Introduction: The workshop session will provide a quick tour covering high-level concepts, commands and processes for using Linux and HPC on our Beartooth cluster. It will cover enough to allow an attendee to access the cluster and to perform analysis associated with this workshop.

Goals:

• Introduce ARCC and what types of services we provide including “what is HPC?”

• Define “what is a cluster”, and how is it made of partitions and compute nodes.

• How to access and start using ARCC’s Beartooth cluster - using our SouthPass service.

• How to start an interactive desktop and open a terminal to use Linux commands within.

• Introduce the basics of Linux, the command-line, and how its File System looks on Beartooth.

• Introduce Linux commands to allow navigation and file/folder manipulation.

• Introduce Linux commands to allow text files to be searched and manipulated.

• Introduce using a command-line text-editor and an alternative GUI based application.

• How to setup a Linux environment to use R(/Python) and start RStudio, by loading modules.

• How to start interactive sessions to run on a compute node, to allow computation, requesting appropriate resources.

• How to put elements together to construct a workflow that can be submitted as a job to the cluster, which can then be monitored.

0 Getting Started

• Users may log in with their BYODs (do you have a computer with you to follow along with the workshop?)

  • Log into UWYO wifi if you can. (Non-UW users will be unable to).

• Logging in:

  • If you have a UWYO username and password: UW Users may test their HPC access by opening a browser and then going to the following URL: https://southpass.arcc.uwyo.edu.

  • Standard wyologin page will be presented. Log in with your

  • UWYO username and password.

  • If you do not have a UWYO username and password: Come see me for a Yubikey and directions allow you to access the Beartooth HPC cluster if you do not have a UW account.
    
    

00 Introduction and Setting the Scope:

The roadmap to becoming a proficient HPC user can be long, complicated, and varies depending on the user. There are a large number of concepts to cover. Some of these concepts are included in today’s training but given time constraints, it’s impossible to get to all of them. This workshop session introduces key high-level concepts, and follows a very hands-on demonstration approach, for you to follow.

Our training will help provide the foundation necessary for you to use Beartooth cluster, specifically to perform some of the exercises later in this workshop over the week.

Because of our limited time this morning, please submit any questions to the slack channel for this workshop and workshop instructors can address them as they are available.

More extensive and in-depth information and walkthroughs are available on our wiki and under workshops/tutorials. You are welcome to dive into those in your own time. Content within them should provide you with a lot of the foundational concepts you would need to be familiar with to become a proficient HPC user.

01 About UW ARCC and HPC

Goals:

• Describe ARCC’s role at UW.

• Provide resources for ARCC Researchers to seek help.

• Introduce staff members, including those available throughout the workshop.

• Introduce the concept of an HPC cluster, it’s architecture and when to use one.

• Introduce the Beartooth HPC architecture, hardware, and partitions.

About ARCC and how to reach us

Based on: Wiki Front Page: About ARCC

• In short, we maintain internally housed scientific resources including more than one HPC Cluster, data storage, and several research computing servers and resources.

• We are here to assist UW researchers like yourself with your research computing needs.

3 ARCC Staff Members will be available through the course of the workshop if you need help using Beartooth:

ARCC End User Support

ARCC End User Support
Simon Alexander	HPC & Research Software Manager	Open
Dylan Perkins	Research Computing Facilitator	Open
Lisa Stafford	Research Computing Facilitator

What is HPC

HPC stands for High Performance Computing and is one of UW ARCC’s core services. HPC is the practice of aggregating computing power in a way that delivers a much higher performance than one could get out of a typical desktop or workstation. HPC is commonly used to solve large problems, and has some common use cases:

1. Performing computation-intensive analyses on large datasets: MB/GB/TB in a single or many files, computations requiring RAM in excess of what is available on a single workstation, or analysis performed across multiple CPUs (cores) or GPUs.

2. Performing long, large-scale simulations: Hours, days, weeks, spread across multiple nodes each using multiple cores.

3. Running repetitive tasks in parallel: 10s/100s/1000s of small short tasks.

What is a Compute Node?

• We typically have multiple users independently running jobs concurrently across compute nodes.

• Resources are shared, but do not interfere with any one else’s resources.

  • i.e. you have your own cores, your own block of memory.

• If someone else’s job fails it does NOT affect yours.

• Example: The two GPU compute nodes part of this reservation each have 8 GPU devices. We can have different, individual jobs run on each of these compute nodes, without effecting each other.

Homogeneous vs Heterogeneous HPCs

There are 2 types of HPC systems:

1. Homogeneous: All compute nodes in the system share the same architecture. CPU, memory, and storage are the same across the system. (Ex: NWSC’s Derecho)

2. Heterogeneous: The compute nodes in the system can vary architecturally with respect to CPU, memory, even storage, and whether they have GPUs or not. Usually, the nodes are grouped in partitions. Beartooth is a heterogeneous cluster and our partitions are described on the Beartooth Hardware Summary Table on our ARCC Wiki.

Beartooth Cluster: Heterogeneous: Partitions

Beartooth Hardware and Partitions

See Beartooth Hardware Summary Table on the ARCC Wiki.

Reservation

A reservation can be considered a temporary partition.

It is a set of compute nodes reserved for a period of time for a set of users/projects, who get priority use.

For this workshop we will be using the following: biocompworkshop:

Important Dates:

1. After the 17th of June this reservation will stop and you will drop down to general usage if you have another Beartooth project.

2. The project itself will be removed after the 24th of June. You will not be able to use/access it. Anything you require please copy out of the project.

02 Using Southpass to access the Beartooth HPC Cluster

Southpass is our Open OnDemand resource allowing users to access Beartooth over a web-based portal. Learn more about Southpass here.

Goals:

• Demonstrate how users log into Southpass

• Demonstrate requesting and using a XFCE Desktop Session

• Introduce the Linux File System and how it compares to common workstation environments

  • Introduce HPC specific directories and how they’re used

  • Introduce Beartooth specific directories and how they’re used

• Demonstrate how to access files using the Beartooth File Browsing Application

• Demonstrate the use of emacs, available as a GUI based text-editor

Based on: https://arccwiki.atlassian.net/wiki/spaces/DOCUMENTAT/pages/1298071553

Log in and Access the Cluster

Login to Southpass

Using Southpass

Exercise: Beartooth XFCE Desktop

Requests are made through a webform in which you specifically request certain hardware or software to use on Beartooth.

Note: While we use a webform to request Beartooth resources on Southpass, later training will show how resource configurations can be requested through command line via salloc or sbatch commands.

Structure of the Linux File System and HPC Directories

Linux File Structure

This is specific to the Beartooth HPC but most Linux environments will look very similar

Linux Operating Systems (Generally)

Compare and Contrast: Linux, HPC Specific, Beartooth Specific

Based on: https://arccwiki.atlassian.net/wiki/spaces/DOCUMENTAT/pages/1714913281

HPC Specific Folders:

1. /home (Common across most shared HPC Resources)

   1. What is it for? Similar to on a PC, and Macintosh HD → Users on a Mac

   2. Permissions: It should have files specific to you, personally, as the HPC user. By default no one else has access to your files in your home.

   3. Director Path: Every HPC user on Beartooth has a folder in on Beartooth under /home/<your_username> or $HOME

   4. Default Quota: 25GB

2. /project (Common across most shared HPC Resources)

   1. What is it for? Think of it as a shared folder for you and all your project members. Similar to /glade/campaign on NCAR HPC.

   2. Permissions: All project members have access to the folder. By default, all project members can read any files or folders within, and can write in the main project directory.

   3. Directory path: get to it at /project/biocompworkshop/

   4. Subfolders in /project/biocompworkshop/ for each user are added to project when a user gets added to the project, but only that user can write to their folder.

   5. Default Quota: 1TB which is for the project folder itself and includes all it’s contents and subfolders.

3. /gscratch (Scratch folder, common across most HPC resources but sometimes just called "scratch")

   1. What is it for? It’s “scratch space”, so it’s storage dedicated for you to store temporary data you need access to.

   2. Permissions: Like /home, contents is specific to you, personally, as the HPC user. By default no one else has access to your files in your /gscratch.

   3. Director Path: Every HPC user on Beartooth has a gscratch directory in Beartooth under /gscratch/<your_username> or $SCRATCH

   4. Default Quota: 5TB

      1. Don’t store anything in /gscratch that you need or don't have backed up elsewhere. It's not meant to store anything long term.

      2. Everyone’s /gscratch directory is subject to ARCC's purge policy.

Beartooth Specific

4. /apps (Specific to ARCC HPC) is like on Windows or on a Mac.

   1. Where applications are installed and where modules are loaded from. (More on that later).

5. /alcova (Specific to ARCC HPC).

   1. Additional research storage for research projects that may not require HPC but is accessible from beartooth.

   2. You won’t have access to it unless you were added to an alcova project by the PI.

Exercise: File Browsing in Southpass GUI

Users can access their files using the south pass file browser app.

Demonstration opening emacs GUI based text editor

03 Using Linux and the Command Line

Goals:

• Introduce the shell terminal and command line interface

  • Demonstrate starting a Beartooth SSH shell using Southpass

  • Demonstrate information provided in a command prompt

• Introduce Policy for HPC Login Nodes

• Demonstrate how to navigate the file system to create and remove files and folders using command line interface (CLI)

  • mkdir, cd, ls, mv, cp

• Demonstrate the use of man, --help and identify when these should be used

• Demonstrate using a command-line text editor, vi

Based on: https://arccwiki.atlassian.net/wiki/spaces/DOCUMENTAT/pages/1596194853

Exercise: Shell Terminal Introducing Command Line

What am I Using?

Remember:

• The Beartooth Shell Access opens up a new browser tab that is running on a login node. Do not run any computation on these.
  [<username>@blog2 ~]$

• The SouthPass Interactive Desktop (terminal) is already running on a compute node.
  [<username>@t402 ~]$

Login Node Policy

As a courtesy to your colleagues, please do not run the following on any login nodes:  

1. Anything compute-intensive (tasks using significant computational/hardware resources - Ex: using 100% cluster CPU)

2. Long running tasks (over 10 min)

3. Any collection of a large # of tasks resulting in a similar hardware footprint to actions mentioned previously.  

4. Not sure?  Use salloc to be on the safe side.  This will be covered later.
   Ex: salloc –-account=arccanetrain -–time 40:00

5. See more on ARCC’s Login Node Policy here

Demonstrating how to get help in CLI

Demonstrating file navigation in CLI

Demonstrating how to create and remove files and folders using CLI

Text Editor Cheatsheets

Note: On Beartooth, vi maps to vim i.e. if you open vi, you're actually starting vim.

Demonstrating vi/vim text editor

Try the vim tutor

04 Using Linux to Search/Parse Text Files

Goals:

• Using the command-line, demonstrate how to search and parse text files.

• Show how export can be used to setup environment variables and echo to see what values they store.

• Linux Commands:

  • find

  • cat / head / tail / grep

  • sort / uniq

  • Pipe | output from one command to the input of another, and redirect to a file using >, >>.

Based on: Intro to Linux Command-Line: View Find and Search Files

Your Environment: Echo and Export

Use Our Environment Variable

Search for a File

Based on: Search for a File

Use Wildcards *

View the Contents of a File

Based on: View/Search a File

View the Start and End of a File

Search the Contents of a Text File

Grep-ing with Case-Insensitive and Line Numbers

Pipe: Count, Sort

Based on: Output Redirection and Pipes

Uniq

Redirect Output into a File

05 Lets start using R(/Python) and RStudio

Goals:

• Using a terminal (via an Interactive Desktop), demonstrate how to load modules to setup an environment that uses R/RStudio and how to start the GUI.

• Mention how the module system will be used, in later workshops, to load other software applications.

• (Indicate how this relates to setting up environment variables behind the scenes.)

• Further explain the differences between using a login node that requires an salloc to access a compute node, and that you're already running on a compute node (with limited resources) via an interactive desktop.

  • Confirm arguments for partition, gres/gpu, reservation.

  • Note that can confirm a GPU device is available by running nvidia-smi -L from the command-line.

• Show how the resources from the Interactive Desktop configuration start mapping to those used by salloc (including defining reservations, and maybe partitions).

Based on Intro to Accessing the Cluster and the Module System

Open a Terminal

You can access a Linux terminal from SouthPass by:

• Opening up an Interactive Desktop (reservation is biocompworkshop) and opening a terminal.

  • Running on a compute node: Command prompt: [<username>@t402 ~]$

  • The reservation is only available for this workshop: StartTime=06.09-09:00:00 EndTime=06.17-17:00:00 Duration=8-08:00:0

  • Only select what you require:

    • How many hours? Your session will NOT run any longer that the amount of hours you requested.

    • Some Desktop Configurations will NOT work with some GPU Types.

    • Do you actually need a GPU?

      • Unless you software/library/package has been developed to utilize a GPU, simply selected one will NOT make any difference - this won’t make you code magically run faster.

• Selecting a Beartooth Shell Access which opens up a new browser tab.

  • Running on the login node: [<username>@blog1/2 ~]$

Setting Up a Session Environment

Across the week, you’ll be using a number of different environments.

• Running specific software applications.

• Programming with R and using various R libraries.

• Programming with Python and using various Python packages.

• Environments build with Miniconda - a package/environment manager.

Since the cluster has to cater for everyone we can not provide a simple desktop environment that provides everything.

Instead we provide modules that a user will load that configures their environment for their particular needs within a session.

Loading a module configures various environment variables within that Session.

What is Available?

We have environments available based on compilers, Singularity containers, Conda, Linux Binaries

Is Python and/or R available?

Load a Compiler

Load a Newer Version of Python

Typically Loading R

Load R/RStudio for this Workshop

You can use module purge to reset your environment, or start a new terminal

Configure your R Environment for this Workshop

Request Interactive Session (Compute Node) from a Login Node

Request Interactive Session (Compute Node) with a GPU

Request what you Need!

06 Create a basic workflow and submitting jobs.

• Since RStudio is a GUI, demonstrate moving from running a script within RStudio to running using Rscript from the command-line.

• Put the various elements of loading modules, moving into a folder, running an R file, that make up a basic workflow, into a script that can be submitted using sbatch to Slurm.

• Map the salloc arguments to #SBATCH.

• Show how to monitor a jobs using squeue as well as using the email related Slurm options.

• Show how to request the DGX nodes and defining gres to specifically request a GPU.

• Provide a basic template.

Based on:

• Intro to Job Scheduling

• Slurm Workload Manager

Why Submit a Job

A single computation can take, minutes, hours, days, weeks, months. An interactive session quickly becomes impractical.

Submit a job to the Slurm queue - Slurm manages everything for you.

Everything you do on the command-line, working out your workflow, is put into a script.

Workflow:

• What resources you require? (Interactive desktop configuration, salloc options)

• What modules are loaded.

• Which folder you’re running you computation within. Where the data is stored. Where you want the results.

• Command-line calls being called.

• Software applications being run.

Submit a Job to the Cluster

Convert salloc command-line options to an sbatch related script.

Options have defaults if not defined.

Additional sbatch Options

Example Script: What Goes into It?

The bash script can contain:

• Linux/bash commands and script.

• Module loads.

• Application command-line calls.

 Lets consider our R workflow. I have:

• R scripts copied into my /gscratch folder.

• R related modules to load.

• R scripts to run.

• to track the time the job starts and ends.

Example Script: Running R Script

Submit your Job

Monitor your Job

Why is my Job Not Running?

Previously we explained: The two GPU compute nodes part of this reservation each have 8 GPU devices. We can have different, individual jobs run on each of these compute nodes, without effecting each other.

So, we can have 16 concurrent jobs all running with a single GPU each.

But, what if a 17th person submitted a similar job?

Slurm will add this job to the queue, but it will be PENDING (P) while it waits for the necessary resources to become available.

As soon as there are, this 17th job will start, and it’s status will update to RUNNING (R).

Slurm manages this for you.

Monitor your Job: Continued…

Alternative Monitoring of Job via Email: Job Efficiency

Example Script 2

This might look like something your cover in later sessions:

Examples and Cheat Sheets

Can be copied from: /project/biocompworkshop/arcc_notes

07 Summary and Next Steps

• We’ve covered the following high-level concepts, commands and processes:

• What is HPC and what is a cluster - focusing on ARC’s Beartooth cluster.

• An introduction to Linux and its File System, and how to navigate around using an Interactive Desktop and/or using the command-line.

• Linux command-line commands to view, search, parse, sort text files.

• How to pipe the output of one command to the input of another, and how to redirect output to a file.

• Using vim as a command-line text editor and/or emacs as a GUI within an Interactive Desktop.

• Setting up your environment (using modules) to provide R/Python environments, and other software applications.

• Accessing compute nodes via a SouthPass Interactive Desktop, and requesting different resources (cores, memory, GPUs).

• Requesting interactive sessions (from a login node) using salloc.

• Setting up a workflow, within a script, that can then be submitted to the Slurm queue using sbatch, and how to monitor jobs.

Further Assistance:

• Everything covered can be found in previous workshops and additional information can be found on our Wiki.

• ARCC personnel will be around in-person for the first three days to assist with cluster/Linux related questions and issues.

• We will provide virtual support over Thursday/Friday. Submit questions via the Slack channel and these will be passed onto us, and we will endeavor to set up a zoom via our Office Hours.