Expert Interview: Inder Monga

[00:00:00] Jenny: Hi, this is Jenny from Berkeley Lab Strategic Communications Team. Today I have Inder Monga with me. Inder is the Executive Director of Energy Sciences Network, or ESnet, which is the Department of Energy’s High Performance Network user facility.

[00:00:28] Inder Monga: Hi, how are you, Jenny?

[00:00:30] Jenny: Good, good. Inder, can you just tell me a little bit about yourself and your role at the lab?

[00:00:35] Inder Monga: I’m a division director of the Scientific Networking Division at Berkeley Lab. We are part of the Computing Sciences Area at the lab. As the Division Director of Scientific Networking, I’m also the executive director of ESnet, Energy Sciences Network, which is a DOE user facility. I also wear a couple of other hats.

I am the lead PI, a principal investigator, for the QUANT-NET project, which is building a quantum networking testbed for distributed quantum computing. And I am the co-PI for an NSF midscale research project called FABRIC. FABRIC is a network testbed that supports more than a thousand users across all the universities for research and advanced network technologies, as well as educating networking to students around the US.

[00:01:25] Jenny: So Inder, for those who don’t know, what is ESnet?

[00:01:30] Inder Monga: ESnet, or Energy Sciences Network, is the Department of Energy’s data circulatory system. What I mean by that is we connect tens of thousands of scientists and researchers at the 17 national labs working on advanced user facilities — 28 of them, like light sources and supercomputers, to get their research done. We move petabyte-scale datasets. We move data from instruments like telescopes and Large Hadron Collider in Europe, and we make sure that the scientists can get their research done quickly and effectively. And no matter where the instrument is that they need the data from, no matter where the computing is that they have access to, they can move the data to computing to get insights as quickly as possible.

That’s what we enable. And we work and collaborate with research and education networks all around the world to make that happen.

[00:02:30] Jenny: So how does ESnet enable scientific collaborations?

[00:02:35] Inder Monga: We work very closely with our science partners. We have a Science Engagement team that holds regular Requirement Reviews where we understand the process of science. We do not ask them what they need from the infrastructure, but we derive that based on how scientists would ideally run the scientific process to get insights. So, we ask them what they expect to see in one year, two years, five years, and we build innovative services to make sure our infrastructure will not be a bottleneck for what they want to accomplish, and it can even enhance or accelerate their scientific enterprise.

[00:03:15] Jenny: So, aside from this reliable, robust infrastructure, can you tell us about some of the innovative services that ESnet provides?

[00:03:23] Inder Monga: Yes, but first I want to emphasize the reliable, robust infrastructure that you mention. The ESnet operations team does a fantastic job of not just  architecting and providing high reliability but also the lossless data movement that really benefits our petabyte-scale scientific datasets to move so much faster. 

In addition, our team also spends a lot of energy creating and providing really advanced services that are tailored for science. One of the services we provide is OSCARS, which is a guaranteed bandwidth service. So, if you want to move a large amount of data reliably, you would use this OSCARS service. And this OSCARS service can also go across different networks and provide reliability. If a network connection goes down, we can build in failovers, so the data stream never gets interrupted.

We also do some cool things like provide large caches for scientific data, so the scientists can get access to the data much quicker, so they can process things faster and they can get to their results faster. One of the new innovative services that we are trying right now, as a prototype, and that we hope to take into production, is an automated interface for scientific workflows. So we have an API that the scientific workflows can request network services, or even reason and see what capacity is available for them and make sure that they can use this automated interface, a programmatic interface, to request services from the network.

[00:05:07] Jenny: So Inder, what are some of the urgent national priorities that ESnet is tackling?

[00:05:13] Inder Monga: Let me talk to you about AI and quantum. The complex and distributed nature of large model AI training requires a robust networking infrastructure to support these workflows effectively. We need to move large data sets to these GPUs for processing and training. So the network that we are building has the capacity and the capability to provide those for training large scientific models, very similar to the LLMs that we are seeing in the environment.

We are not only focused on enabling AI, but we are also building AI and AI tools within the network to be able to offer these services effectively. So we are collecting a whole bunch of data on what flows over the network, and through AI we are finding proactively problems in the network, answering questions, and handling tickets or bugs that may exist in the network, in a way that is much faster and better than what we can do today.

In addition, we are researching and investing in quantum networking. One of the big challenges is to build a utility-scale quantum computer that can actually get the quantum advantage, process algorithms of scientific importance. And in order to do that, you need to connect multiple quantum computers up.

So how can you connect multiple quantum computers in a way that allows us to get a quantum advantage, is research that we are currently doing under the quantum networking umbrella in a project called QUANT-NET. So we are really trying to see how distributed quantum computing can bring the power of quantum quicker to the scientific enterprise.

So these are the ways we are tackling the national priorities.

[00:07:06 ] Jenny: So lastly, I just want to ask, what is one thing that you might want someone to take away from this interview? 

[00:07:12] Inder Monga: What I would like people to know is when the network works, it is invisible. But at ESnet we are not just a network, we are a partner to all the scientific enterprises and really work hard to collaborate with them to accelerate their science.

[00:07:30] Jenny: I just want to thank you so much for being here today, Inder. I learned so much about ESnet, and I hope our listeners did too.

[00:07:37] Inder Monga: Thank you, Jenny. It was a wonderful conversation.

[00:07:47] Jenny: If you want to learn even more about ESnet, visit es.net or head to lbl.gov.