Skip to main content

Mark Reed at ITS Research Computing, Mark Reed

Title: Research Associate, ITS Research Computing

Number of years with ITS: 11

Personal: Lives in Durham with his fiancé, Sandy. They have six grown children and three grandchildren.

Education: Ph.D., masters and bachelor’s degrees in physics from UNC-Chapel Hill


What does your job entail?

I manage the scientific engagement team here. We have a mix of a couple people with some program experience and people with Ph.Ds in different scientific fields as well. We’re a very strong and diverse staff of six people. And I make seven.

Our task is really to interact with the faculty and sort of help partner with them in using the using the resources – using both the compute resources – the clusters—and generally meeting their research needs whether it’s for large amount of storage or doing heavy amounts of computation or needing some type of specialized services.

We work with faculty, both in helping them get the models running and the resources in importing codes. We have a couple members who are kind of imbedded in different research groups. They work very closely with faculty on different research projects and co-author papers with them.

The faculty clearly lead the research, but we’re here to help them with all that we can.

Tell us a little bit about computational science.

Some look at it as the third leg of science, where you have traditional theory and traditional experiment. This is in some ways a melding of the two. You can use it for theoretical predictions and you can use it test out theories as well.

Building computational models is a large part of what many researchers do today. They develop these sophisticated mathematical models that solve a set of equations. They’ll use a cluster like KillDevil to do these because usually once you solve them, the model becomes quite large and complex. So you really want to run these at scale where you can run many thousands or tens of thousands of jobs, or you want to run a single job on many hundreds or thousands of cores at one time so you can quickly make progress on the problem at hand.

What excites you about your work?

Just getting to work with faculty and be part of the larger University system. I feel like we’re helping to achieve the University’s mission of research. We also work with people to do training for classes. I find both of those to be pretty fulfilling.

Are customer expectations changing?

I think we are seeing some changes in the type of computing people are doing. There’s always new people coming into what we loosely call research computing. We’re seeing a mix: high-performance computing, which people tend to think of as doing large parallel calculations, and then high-throughput computing, which is basically doing many thousands or tens of thousands of serial computations, serial jobs that are run on a single core or perhaps on a single node.

As you have new people come into the field, they may have more or less expertise in computational matters. We’re seeing more people coming in who are not as experienced in computing or maybe it doesn’t come as natural to them so we’re trying to expand and offer services to them as well. We’re seeing a need for lots of high-throughput type computing. Other people need large memory jobs. One of the areas that we will see in the future is computing on multi- and many core architectures.

Ultimately, researchers want to get their research done right. They don’t care so much as how it gets done. They want the results. Our job is to make it as seamless and as painless as possible.

What are some of the things you’re doing to make Research Computing a stronger enterprise?

We’ll be retiring the Kure cluster and so we’ll be looking for a replacement for that. We will talk to faculty members across the University to see what makes most sense to enhance capabilities of researchers here at UNC-Chapel Hill. We’ve talked to individuals. We haven’t reached any consensus. We don’t have a hard deadline. We’d like to replace it in the next year. Kure is now five years old. It goes back to 2010.

What’s the biggest challenge Research Computing faces?

Meeting the diverse needs of the many researchers because they all have very different needs and you have to balance what each one needs with our limited amount of resources and staff.


Comments are closed.