5GC Architecture and Dimensioning
Cloud-Native Approach
The 5GC network functions (NFs) are designed as cloud-native applications, which allows for flexible
deployment and scaling on cloud platforms like Google Cloud. These NFs are containerized and can be
orchestrated using Kubernetes, which Google originally developed.
Dimensioning Process
Resource Estimation
1. Workload Analysis:
- Analyze expected traffic patterns and user behavior
- Consider peak and average loads for each NF
2. Performance Requirements:
- Determine latency, throughput, and reliability requirements for each NF
- Consider Service Level Agreements (SLAs) for different network slices[4]
3. Interdependencies:
- Map out interactions between NFs
- Account for how the performance of one NF affects others[5]
Machine Learning-Based Dimensioning
Google Cloud's AI and ML capabilities can be leveraged to optimize resource allocation:
1. Reinforcement Learning (RL) Model:
- Train an RL model to understand NF behavior and resource needs
- Use historical data and simulations to improve accuracy[5]
2. Dynamic Adjustment:
- Continuously monitor NF performance
- Adjust resource allocation based on real-time demands[5]
A high-level design for a 5G Packet Core architecture on a cloud platform incorporates several key
components and principles to enable the advanced capabilities of 5G networks. Here's an overview of
the architecture:
Core Components
Control Plane Functions
1. Access and Mobility Management Function (AMF)
- Handles connection and mobility management
- Authenticates and authorizes user access
2. Session Management Function (SMF)
- Manages user sessions and IP address allocation
- Controls the User Plane Function (UPF)
3. Policy Control Function (PCF)
- Provides policy rules for network functions
- Manages Quality of Service (QoS) policies
4. Unified Data Management (UDM)
- Stores subscriber data and profiles
- Handles authentication credentials
5. Network Exposure Function (NEF)
- Exposes network capabilities to external applications
- Enables secure access to network services
User Plane Function (UPF)
- Handles user data packet routing and forwarding
- Enforces QoS policies
- Serves as the interconnection point between the RAN and data network
## Architectural Principles
### Service-Based Architecture (SBA)
- Implements a modular design with independent network functions
- Utilizes REST APIs for inter-function communication
- Enables flexible scaling and deployment of individual functions
### Cloud-Native Design
- Leverages containerization for network functions
- Utilizes Kubernetes for orchestration and management
- Enables microservices-based implementation
### Network Slicing
- Supports creation of virtual network slices for different use cases
- Allows customization of network resources for specific requirements
### Edge Computing Integration
- Enables deployment of core functions at the network edge
- Supports low-latency applications and services
## Deployment Model
### Multi-Tier Architecture
1. Far Edge
- Hosts RAN functions and initial aggregation points
- May include limited core functions for ultra-low latency use cases
2. Near Edge
- Hosts user plane functions (UPF) and select control plane functions
- Serves as aggregation point for far edge sites
3. Regional/Central Data Centers
- Hosts majority of control plane functions
- Manages overall network orchestration and management
### Cloud Infrastructure
- Utilizes a mix of private and public cloud resources
- Implements Network Function Virtualization (NFV) for flexible resource allocation
## Key Interfaces
- N1: Between UE and AMF
- N2: Between RAN and AMF
- N3: Between RAN and UPF
- N4: Between SMF and UPF
- N6: Between UPF and Data Network
This high-level design provides a flexible, scalable, and efficient 5G Packet Core architecture that can be
deployed on cloud platforms, enabling the advanced capabilities and use cases of 5G
networks[1][3][4][7].
Citations:
[1] https://www.ericsson.com/en/core-network/5g-core
[2] https://www.3g4g.co.uk/5G/5Gtech_6004_2017_11_Service-Based-Architecture-for-5G-Core-
Networks_HR_Huawei.pdf
[3] https://www.digi.com/blog/post/5g-network-architecture
[4] https://infohub.delltechnologies.com/en-us/p/the-5g-core-network-demystified/
[5] https://www.redhat.com/en/topics/5g-networks/evolution-to-a-5g-core
[6] https://niralnetworks.com/5g-core-network-architecture/
[7] https://docs.vmware.com/en/VMware-Telco-Cloud-Platform/2.7/telco-cloud-platform-5G-edition-
reference-architecture-guide-27.pdf
[8] https://www.vamsitalkstech.com/5g/5g-core-5gc-platform-architecture/
## Provisioning on Google Cloud
Infrastructure Setup
1. Google Kubernetes Engine (GKE):
- Deploy GKE clusters to manage containerized 5GC NFs
- Utilize GKE's auto-scaling and self-healing capabilities
2. Google Distributed Cloud:
- Leverage Anthos for consistent deployment across edge, on-premises, and cloud environments[3]
Network Function Deployment
1. Containerization:
- Package each NF as a container image
- Store images in Google Container Registry
2. Kubernetes Deployment:
- Use Kubernetes manifests to define resource requirements
- Implement pod autoscaling for elastic capacity
3. Service Mesh:
- Deploy Istio or Anthos Service Mesh for inter-NF communication
- Implement traffic management and security policies
Specific NF Provisioning Examples
1. Access and Mobility Management Function (AMF):
- Deploy multiple instances for high availability
- Allocate resources based on expected number of connected devices
2. Session Management Function (SMF):
- Provision based on expected number of concurrent sessions
- Implement horizontal scaling for session management capacity
3. User Plane Function (UPF):
- Deploy at edge locations for low latency
- Allocate high-performance compute resources for packet processing
4. Unified Data Management (UDM):
- Utilize Google Cloud's managed databases for subscriber data
- Implement caching layers for frequently accessed information
Network Slicing Implementation
1. Slice Configuration:
- Define network slice templates in Google Cloud
- Allocate dedicated resources for each slice based on QoS requirements
2. Dynamic Slice Management:
- Use Google Cloud's orchestration tools to manage slice lifecycle
- Implement automated scaling based on slice-specific metrics
## Monitoring and Optimization
1. Google Cloud Operations Suite:
- Implement comprehensive monitoring for all NFs
- Set up alerts for performance thresholds and anomalies
2. AI-Driven Optimization:
- Utilize Google Cloud's AI Platform to analyze performance data
- Implement predictive scaling based on ML models
3. Continuous Improvement:
- Regularly update RL models with new data
- Refine resource allocation strategies based on observed patterns
By leveraging Google Cloud's robust infrastructure and advanced capabilities, Communication Service
Providers (CSPs) can efficiently dimension and provision their 5GC networks, ensuring optimal
performance, scalability, and cost-effectiveness. The cloud-native approach, combined with AI-driven
optimization, allows for a flexible and future-proof 5G core deployment.
Citations:
[1] https://infohub.delltechnologies.com/en-us/p/the-5g-core-network-demystified/
[2] https://jep-asset.akamaized.net/jiostaticresources/5g-pdfs/solution-brief-hyperscale-cloud-native-
5g-core.pdf
[3] https://cloud.google.com/blog/topics/telecommunications/how-csps-can-use-cloud-networks-to-
deliver-5g
[4] https://www.ericsson.com/en/core-network/5g-core
[5] https://www.ericsson.com/en/blog/2023/11/how-to-automate-resource-dimensioning-in-cloud
