NEWTON – Cluster

Newton cluster is available to qualified faculty, students (under faculty supervision), and researchers from the Computer Science Faculty.

Access requests are handled through the CS GPU Cluster System.

Contents

• • NEWTON – Cluster
  • NEWTON – Status
  • NEWTON – Hardware Specs
  • DARWIN – Hardware Specs
  • NEWTON – Storage
  • NEWTON – Connection
  • DARWIN – Connection
  • NEWTON – How to launch jobs
  • NEWTON – ENROOT containers
  • NEWTON – Partitions, accounts and QoS
  • DARWIN – Partitions, accounts and QoS
  • NEWTON – FAQ

Introduction to the Newton Computing Cluster
The HPC environment consists of dedicated login nodes and a large number of compute nodes. Login nodes are intended only for accessing the system, preparing files, and submitting jobs. All development, testing, and computational workloads should be performed on the compute nodes.
Jobs are submitted through the Slurm scheduler, which automatically allocates the required resources and runs the job on available compute nodes.
The environment is divided into two complementary clusters:
Newton is a GPU-accelerated AI and machine learning cluster providing approximately 350 GPUs for deep learning, data science, and accelerated computing workloads.
Darwin is a CPU-focused high-performance computing cluster providing approximately 600 CPU cores (1,200 hardware threads) for serial, multithreaded, and distributed computing applications.
Both clusters share the same authentication system, user accounts, home directories, software environment, and shared storage infrastructure, allowing users to seamlessly access resources across the HPC platform.
Users may request both batch and interactive allocations on compute nodes. Home directories are intended for personal files and small-scale storage, while large datasets, checkpoints, and research data should be stored on the shared and Datasets volumes or designated laboratory storage areas.