Welcome to the first interview in our 7 Questions for Chapel Users interview series for 2026! In this edition, we hear from Dr. Oliver Alvarado Rodriguez about his experiences using Chapel in his Ph.D. thesis to write Arachne, a graph analytics package for Arkouda. This article is a logical successor to our earlier interview with David Bader, who served as Oliver’s Ph.D. advisor. After Oliver graduated, we were very happy to have the opportunity to continue working with him within HPE’s Advanced Programming Team that Chapel is a part of.
1. Who are you?
My name is Oliver Alvarado Rodriguez, and I am a software engineer at Hewlett Packard Enterprise. I received my Ph.D. in Computer Science from the New Jersey Institute of Technology in Newark, NJ. My dissertation, On the Design of a Framework for Large-Scale Exploratory Graph Analytics, focused on extending the Arkouda framework to handle more complex and sparse problems such as graph analysis. Rather than modifying Arkouda directly, we implemented an add-on module called Arachne.
2. What do you do? What problems are you trying to solve?
Since I started my Ph.D. in September 2020, and continuing into my current role, I live and breathe graph data structures and algorithms. The great thing about graphs is that anything modeled as an interconnected network can be represented and analyzed with graph algorithms. Once you’re comfortable with graph data structures and algorithms, you’re well-equipped to tackle a wide range of problems in computer science and related fields.
During my Ph.D., I focused on making graph analysis more accessible to Arkouda users. At Hewlett Packard Enterprise, I’ve been improving communication and performance for distributed graph patterns via sparse matrices, and now I’m working on accelerating simulation software by improving underlying distributed graph data structures as well as building tools to facilitate hierarchical simulation models.
“Chapel’s biggest benefit for me is how quickly it lets me prototype high-level parallel and distributed code, and then incrementally optimize it.
”
3. How does Chapel help you with these problems?
Chapel’s biggest benefit for me is how quickly it lets me prototype
high-level parallel and distributed code, and then incrementally
optimize it. For graph algorithms, Chapel allows you to represent a
graph with built-in arrays and standard library containers in a single
line, for example: var G: [0..<n] list(int). That gives you a
compact, high-level representation you can iterate over in parallel
with a simple forall loop. If you want a distributed graph and don’t
want to do special partitioning, you can make that array
block-distributed and still use the same forall loop; the
distribution is handled for you.
This rapid prototyping is invaluable for producing working implementations and for onboarding people who are new to parallel programming or graph algorithms. Once an algorithm is prototyped, you can dig into optimizations. For instance, instead of an array-of-lists, you can use Chapel’s standard library to represent the graph as a sparse matrix and automatically apply CSR, CSC, or COO layouts, giving you bespoke sparse graph representations essentially out of the box.
4. What initially drew you to Chapel?
Before Chapel, my experience with parallel programming was limited to simple task and thread-based programs in C. I didn’t know Chapel existed—or much about OpenMP or MPI—until I began my Ph.D. I came to NJIT intending to do network analysis, but I hadn’t expected the depth of parallel and distributed programming involved.
Chapel made it easy to learn general parallel programming and to see parallelism clearly without wrestling with the more verbose or cumbersome syntax of OpenMP and MPI. That clarity helped me understand how parallel and distributed codes work, and later made it much easier to read and work with codes written in other parallel frameworks.
“Chapel helped me see how parallelism maps to real problems without needing to first understand every detail of message passing, the underlying runtime, etc.
”
5. What are your biggest successes that Chapel has helped achieve?
The code I developed for my dissertation was written entirely in Chapel, and that is the biggest success Chapel helped me achieve. I went from knowing little about parallel programming to becoming comfortable both as a graph scientist and a parallel programmer. Chapel let me learn parallel programming within the domain I love, graphs, and that foundation makes it significantly easier to understand and port ideas to OpenMP, MPI, or other frameworks. At the end of the day, these tools are all about parallelizing operations, and Chapel helped me see how parallelism maps to real problems without needing to first understand every detail of message passing, the underlying runtime, etc.
6. If you could improve Chapel with a finger snap, what would you do?
I would add a general aggregation framework that lets users define their own buffer-flushing functions and apply them to any distributed data structure—even their own user-defined ones. Relatedly, I’d like to see a wider variety of distributed data structures in the standard library that can be used with custom aggregators for moving data into and out of them.
7. Anything else you’d like people to know?
Chapel is awesome, and it’s created by some of the most helpful, friendly, and passionate people I’ve ever met. I was fortunate to intern with the Chapel team in the summer of 2024 and work with them when I joined Hewlett Packard Enterprise. Those experiences rank among the best in my career. Even though I’m no longer directly on the Chapel team, I still lurk in the communication channels and learn something new almost every day. It takes a truly dedicated team to build a language packed with features like Chapel’s. After all, what other language has dedicated, parallel, distributed sparse matrix data structures with all your favorite layouts? (I’ll wait).
We’d like to thank Oliver for participating in this 7 Questions for Chapel Users interview. Stay tuned for future installments of the series!
