Getting Started

Overview of the NowWhat cluster layout, access model, and standard workflow.

This page gives a compact orientation to the cluster. Detailed procedures live on the dedicated access, storage, software, and SLURM pages. For browser-based access, follow the NowWhatHub User Guide.

System Layout

NowWhat currently consists of:

  • one login node
  • one CPU compute node
  • one GPU compute node

Login Node

  • 8 AMD EPYC 9654 cores
  • 32 logical cores
  • 64 GB RAM

CPU Node

  • 88 AMD EPYC 9654 cores
  • 352 logical cores
  • 704 GB RAM

GPU Node

  • 2 AMD EPYC 9135 processors
  • 1.5 TB RAM
  • 4 NVIDIA H200 MIG 1g.35gb instances, each with about 35 GiB GPU memory
  • 1 full NVIDIA H200 NVL GPU, about 141 GiB GPU memory

Access Model

NowWhat supports two access paths:

  • NowWhatHub starts a Slurm allocation from a web browser and provides JupyterLab, notebooks, RStudio, VS Code, a remote desktop, and terminals. See the NowWhatHub User Guide.
  • SSH connects to the login node for file management, job submission, and command-line workflows. The endpoint and first-login procedure are documented in Access and Login.

Compute nodes are used through SLURM allocations. Batch and interactive workflows are documented in Running Jobs with SLURM.

Standard SSH Workflow

The normal workflow on NowWhat is:

  1. connect to nowwhat.stat.unipd.it
  2. organize files according to Files and Storage
  3. prepare software according to Software
  4. submit a batch or interactive job through SLURM
  5. monitor the job with SLURM tools
  6. inspect logs, outputs, and resource usage
  7. clean up obsolete files

For the equivalent browser-based workflow, see How NowWhatHub works.