Hazel Quick Start Tutorial

This tutorial works through the steps for porting a workflow from a desktop computer to the NC State High-Performance Computing (HPC) cluster, called Hazel. You will be guided through the various steps using the current website documentation.

Here is a link the Basic HPC Workshop in video format.

Case scenario

You are developing R code with a faculty advisor. The calculation is getting so big that it ties up your desktop computer for days and sometimes crashes. Your advisor creates an HPC project and adds you to the project. You log in to Hazel and transfer your code from a desktop machine to Hazel. You run your R code using the batch scheduler. After making some changes to the code, you find that you need to do some interactive debugging. Many R functions can be run in parallel, so you take a break from R to learn the different types of parallel processing and how to submit those types of jobs. Next, you request a permanent place on Hazel to store your growing repository of R scripts - now shared with your research group - as well as mass storage space for simulation inputs and temporary outputs. Finally, you want to visualize the output in real time through a graphical user interface (GUI), so you request access to the HPC-VCL in order to visualize R output with GUI applications

Tutorial prerequisites

This HPC tutorial has no prerequisites. You will receive instructions on all steps, such as gaining access to Hazel and downloading necessary software. Each step in the tutorial does require that you have successfully completed the previous steps. Source code will be provided; you do not need to know R or any other programming language.

Tutorial - step by step

1. Gain access to Hazel through your project advisor.
   • A faculty member must create a project and add members to projects.
2. Log in to Hazel.
   • Learn how to log in via Windows, macOS, or Linux.
   • Learn how to log in via web browser
3. Transfer your code from a desktop machine to Hazel.
   • Learn file transfer methods for your OS.
4. Run R using the batch scheduler.
   • Check the Software webpage for R specific instructions and the scheduler (LSF) webpage for additional job scheduling options.
5. Run R interactively at the command line for a quick debugging session.
   • Running applications on a login node - even for a couple seconds - is prohibited, but you may reserve a short interactive session on a compute node for the purposes of debugging.
6. Run a few parallel programs.
   • Learn the different types of parallelism and how they affect the parameters in a job submission script.
7. Request a permanent place on Hazel to store and share your scripts, and request mass storage space to archive your simulation output.
   • Space in the home directory is limited, and scratch storage is not permanent.
8. Use Open onDemand or the HPC-VCL to run GUI-based applications and for interactive visualization (optional).
   • Using a remote display through X-windows is slow. Open onDemand and the HPC-VCL uses a different transfer protocol (RDP) that enables the use of GUIs with acceptable performance.