How to have concurrent users on a single EC2 Instance of Unreal Engine projects of Pixel Streaming?

For your Unreal Engine 5 Pixel Streaming scenario with multiple users, you should consider implementing a multi-session approach to provide each user with their own private experience rather than a shared one.

Based on best practices for multi-session environments, you'll need to:

1. Right-size your instance: For graphics-intensive applications like Unreal Engine, consider using graphics-optimized instances such as the G4dn family. The recommendation for graphics-intensive workloads is approximately 1 user per 2 vCPUs on Graphics.g4dn.* instances.

2. Determine optimal user density: When deciding how many users to host on a single instance:

• Consider the resource requirements of your Unreal Engine project
• Evaluate the instance specifications (CPU, memory, GPU)
• Conduct performance testing with real-world usage patterns
• Monitor resource utilization during testing

3. Scale appropriately: If you need to support 10 concurrent users:

• You may need to use a larger instance size or multiple instances
• Consider implementing auto-scaling to handle varying user loads
• Use load balancing to distribute users across instances

4. Cost optimization strategies:

• Use a mix of instance types (on-demand, spot, reserved) based on your usage patterns
• Implement auto-scaling to add or remove instances based on demand
• Consider scheduling instances to shut down during periods of no traffic
• Monitor and optimize using AWS Cost Explorer and CloudWatch

For your specific Unreal Engine Pixel Streaming use case, you'll likely need to implement a solution that launches separate Unreal Engine processes for each user session, with proper resource allocation to ensure good performance for each user while maintaining cost efficiency.
Sources
Multi-Session Recommendations - Amazon WorkSpaces Applications
Optimizing Costs for EC2 in High-Traffic Scenarios? | AWS re:Post