Welcome to ASTRA-sim’s documentation!

ASTRA-sim is a distributed machine learning system simulator. It enables the systematic study of challenges in modern deep learning systems, allowing for the exploration of bottlenecks and the development of efficient methodologies for large DNN models across diverse future platforms.

Here is a concise visual summary of our simulator:

• Getting Started
  • Dependencies Setup
  • Build ASTRA-sim
  • Run ASTRA-sim
  • ASTRA-sim Output
• Workload Layer
  • Overview
• System Layer
  • Collective Scheduler
  • Collective Implementation
  • Input Files for Collective API
• Network Backend
  • Analytical Network Backend
  • ns-3 Network Backend
  • Garnet Network Backend
• Compute Backend
  • Scale-sim Compute Backend
  • Roofline Compute Backend
• Remote Memory Backend
  • Analytical Remote Memory Backend
• Validation
  • Validation on GPU Systems - NCCL over HPE ProLiant Gen10
  • Validation on GPU Systems - NCCL over HGX-H100 Systems
  • Validation on Google’s TPU-v3

Note

ASTRA-sim documentation project is currently under active development. Please stay tuned!