Hello, everyone and welcome. My name is uh S Clar. I uh have 20 minutes and 21 slides. I tend to speak pretty fast. Uh So I'm gonna try to get through all the slides in less than 20 minutes and then pop out for some Q and A.Uh But if that doesn't happen and if I do take up all 20 minutes, uh my linkedin information is at the bottom of the slide, feel free to reach out to me afterwards. I'm happy to answer any questions. So uh my talk today is about high performance computing and research support at a liberal arts school. And I just gave this talk a few weeks ago at something called Virtual Residence uh where it was a group of about 600 people, um researchers and uh people that support researchers uh at uh academic institutions across the United States. And I added this uh qualifier for this slide deck because I feel like this is the audience that can appreciate my experience. And uh I uh would like to share it with you and then maybe have a conversation about how we can change it. So a little bit about me. I've been in it, uh, for more than 20 years in various roles since, uh, 95. Um, I've worked at start ups corporation or large corporations. Uh, I've done consulting and I am now working in a higher ed institution for about 2.5 years.

Pomona College is located in Claremont, California. The picture on the bottom is the art installation right outside the IT department, uh, by James Tyrrell who is an alum of Pomona College. It's called Sky Space and it changes colors at night. Uh When the sun sets, I'm the director of high performance computing, which is a really niche job. Basically, I like to explain it as when you need more than one computer. That's probably when you need to talk to me. Uh Pomona College is a little bit unusual uh because it's an undergraduate school. But uh that also does research. Uh So I support both research and teaching. Sometimes they overlap. I also founded the technology exploration lab uh to learn more about uh HPC at Pamona. You can go to H BC dot Tamon ad U and uh I'm also trying to be active on our Twitter account, Pamona HPC. So uh tweet at me, uh the painting that you see is of me some years ago uh made by someone who didn't know me very well. That is more likely to be beer. And uh uh also I don't quite look so casual. I'm usually a lot more intense. Um And uh I'm holding a graphics card in my hands because that's the workforce of high performance computing these days. That's happens to be a 1080 T that I'm installing into an external GP U.

Uh So part of my talk is going to be technical and part of it is going to be a reflection. And I'm kind of curious to hear your reactions about how this is gonna go. So this is my path to where I am today. I did not have any access to technology of any sort till about 17 years old. And uh I don't count the construction director set and a Sony tape and then CD player um as technology. Uh but we had no car. Um and I didn't have access to any electronics or any um anything technical really. And then at around 17 years old, I got thrown into uh a project to help out a friend of my dad's to assemble a network um star topology back then. Uh work with databases like B three databases and uh come up with some web presence. This is the era of uh redhead four and uh D OS 6.2 maybe some of you in the audience uh relate to those words. Um And this seems complicated and uh this was an unusual situation, everybody around me was too busy with their real job. So this was kind of a toy. And so I built all of this without really knowing how much usually goes into learning about it. And then I migrated between continents because uh in the nineties, there was a start up boom and the US lacked it talent. Um And I came to the United States uh on Parole visa. Uh It, it's a really curious thing. No one really knows about it.

If you Google it, you find the more explanation. Uh And I didn't really notice the uh uh start up crash because I've always had skills that were in demand. I've never been without a job. Fortunately, um I decided to stay in it despite the constant distrust and judgment because I enjoy solving difficult challenges. And um I just had a conversation with my partner the other day where he tried to convince me I'm successful, but I still find myself thinking that I don't belong or I'm not sure I belong a little bit about my other identities. Um I'm an exceed campus champion. Uh exceed is uh an infrastructure shared uh among academic institutions uh that provides access to more computing and storage capacity than you have access to. Um And uh there's a lot involved in that education efforts. Um And there's also something called empower program that uh peers uh mentors and mentees, not necessarily the same institution. So I've done that. Uh There is a conference that we get together at every year and share our experiences uh practice and experience, advanced research, computing.

I mentored at the conference as well. And uh I'm still in conversations with the uh my mmt from uh Chicago area. Last year, Pomono College hosted Snap The Gap event, which was meant to bring uh 10 to 20 girls. We actually ended up uh hosting a group of 20 into STEM and Steam and change their mind about uh being engineers. Um Majority of them came in not thinking that they can be in any new technical role. But I think we succeeded after about 10 sessions um of mentoring them that it's possible and we got them excited and that was very rewarding. I participate in big data and data science meetups because uh it's a constant source of education. I never think that I know nearly enough. Um I mentor my coworkers who are curious about what I do and uh want to learn and help. I also help artists in Southern California learn about technology and how it applies to them. And uh there's a conference called Southern California, Linux Expo. I spoke at that conference as well. Uh So what you see here in the pictures is uh Zintan, that's one of the students at Pomona College that I mentored. And uh he also became a campus champion. He went to uh boot camp at San Diego. He was the only undergraduate student there uh among a bunch of phd S. Um But uh learning about high performance computing and how to use this technology effectively as a graduate student or a phd is too late.

You really have to start earlier. So Zintan is proof uh that uh the efforts work. Um And on the bottom, you see a picture of a girl from Snap the Gap program and her parent uh surveying her creation, which uh I believe had a sensor counting uh the balls that were thrown into the plastic cups. And then I have a couple of pictures. I was um reminiscing about uh my experience at conferences. Um Those of you who had had been to technical conferences, know what I'm talking about. And so this picture is about uh me in my natural habitat. So to say, I noticed a thing like a rock and I will see something about it. No, no one else sees and I'll uh take a picture and I'll think about it and then something interesting comes out of that experience. And uh so there's another picture which is a po post uh conference day full of activities. Um The only woman I'm excited talking about something technical. Um don't quite know what um But uh everybody is listening and paying attention. And uh I realized when looking at this, that when that picture was taken, I was not aware of the fact that I was the only woman in that group that came way later. Um And I'd like to change it.

So it feels pretty lonely when you realize that you are the only one in the group and even not considering, um, how you're not listened to or how you're sometimes shut down or ignored. Uh, and your opinion is not considered valid. Um, but it's a lonely experience. I would like to see more women in, uh, technical jobs switching gears a little bit. Uh, I'm gonna tell you about Pomona College infrastructure and, uh, the pictures are of our data center and the students and I like to give them tours. Uh So we have about 1000 local cores. Uh We use exceed, we use cloud. I do not engage with that conversation of moving everything to cloud. In fact, I moved some researchers out of cloud because that became too expensive. It was funded by a great expire. Uh When I was building infrastructure, I made the decision to uh make it uh diverse. So I didn't get uh 10 of the same thing. I got all the unique technology that was available at the time, the house amount of cores in a single two sacket server like am D's Epic, which came out at the time and then the 28 core Sky Lake from Intel. And of course, comparing how A MD and Intel work with different software. And then NVIDIA Cards, uh Pascal generation P 100 was a staple at the time. And then Volta came out.

So we had both P 100 V 100 of course, the Volta architecture became uh very useful for artificial intelligence and inference workload. And uh we got really expensive flash storage for uh bioinformatics step workloads that are really uh high in IO usage. Uh And we're experimenting with mo as an on premises cloud storage and also it's a pretty good performance for most most workloads. Uh We are using SLRM as a scheduler for jobs and experimenting with Kubernetes for scaling things like git lab uh and doing auto grading for a computer science professor. We are using our and our studio, both desktop and web-based versions and which is capable of submitting jobs to our HPC infrastructure. Uh We are also using uh Jupiter uh in all of its situations, Jupiter hub, Jupiter lab. Uh And we try to support all the researchers um and recognize that their workflow is comfortable to them. And there's got to be a way to convert what they're doing um on a day to day basis into something that's more scalable. So that's part of my job. The primary languages that uh we work with right now are Python Kaya and Julia. And this is a slide I stole from another presentation I gave about the lab. Uh And to explain to you why the lab started. Uh I'm a geek. I like playing with technology.

I have basically the same setup at home and I wanted to share it uh with the students, especially students that come from the backgrounds that did not have exposure to technology. Um And also I uh thought of it as a way to make learning about Linux command line and, and my actual job more fun. Uh because when you're having fun, you're gonna forget about the complexity of the task. So this is a little bit of uh uh what we have in the lab. We of 3D printing, we have virtual and augmented reality headsets. Uh Students get very excited about the fact that they can custom 3D print things for their dorms. Um And also it's hard to just choose one technology, you kind of have to consider uh all of them together. So we do comparative analysis of tethered and wireless headsets and 3 61 80 cameras. We're looking at 3D scanning technology and uh we built a library of um educational resources that can be used to uh learn and then apply that knowledge. Uh another lab slide where you see students working on projects and um some examples of projects that were completed where an art history and politics student learned from scratch, how to D print, how to design things in Tinkercad uh in Doodle 3D.

And she project managed six people and they 3D printed outlines of 60 countries for a bio art project. And then uh there was a media studies student who was designing jewelry, belt buckles and sunglasses. And we made it easier for him to learn about technology. And then uh we had a media studies and CS students that used uh 3 61 80 cameras to create a theater play with augmented reality elements. So that was a really good uh learning experience there as well. The link to the play is at the bottom of the slide. And um so this is an example of what's affordable external GP U is somewhat affordable, more affordable than the full system. Uh Raspberry Pis are affordable. And with this technology, people can be taught how clusters are built and how they work. Uh Also uh when you uh build something like this, it feels like a toy, but it's a real learning experience. It teaches you about command line, distributed computing, uh sharing keys, version control.

This is another example of a typical asset slide where I like to provide information uh for people to follow up on certain concepts. Like for instance, uh the concept of power over internet where you don't need to have a separate power adapter for each Raspberry Pi, you can power it literally via the switch that is plugged into. Uh and I've seen time and time again how to close students minds and they uh get engaged with technology and realize that it doesn't have to be difficult. And uh this is the list of other uh technical toys we have. There's a Merge Cube, there's an intel compute stick uh which is uh uh an addition of a TPU or a special processing unit to a Raspberry by that doesn't have a built in graphics card or a GP U. And that's a uh meeting speaker with 360 ability to record in 360 also spatial sound recognition so that it focuses the camera on somebody who speaks uh and lots more in the slide that I don't have time to get into. But you can Google. all these, these are really fun. And some of the projects that were in uh our lab were ORAC Vina, which is used in uh chemical modeling uh that can run on Raspberry Pi clusters. So that was a really good use case.

And then one of the students was studying race in DR is it represented, is it represented? Well, uh some of these are obvious answers but um that was the step number one for the students to actually get engaged and start creating something uh in a new medium. And this was a slide from meeting uh something called Miami Posse at Pomono College. The students uh get a chance to visit the college uh before they come to study. Uh So during the summer, these groups would come through and I would meet with them and tell them about the lab and high performance computing and uh sneakily tried to recruit them so that they can work for me when they're uh on campus. And uh I like telling the stories and pictures. So this is more data center tours. There's my colleague, Angelina helping me at the job fair, uh talking to students. There's uh max uh showing people raspberry pies, I believe. Uh And there's Lindsay and Nicole holding a hard drive on their hands for the first time and there's Zintan installing parts into a laptop and um I like to see people happy. I like that. Everybody is smiling and there's more examples here, there's Ariane uh working on uh one of her art history projects and there's Kevin printing in a resin uh 3d printer. And uh I will briefly go over interesting use cases that we are working on right now.

There is uh a psychology project that is studying decision making. There are several projects in economics that they're dealing with things like topic modeling and sentiment analysis. Um We are working with GIS uh with various different software packages and uh Ortho mosaics uh in physics, we have some really interesting projects with nuclear fusion simulations and quantum IIL. And in computer science, we are using virtual GP US and in environmental analysis, we're using sensors and cues and no SQL and graph databases and data visualization. And our museum wants to build an augmented reality app because uh not everybody can visit it or at least it's limited for the time being. So that way we can maybe bring art to more people in terms of the future summer research just kicked off. It's been an incredibly busy time. Uh The college is gonna be partially teaching online. So I'm helping out my team to figure out things like tablets and apps that run on tablets. Um 64 bit Raspberry Pi is out. So that's probably the most exciting thing that happened lately.

For me, I'm looking forward to 100 and 28 cores in a single two cir server. Uh I'm experimenting with opt in which can be used for memory or storage. Um I like to build science gateways specifically for economics, but maybe other disciplines to make access to high performance computing, easier to people that don't necessarily want to spend time on command line. I'm continuing to train undergraduate students in my job and I'm participating in the high performance community diversity efforts. And uh this is second to last slide, I think. So, students programming the DG I Robo master uh using Python. Uh everybody is engaged and they're not even realizing their learning skills they can use later in real jobs. And then um a group of students created an app called HPC Ponder. And that will be my last slide. I'll tell you a bit more about that. And uh this is a what is human class where people are exploring technology and uh trying to see how it relates to us as people, as humans and a picture of our lab jam packed with equipment. And that's my last slide. Uh This is the app that was created by the students because we believed that uh in order to be successful, something you have to have the language of that area. So this was our attempt to onboard students as quickly as possible.

But by giving them access to words that they will need to use in their job and uh explanations and link links to where to learn more. Um So I encourage you to download it right now. It's only I OS, but it's coming in Android as well and uh contact us and let us know what you think and I'm gonna post here with, I think one minute remaining.