When to use ARC clusters¶
Is a cluster appropriate?¶
This page contains descriptions of kinds of work that can be done on a cluster. If any of them seem like they apply to your situation, please contact us, and we can discuss your situation, whether the cluster might be an appropriate tool, and how you might get started to see if it really is. Write us at arcts-support@umich.edu and we will set up a time to meet, by phone, by Zoom, in person, or just over e-mail.
What is a cluster, and when should I use one?¶
A cluster is, literally, a bunch of PCs connected by a fast network. What makes a cluster useful is the software that runs on it. Anything that you would consider running on more than one PC at the same time is a good candidate.
Why should you care?¶
You should care if your own computer is bogged down and unable to do more than one thing while something runs on it.
You should care if you are not considering research projects because your computer cannot handle the work needed.
You should care if you have many people working on the same project at the same time.
You should care if you have tens, hundreds, or thousands of analyses to run on similar data -- as in, many subjects, many samples, many locations -- and the operations or analysis is the same for them all.
When you should care¶
When you need more disk space
When you need more memory
When you need more computers
When you have many people who need to run the same thing repeatedly
Example scenarios¶
I have a statistical problem that needs to have more than 24 GB of memory to be able to do the computation.
I have a data collection that exceeds the disk capacity of my computer.
I have data from hundreds of subjects that needs to be processed by my students; they cannot all use the same computer, and I need to make sure each of them processes the data in exactly the same way as all the others.
I have 1,000 subjects, each of which takes eleven hours to process. I need to process them all in couple of weeks. I wish I had 100 computers!
I want to just start this analysis and check back on it in a couple of days to see how it is doing.