Skip to content

Latest commit

 

History

History
78 lines (54 loc) · 2.93 KB

cost-and-scaling.md

File metadata and controls

78 lines (54 loc) · 2.93 KB

EC2 Cost and Scaling Considerations for OpenSpace WebRTC

This document provides guidance for selecting AWS EC2 instances to run OpenSpace WebRTC efficiently. It covers minimum requirements, recommended instance types, scaling strategies, and estimated costs.

Minimum Requirements

Component Minimum Requirement
CPU Intel i5 equivalent or better
GPU NVIDIA GTX 1060 or comparable
RAM 8 GB
VRAM 4 GB
Software Size ~10 GB (including sync folder)
OS Windows
Instances per Machine Multiple OpenSpace instances supported

Recommended EC2 Instances

Instance Type GPU vCPUs RAM VRAM Max OS Instances Hourly Price Notes
G4dn.xlarge NVIDIA T4 (16GB) 4 16 GB 16 GB 1-2 ~$0.76 Best for MVP
G3s.xlarge NVIDIA M60 (8GB) 4 30.5 GB 8 GB 2-3 ~$0.97 Older but supports more instances
G5.xlarge NVIDIA A10G (24GB) 4 16 GB 24 GB 3-4 ~$1.20 Overkill but future-proof
G5.2xlarge NVIDIA A10G (24GB) 8 32 GB 24 GB 5-6 ~$2.40 Best for production

Legend:

  • MVP: Minimal Viable Product (2 developers, limited instances).
  • G3: Older GPU generation for graphics-heavy tasks.
  • G4: NVIDIA T4 GPUs, balanced performance.
  • G5: Latest NVIDIA A10G GPUs, high performance for advanced graphics.

Monthly Cost Estimation

Assuming development/testing for ~6 hours/day:

Instance Type Monthly Cost (24/7) Monthly Cost (~6 hr/day)
G4dn.xlarge (1-2 instances) ~$547.20 ~$136.80
G3s.xlarge (2-3 instances) ~$700.80 ~$175.20
G5.xlarge (3-4 instances) ~$864.00 ~$216.00
G5.2xlarge (5-6 instances) ~$1,728.00 ~$432.00

For pricing details, see AWS EC2 On-Demand Pricing

Recommendation Based on Use Case

  • MVP (2 Developers, Limited Instances) → G4dn.xlarge
    Runs 1-2 OpenSpace instances smoothly, best price/performance balance.

  • Scaling Beyond MVP (More Users, Heavier Load) → G3s.xlarge or G5.xlarge

    • G3s.xlarge: Supports 2-3 instances at moderate cost.
    • G5.xlarge: Handles 3-4 instances, better for larger tests.

  • Future Production Deployment (Multiple Concurrent Users) → G5.2xlarge
    Handles 5-6 instances, future-proof with 8 vCPUs & 32GB RAM.

Scaling Strategy

  • Launch additional EC2 instances when CPU usage exceeds ~75%.
  • Scale down instances when CPU usage drops below ~75%.
  • This ensures OpenSpace is available continuously while optimizing cost.
  • Scaling can be managed via AWS Auto Scaling Groups (ASGs).

Developer Suggestion:
For initial development, use G4dn.xlarge. Upgrade to G5.xlarge or G5.2xlarge as needed for larger workloads or production use.