Lab 2 – Distributed Open5GS + UERANSIM on AWS (Two‑EC2 Architecture)

~The pre‑configured AWS AMI for this lab is available to share. For details, see the About page.~

This lab extends the work from Lab 1, but introduces a major architectural change:
instead of running both Open5GS and UERANSIM on a single EC2 instance, the system is now distributed across two EC2 instances, each in a separate Availability Zone.

The goal of Lab 2 was to validate that Open5GS and UERANSIM can operate across AWS infrastructure boundaries, with correct routing, SCTP/GTP‑U reachability, and proper N2/N3 interface binding.

Image title — Lab 2 – Distributed AWS Architecture

1. Architecture Overview

Lab 2 uses two EC2 instances inside the same VPC:

EC2 #1 – UERANSIM

Runs gNodeB
Runs UE
Handles RRC, NGAP, and GTP‑U uplink/downlink
Must reach AMF (N2) and UPF (N3) over private IPs

EC2 #2 – Open5GS Core

Runs AMF, SMF, UPF, NRF, SCP, UDM, UDR, AUSF
UPF exposes GTP‑U (UDP/2152)
AMF exposes NGAP (SCTP/38412)

Both instances remain inside the same VPC, so routing is automatic.
However, AWS does not allow SCTP or GTP‑U by default, so Security Groups must be configured manually.

2. Key Differences from Lab 1

Lab 1 used a single EC2 instance, so all interfaces were loopback‑local or internal to the same machine.

Lab 2 introduces the following changes.

(A) Distributed N2 and N3 interfaces

You must update the following in UERANSIM configs to use EC2 private IPs.