Lukas Breitwieser successfully defends his PhD

We were very excited to celebrate two successful PhD defenses on Monday, November 11, 2024! We’d like to wish Lukas Breitwieser and Can Firtina a very warm congratulations on their achievements.

[You can read more on Can’s PhD defense and work here]

In Lukas’s thesis work, “Design and Analysis of an Extreme-Scale, High-Performance, and Modular Agent-Based Simulation Platform” he developed the high-performance simulation engine, BioDynaMo.

Since its publication and open source release, researchers have used BioDynaMo to study radiation-induced lung injuries, vascular tumor growth, invasion of Gliomas into surrounding tissue, the formation of retinal mosaics in the eye, freezing and thawing of cells, neuronal geometries, the formation of the cortical layers in the cerebral cortex, the spread of viruses on a country-scale, and more. BioDynaMo sets a new standard for agent-based simulation platforms by combining high performance, scalability, modularity and user-friendliness. It is the first agent-based simulation platform that supports more then 500 billion agents, and is poised to enable even larger simulations with current, ongoing work.

Background:
Agent-based modeling is an indispensable tool for studying complex systems in biology, medicine, sociology, economics, and other fields. However, existing simulation platforms exhibit two major problems: 1) performance: they do not always take full advantage of modern hardware platforms, which leads to low performance and 2) modularity: they often have a field-specific software design. BioDynaMo was developed to overcome these limitations by maximizing parallelization, reducing the memory access latency, and omitting the collision force calculation. A distributed simulation engine also allows BioDynaMo to scale out the computation of one simulation to multiple servers.

The core contributions of Lukas’s thesis work are:
BioDynaMo significantly reduces the simulation runtime, enabling larger and more complex simulations, faster iterative development, and more extensive parameter exploration. BioDynaMo also significantly improves adoption, thereby enabling advances in many different domains.

Specifically, BioDynaMo can 1) simulate 500 billion agents (a 84x improvement over the state-of-the-art), 2) scale to 84’096 CPU cores, 3) significantly reduce the simulation time (e.g., BioDynaMo simulates an iteration of 800 million agents in 0.6 s instead of 5 s), and 4) significantly increases the visualization performance by 39x.

Lukas’ work on BioDynaMo with co-authors Ahmad Hesam, Fons Rademakers, Juan Gómez Luna, Onur Mutlu received the Best Artifact Award at PPoPP 2023 [Read more].

Did you know: “BioDynaMo is both the name of a multidisciplinary project bringing together European and International Institutions, and researchers, united by a shared mission to bring a new model of computational practices to digital health AND of the software suite it develops, maintains and distributes under the Apache 2.0 license.  The software supports scientists to easily create, run, and visualise multi-dimensional agents-based simulations (be them biological, sociological, ecological, financial, …). Built on top of the latest computing technologies, the BioDynaMo suite will enable users to perform simulations of previously unachievable scale and complexity, making it possible to tackle challenging scientific research questions.” BioDynaMo project website.

You can follow Lukas’s activity and see what he’s up to on LinkedIn and follow the BioDynaMo project on the homepage.

Advisor: Onur Mutlu (ETH)
Second Advisor: Fons Rademakers (CERN)

Co-Examiners:
Can Alkan (Bilkent University)
Arnau Montagud (CSIC)
Mohammad Sadrosadati (ETH Zurich)

Lukas Breitwieser (defended 11 November 2024)
Thesis title: “Design and Analysis of an Extreme-Scale, High-Performance, and Modular Agent-Based Simulation Platform”
[Slides (pdf)]