Dr. Alexei Razoumov creates three-dimensional numerical models of galaxy formation, supernovae and other phenomena.

Astrophysics research Compute Canada


Researcher Alexei Razoumov

Dr. Alexei Razoumov
Visualization Coordinator, WestGrid

Research area
Dr. Razoumov specializes in computational astrophysics. His work has relied on Compute Canada infrastructure to create three-dimensional numerical models of galaxy formation, supernovae and other phenomena. Prior to joining WestGrid in 2014, he spent five years as an HPC Analyst at SHARCNET helping researchers from diverse backgrounds to use large clusters.

Relevance to other sectors
Dr. Razoumov is sharing his expertise in transforming massive amounts of data into 3D “pictures” to help Compute Canada users from other scientific fields. For example, material scientists increasingly rely on computer visualizations to understand physical processes — from the metre to the micro scale — that effect material stress in airplane wings or bridges.


Science is producing an unprecedented amount of data on everything from predicting weather and ocean changes to searching for distant galaxies. How are visualization tools helping researchers make sense and maximize utilization of such vast amounts of data?
Visualization is all about converting numbers into a visual form that people can understand easily. If you have a billion numbers from sensors measuring changes in ocean currents, you need some way to make sense of these numbers. Visualization tools allow researchers to explore large datasets interactively without being overwhelmed by the sheer amount of data. These datasets can come both from measurements and from numerical simulations. For example, in my area of computational astrophysics we didn’t have the tools or the computing power 25 years ago to model an entire galaxy. Now we do, and these models generate a huge amount of data. To answer the important scientific questions, for example, how the formation of the first generations of stars some 12 billion years ago dispersed heavy elements (e.g. carbon, silicon and oxygen) across the galaxy allowing the first planets to form, we need to analyze these models to visualize in 3D the distribution of heavy elements.

What do these visual representations look like?
It’s similar in a way to Google Earth, where you have an object that you can interact with and zoom in on regions where you want more detail. But with advanced visualization, you can explore data in three-dimensional space: you can rotate your dataset vertically and horizontally, change your camera position fly through the dataset, or turn on and off various features and variables.

Why is this so exciting, compared to what researchers used in the past?
It’s exciting because you are looking at the actual scientific output. You’re not looking just at the numbers. You’re dealing with some complex phenomena that you can analyze and present visually in an interesting way that makes sense.

In your role as Visualization Coordinator at WestGrid, how are you helping other disciplines use visualization tools and technologies?
My job is to help researchers find the visualization tools they need, teach them how to import their data and use these tools, and how to scale these tools to analyze large datasets on the WestGrid systems. There are many diverse fields where it is paramount to visualize data to answer the important scientific questions, from oceanography and atmospheric sciences to medical imaging to physics, biology and chemistry. Recently, there is also growing interest in information visualization and visual analytics types of problems from linguistics and other disciplines within the digital humanities.

Is it essential to have access to Compute Canada infrastructure to use these visualization tools?
You could use these tools on your laptop or on a workstation but then you wouldn’t be able to load very large, multidimensional datasets. Without Compute Canada it would be absolutely impossible to produce and analyze these datasets, simply because you just don’t have enough memory and CPU power in a stand-alone machine. A typical large simulation could use tens of thousands of processors, and visualizing such model would also require resources beyond a single desktop machine.

What kind of assistance will you be providing to researchers and their teams?
I’ll be contacting people who use WestGrid a lot to understand what type of visualization needs they have and suggest how they can extract nice pictures from complex simulations. Some people are still using visualization tools that are 10 to 15 years old and running them on their laptop. I want them to be aware of the range of tools that exist these days and the value of scaling up to tackle much larger problems.

What will it mean for the graduate students and postdoctoral fellows who use these advanced visualization tools? Will it make them more employable?
If you learn how to use these tools for a scientific problem then you’ll be able to use them for any sort of technology problem where you have a lot of data. These tools are somewhat agnostic — they can be used across a wide variety of disciplines.

Are Canadian researchers big users of these tools compared to researchers in other countries?
In some sense we’re playing catch-up. When you go to a super computing conference in the U.S., for example, I’d say that almost all projects using advanced visualization are from the States, Europe or Asia. That’s starting to change with some Canadian projects with larger datasets but, for the most part, large visualizations come from researchers in other countries.

So how do you get the word out that these resources are available through Compute Canada?
That’s my job. I’ll start by contacting research groups that either already use Compute Canada a lot for computing, or have large datasets on their own facilities. I’d like to build a visualization showcase online to show larger visualizations that we’ve worked on so far to demonstrate what is possible. I predict the next decade will be an exciting one for visualization and Canadian research.