In this tutorial, we talk about the overall process of finding, building, and compiling software. These days, installing and managing scientific software is taking more and more time, thus we need to specifically talk about it.
Clusters, being a shared systems, has more complicated software requirements. In this tutorial, you will learn how to use existing software. Be aware that installing your own is possible (and people do it all the time), but does require some attention to details. Either way, you will need to know the basics of software on Linux.
There are many ways to get software on Triton: we use the standard module system, have Singularity containers, and you can install your own.
Modules allow you to activate a lot of software.
Singularity containers allow you to run other hard-to-install software.
Ask us to install software if you need it. You can also install software yourself, but you need to update instructions to do a user install (as opposed to admin install).
Main article: Applications: General info
Almost every site will use modules. The exact module names, and anything beyond that, will be different. Containers are becoming more common, but they are less standardized.
How to find the software you need¶
You can find what softwares we have available in different ways:
First, you should check our Applications page and see if the software you need is already available and has instructions.
If you find the software you need available, you can usually load it via a module. The next tutorial, Software modules explains what modules are and how to work with them.
You can also search this tutorial to see what you can find (though note that not everything is in the Triton section here - some applies to Aalto workstations or own computers).
It’s always a good idea to search the issue tracker to see if there are previous issues about it - not everything is always updated.
Ask other users in the Zulip chat. We hope that we can facilitate user group meetings and discussion among users of similar software suites.
Ask us admins/Research Software Engineers in garage.
Throughout this process, try to remember these things:
Scientific software, like scientific process itself, is collaborative. Work on sharing and seeking knowledge among other users. They might have the answer you need.
Interesting problems draw people together independently. If you’re working on a certain type of a problem, it is quite likely that some other researcher is working on a similar problem. You might not be alone with your problem.
Try to form connections between users of similar software. The same software that you use can be used by a researcher in completely different field. Many software suites e.g. statistical modelling, machine learning, is common to many other fields. If you cannot find similar users within your field, look across fields.
If you find something useful or interesting, share it. If you do not know who to share it with, share it with us in SciComp. When we hear of a tool, a method, a success story or a problem encountered by one of our users, we often try to share it among other researchers.
Common applications are available as modules¶
This is Aalto-specific. Some of these will work if you
load fgci-common at other Finnish sites (but not CSC). This is
introduced in the next lesson.
Here is a sample of our most commonly used software:
module load anacondafor the Anaconda distribution of Python 3, including a lot of useful packages. More info.
module load rfor a basic R package. More info.
module load matlabfor the latest Matlab version. More info.
module load juliafor the latest Julia version. More info.
If one of these
module load commands does not work at your site,
module spider NAME and see if you can find it. More information
on these commands will be
actually covered under the upcoming modules tutorial.
We try to install commonly used software for all of our users, so that everyone can benefit from them. If you cannot find what you’re looking for, do let us know.
Main article: Singularity Containers
Some software packages are either very complicated to install or they have been designed with certain operating systems in mind. For these kinds of software we often use containers. A software container is basically a complete self-contained operating system environment. Another advantage of containers is that they make it easy to move installed software from system to system, so that you can have the same environment everywhere.
If your program is usually deployed using Docker or it is hard to maintain, do read our documentation on Singularity containers and contact us for more information.
We also provide some containers built by NVIDIA. These containers are from NVIDIA’s NGC-repository and meant for GPU computations.
Requesting new software¶
We aim to install a good base of software for our users - but it’s not possible to keep up with all requests. If you need something, submit a request to our issue tracker, but be aware that despite best efforts, we can’t do everything. See the main Applications page for more information.
Not everyone has to, but many people either write there own software or writ scripts to automate the running and analysis. Yet, these skills are often not developed as well as they should be. Contact the Research Software Engineers (part of Science-IT) for help here - basic service is free.
A plea: make your software reusable!¶
Five years from now, when you are releasing your own software that you want others to use, make it easy to install and reusable.
If you are at Aalto, everything will work. Otherwise, if you are in
Finland (but not at CSC)
module load fgci-common will make our
modules available on your cluster.
Applications-1: Find Tensorflow documentation
This is an exercise in finding out how to use a new program on Triton. The goal is not to actually use tensorflow, but practice searching for information on it.
Figure out how to use
tensorflow: What is the documentation
page on scicomp.aalto.fi? Can you see any issues where others have
asked for special support?
If you want to use Python, see if you can import it. For example,
try to run
python and then in Python run
Applications-2: Check your needs
Find the Applications page link above, and check the list for ways to find out if we already have your software installed. See if we have what you need, using any of the strategies on that list.
(optional) Applications-3: Spack package list
From the Applications page, find the Spack package list (warning: it’s a very long page and takes a while to load). Does it have anything useful to you?
(optional) Applications-4: Your group’s needs
Discuss among your group what software you need, if it’s available, and how you might get it. Can they tell you how to get started?
The next tutorial covers software modules in more detail.