |
| Data Sovereignty | Complete | Partial | Carrier-dependent |
| Best For | Large enterprises, critical IoT | Mid-size, mixed requirements | SMBs, testing |
32.4.4 Spectrum Options for Private 5G
| Spectrum | Region | Bandwidth | License |
|---|---|---|---|
| CBRS (3.5 GHz) | USA | 150 MHz | Light licensing |
| n78 (3.5 GHz) | Europe, Asia | Varies | Country-specific |
| mmWave (26-28 GHz) | Global | 400+ MHz | Licensed/shared |
| n79 (4.5 GHz) | Japan, China | 100 MHz | Licensed |
| Unlicensed (5/6 GHz) | Global | 500+ MHz | Unlicensed |
32.4.5 Private 5G Cost Analysis
Typical Standalone Private 5G Deployment:
| Component | Cost Range | Notes |
|---|---|---|
| Small cells (8-12) | $80K-180K | Indoor/outdoor coverage |
| Private 5G core | $30K-100K | On-premises or cloud |
| Edge computing | $20K-50K | Local processing |
| Spectrum license | $5K-50K/year | CBRS, local licensing |
| Integration | $30K-100K | One-time setup |
| Year 1 Total | $165K-480K | |
| Annual Ongoing | $25K-75K | Maintenance, spectrum |
32.5 Understanding Check
Scenario: A logistics company with a 2 km² distribution center needs connectivity for: - 200 autonomous forklifts (URLLC requirement) - 50 loading dock cameras (10 Mbps each) - 5,000 package tracking tags (location every 10 seconds)
Questions:
- Should they use private 5G or carrier network slicing?
- How would you design the network slices?
- What deployment model fits best?
32.6 Worked Example: Private 5G vs Public 5G for Logistics Hub
Scenario: A logistics company operates a 2 km² distribution center with: - 200 autonomous forklifts (URLLC: <5 ms latency) - 50 cameras (10 Mbps video streams) - 5,000 package tracking tags (location every 10 seconds)
Given:
- Annual budget: $200,000
- Carrier offers public 5G slice at $50/device/month for URLLC
- Coverage area: 2 km² (outdoor yard + indoor warehouse)
Steps:
Calculate public 5G slice costs (Option A):
URLLC devices (forklifts): 200 × $50/month × 12 = $120,000/year eMBB devices (cameras): 50 × $30/month × 12 = $18,000/year mMTC devices (tags): 5,000 × $5/month × 12 = $300,000/year Total annual cost: $438,000 (exceeds budget by 2.2×)Calculate private 5G costs (Option B):
Infrastructure: 8 small cells × $15,000 = $120,000 (CAPEX) Private 5G core: $50,000 (CAPEX) CBRS spectrum license: $5,000/year Maintenance and support: $20,000/year Year 1 total: $195,000 Year 2+ total: $25,000/year 5-year TCO: $195,000 + 4 × $25,000 = $295,000Compare capabilities: | Factor | Private 5G | Public Slice | |——–|————|————–| | Latency | 2-5 ms (local edge) | 10-20 ms (carrier core) | | Data sovereignty | All on-premises | Carrier network | | Capacity control | Dedicated | Shared |
Calculate 5-year savings:
Public 5G: $438,000 × 5 = $2,190,000 Private 5G: $295,000 Savings: $1,895,000 (87% reduction)
Result: Deploy private 5G with CBRS spectrum. Year 1 fits budget at $195,000, and years 2-5 cost only $25,000/year. Additional benefits: true URLLC latency for forklift safety, complete data sovereignty, no per-device fees for scaling.
The 5-year TCO comparison reveals the break-even point for private 5G. Using the logistics hub numbers (5,250 total devices):
Per-device annual public cost (weighted average): \[ \text{Public annual} = \frac{\$438{,}000}{5{,}250} = \$83.43 \text{ per device per year} \]
\[ \text{Public 5-year cost} = \$83.43 \times 5 = \$417 \text{ per device over 5 years} \]
\[ \text{Private 5-year TCO} = \$295{,}000 \text{ (from calculation above)} \]
Break-even device count (\(N\)) occurs when total private 5G cost equals total public 5G cost:
\[ \$295{,}000 = N \times \$417 \]
\[ N = \frac{\$295{,}000}{\$417} \approx 707 \text{ devices} \]
Above 707 devices, private 5G delivers lower TCO. With 5,250 devices, the logistics hub saves $1,895,000 over 5 years – a 6.4x return on the $295,000 private 5G investment.
Key Insight: For dense, geographically bounded deployments with URLLC requirements, private 5G often delivers better economics and performance than carrier slices. Break-even is typically 500-1,000 devices.
32.7 Worked Example: Network Slice Configuration for Smart Hospital
Scenario: A 500-bed hospital needs network slices for: - 200 patient vital sign monitors (99.99% reliability, <50ms latency) - 50 mobile medical imaging devices (100 Mbps per device) - 2,000 asset tracking tags (best-effort acceptable)
Given:
- Budget: $75,000/year connectivity
- HIPAA compliance required
- Carrier offers network slicing with various 5QI levels
Slice Design:
Slice 1: URLLC for Patient Monitors
5QI: 82 (Delay Critical GBR)
Guaranteed Bit Rate: 100 kbps per device
Packet Delay Budget: 10 ms
Packet Error Loss Rate: 10^-5 (99.999% reliability)
Priority Level: 19 (highest)Slice 2: eMBB for Medical Imaging
5QI: 6 (Non-GBR, TCP-based)
Maximum Bit Rate: 150 Mbps per device
Packet Delay Budget: 100 ms
Priority Level: 60 (medium)Slice 3: mMTC for Asset Tracking
5QI: 79 (Non-GBR, Low Priority)
Data rate: 10 kbps per device
Packet Delay Budget: 500 ms
Priority Level: 90 (lowest)Cost Optimization: Pure carrier slicing exceeds budget at $270,600/year. Optimized hybrid approach: - Private 5G for imaging (on-premises DICOM data) - URLLC slice for 50 critical ICU monitors only - NB-IoT for asset tags
Optimized Annual Cost: $136,000 (45% reduction from pure slicing)
Key Insight: Use carrier URLLC slices only for truly life-critical applications (ICU, OR monitors) where contractual SLAs are essential. Handle high-bandwidth imaging via private 5G and use NB-IoT for non-critical tracking.
32.8 5G vs LPWAN Technology Positioning
32.9 Slice Isolation and Security: Why It Matters for IoT
Network slicing promises logical separation on shared physical infrastructure, but the degree of actual isolation varies significantly between deployment models. For IoT architects, understanding isolation guarantees is critical – especially when safety-critical devices (medical, industrial control) share infrastructure with best-effort traffic.
Three levels of slice isolation:
| Isolation level | Mechanism | Guarantee | Use case |
|---|---|---|---|
| Soft isolation | QoS scheduling priority | Best-effort SLA; no hard resource reservation | Smart meters, environmental sensors |
| Hard isolation | Dedicated PRBs (Physical Resource Blocks) | Guaranteed minimum bandwidth and latency | Factory automation, fleet telematics |
| Physical isolation | Dedicated gNB hardware or frequency band | Air-gapped; no shared resources | Critical infrastructure, defense |
Why soft isolation fails for URLLC:
In 2022, a European automotive OEM tested autonomous guided vehicles (AGVs) on a shared 5G network with an eMBB slice for employee devices and an mMTC slice for warehouse sensors. The AGV URLLC slice had soft QoS isolation (priority scheduling).
During a stress test, 200 employees simultaneously streamed video during a lunch break, saturating the eMBB slice. The scheduler correctly prioritized URLLC traffic, but the physical uplink control channel (PUCCH) contention caused 12 ms scheduling delays that propagated to the URLLC slice – violating the 10 ms end-to-end latency requirement for AGV collision avoidance.
The fix required upgrading to hard isolation with 4 dedicated PRBs reserved exclusively for the URLLC slice (out of 52 total PRBs on a 20 MHz n78 carrier). This reduced available eMBB capacity by 7.7% but guaranteed URLLC latency remained below 5 ms regardless of eMBB load.
Cost of isolation (per site, annual):
| Isolation level | Additional cost | Capacity impact |
|---|---|---|
| Soft (QoS only) | USD 0 | 0% capacity loss |
| Hard (dedicated PRBs) | USD 2,000–5,000 (configuration) | 5–15% capacity reserved |
| Physical (dedicated hardware) | USD 50,000–200,000 (separate gNB) | N/A (fully separate) |
Decision guidance: Use soft isolation for non-safety applications where occasional SLA violations are acceptable (smart building, agriculture). Use hard isolation when latency guarantees are contractual or safety-related (manufacturing, logistics). Reserve physical isolation for critical infrastructure where regulatory requirements mandate air-gapped networks (power grid, military).
32.10 Summary
Network slicing creates virtual networks with different SLAs on shared infrastructure
Three standard slice types: eMBB (throughput), URLLC (latency/reliability), mMTC (density)
Private 5G models: Standalone (full control), Hybrid (shared core), Network Slice (carrier-managed)
CBRS spectrum (USA) enables light-licensed private 5G at 3.5 GHz
Private 5G ROI: Break-even typically at 500-1,000 devices vs carrier subscriptions
Hybrid architectures often optimal: URLLC slice for critical, private for high-bandwidth, NB-IoT for massive
32.11 Concept Relationships
32.12 See Also
32.13 Try It Yourself
Common Pitfalls
Network slicing requires 5G Standalone (SA) architecture with the 5G Core (5GC). The majority of 2024–2025 deployments use Non-Standalone (NSA) architecture (5G NR radio + 4G core), which does not support true network slicing. Verify SA deployment with the operator before designing applications that depend on specific slice QoS guarantees. NSA networks use LTE QCI (QoS Class Identifier) for differentiation, not slicing.
Network slice SLAs specify committed (average) QoS parameters, not instantaneous guarantees. During peak load or radio congestion, individual packets in a URLLC slice may still experience >1 ms latency. For true deterministic latency requirements (factory automation, surgical robots), complement 5G slicing with application-layer scheduling and edge computing to absorb network jitter.
Enterprises often request maximum-bandwidth slices for all IoT devices “just in case,” resulting in 10–100× overprovisioning and corresponding SLA cost. Conduct traffic profiling first: measure actual peak bandwidth, burstiness (burst duration and inter-burst interval), and latency sensitivity per device type. Then provision slices for measured 95th-percentile demand with burst allowances, not theoretical maximum.
5G UE firmware must be configured with the correct S-NSSAI to join the intended slice. Devices that do not specify S-NSSAI fall into the default slice (typically eMBB), bypassing the IoT URLLC or mMTC slices with their QoS guarantees. Ensure modem AT commands (AT+CNSSAI for Quectel, AT#SLPSCFG for Telit) are set to the operator-assigned S-NSSAI values during provisioning.
32.14 What’s Next
| Chapter | Focus Area | Link |
|---|---|---|
| 5G URLLC and Future | Mission-critical IoT with sub-1 ms latency and 6G vision | 5G URLLC and Future |
| Private 5G Networks | Detailed deployment guide for enterprise private 5G | Private 5G Networks |
| 5G Device Categories | NB-IoT, LTE-M, RedCap, and full 5G NR device selection | 5G Device Categories |
| Cellular IoT Applications | Real-world cellular IoT deployments and use cases | Cellular IoT Applications |