Why Software Modules?

Overview

Questions

• Why do we need environment modules on shared computing systems?
• What is an environment module and how does it work?
• How does Lmod modify your shell environment?

Objectives

• Understand the problem that environment modules solve
• Learn how Lmod modifies your shell environment
• Understand the difference between loading and installing software

The Problem: Conflicting Software Versions

---

Callout

The Version Conflict Problem

Imagine you’re working on Sagehen HPC with two research groups. One group needs R version 3.6.0 for legacy analysis scripts that haven’t been updated in five years. Another group requires R version 4.2.3 for their latest genomics pipeline. Your colleague needs Gaussian 16 to run quantum chemistry simulations, but you need Gaussian 09 for continuity with your previous research.

Without a way to manage these different versions, installing them all in the same locations (like /usr/bin or /usr/lib) creates conflicts: - Programs and libraries overwrite each other - Running R might load the wrong version - Dependencies break unpredictably - Users can’t work on different projects simultaneously

On a shared system with hundreds of users, this would be chaos.

Four commands cover almost everything you will do with modules.

The Solution: Environment Modules

---

Environment modules (managed by Lmod on Sagehen) solve this problem by dynamically modifying your shell environment. A module is essentially a collection of instructions that:

1. Adds software locations to your PATH (where the system looks for executable programs)
2. Sets LD_LIBRARY_PATH (where the system looks for shared libraries)
3. Sets other environment variables needed by the software
4. Possibly loads prerequisite modules

When you load a module, it’s like installing software temporarily in your current shell session without actually changing the filesystem.

How Lmod Modifies Your Environment

---

Let’s look at a practical example. When you load the R module:

BASH

$ module load r

Lmod runs a script that: - Adds the R installation’s bin directory (e.g. /opt/.../r/4.5.1/bin) to your PATH - Sets LD_LIBRARY_PATH to include R’s libraries - Sets R_HOME to the installation directory - Sets other R-specific variables

Now when you type R at the command line, the shell finds the R executable in the path that was just added.

If you then load a different version:

BASH

$ module swap r r/4.2.3

Lmod removes the 4.5.1 entries and adds the 4.2.3 entries instead. No files are actually installed or removed; just your shell environment changes.

Module Naming Conventions

---

Modules follow the pattern software/version:

• r/4.5.1: R version 4.5.1
• r/4.2.3: R version 4.2.3
• r: Same as the default, currently r/4.5.1 (marked with D)
• miniconda3: The default version of miniconda3
• miniconda3/py313_26.3.2-2: Specific version
• cuda/12.2.1: NVIDIA CUDA toolkit version 12.2.1

When you run module load r, you get the default. To load a specific version, be explicit.

Challenge

Challenge 1: Explore Available Modules

Connect to Sagehen and run module avail. Scroll through the output and identify:

1. How many versions of R are available?
2. What CUDA toolkit versions are available?
3. Which modules have a default version marked with (D)?

BASH

$ ssh <myusername>@sagehen.hpc.pomona.edu
$ module avail

Show me the solution

Output will vary, but a typical response shows:

R versions:
- r/4.2.2
- r/4.2.3
- r/4.4.1
- r/4.5.1 (D)

CUDA toolkit versions:
- cuda/11.8.0
- cuda/12.0.0
- cuda/12.2.1 (D)

Modules with defaults:
- miniconda3/py313_26.3.2-2 (D)
- r/4.5.1 (D)
- go/1.23.1 (D)

The important point is that you can see all available options before loading.

Key Points

• Environment modules solve the problem of multiple software versions on shared systems
• Modules temporarily modify your shell environment (PATH, LD_LIBRARY_PATH, etc.)
• No files are installed or removed; only your environment changes
• Modules follow the naming convention software/version
• Modules with (D) are defaults; specify versions explicitly to get other versions