Hi there! My group at Harvard, under Academic Technology, is interested in submitting a proposal for the GOOD conference for a Birds of a Feather session on the use of OOD in a classroom or course setting, which is what we do for our community. The nature of a BoF presentation seems like it would benefit from a more diverse panel from other institutions, so I’m posting here to see if anyone else is interested in talking about how they’re working with OOD in classrooms, and the different needs for application setups and software that can come along with it.
For context, we’ve been running a cluster based in AWS Parallel Cluster since spring of 2024, and have moved from an on-prem environment managed by a research computing group (with all of the pre-existing software that entails) to a new, blank slate that we’ve had to fill up with stuff that our instructors and their teaching staff need. OOD just about the only way into this cluster, and we’ve been doing a lot of work figuring out how best to configure our apps so that our users are happy and we stay (somewhat) sane.
If anyone here is interested in collaborating, please let me know! Also, if there’s a better venue for this kind of discussion somewhere else, I’m happy to move it there.
Hi Jeremy, I’d be up for participating in a discussion on this. We have a setup that works fairly well, more-less doing what I presume other sites do as well - auto-creation of student accounts in our HPC space via a token, tied to their campus AD identity, confined to a class unix group and class SLURM account, and then having class-specific OOD apps that have very simple job submission form to minimize chances for mistake when submitting the job.Happy to share that and hear what others may be doing.
Hi Martin! That does sound pretty similar to what we do, although I’m interested in how things differ. For instance, our local IAM group hooked us up with an LDAP server, and we manage access through Grouper groups. We’ve also had two different user setups: In our current setup, users have one account, regardless of how many classes that use the service they’re enrolled in. We’ve also had user setups where users have separate accounts for each class they’re in. We also just implemented sub-apps to create different versions of common applications, like Jupyter Lab configurations with different Python dependencies, which has been great at simplifying our setup process and app repos.
Hopefully there’s more folks interested in those nuances! We’ll have to put together a proposal by Dec. 9 to submit for consideration. My current thought is to have different groups give short presentations on how they make OOD apps available to students, and then open it up to discussion (with some prepared questions for discussion). Open to revision and working together on the proposal too!
your thoughts on the BoF format are close to mine. I think we can solicit a handful of more sites for this, I can advertise this at the Tips and Tricks calls. We’ve had a discussion on this in the past but it was a while ago and not that many sites participated. Being in person would help more people to speak out, and exchange ideas.
How does this title sound: “Using OpenOnDemand in classroom”, with the contents being, that in the first part several sites will go over their OOD classroom support followed by discussion on best practices. We could also try to create some kind of working group to help to create/promote best practices and support sites that are new to this.
Though I suspect there may be differences in how sites implement various aspects of this so perhaps a general approach may not work the best, but at least it may be worth discussing.
I would also be interested in participating in your BoF. I work at Wake Forest University and we are going to put OOD into classrooms (for the first time) this upcoming Spring semester. This first round of courses will include Computer Science and Physics offerings. WFU is a private liberal arts school, but we have a robust centralized HPC system that serves 15 departments across campus.
Obviously, I will not be able to gather a ton of data by the March conference, but I would be happy to share any metrics, experiences, and perspectives on this process.