1

I am familiarizing myself with conda and linux environment modules (lems henceforth), largely for the purpose of enabling reproducible research. I know that conda is more geared towards setting up user-level environments compatible across many platforms (HPC, laptops, etc), where as lems are generally geared towards HPC and usually deployed system-wide.

However, what is unclear is whether I can use conda to largely replace lems, given the caveat that when using conda many binaries will be duplicated if each user manages their own conda stack.

So, from a user's perspective, are there any distinct limitations to conda or lems that would require learning how to use both?

Vince
  • 3,325
  • 2
  • 23
  • 41
  • 2
    One HPC I use is moving away from modules and using [singularity containers](https://en.wikipedia.org/wiki/Singularity_(software)) instead. Containers are much more reproducible, though there is a learning curve. – jkr Mar 24 '21 at 15:32
  • Yup. The idea we have here is also that, specifically to define environment using conda and put conda within docker/singularity. Issue I see is whether we still need linux modules, as this can still be used within docker/singularity to load specific versions of gcc, for example. Or just use sigularity/docker to install the specific version of whatever conda can't handle? Clearly, there are many ways to approach this... – Vince Mar 24 '21 at 17:26

0 Answers0