-Computer simulations have become a very powerful tool for scientific research. In order to facilitate research in computational biology, the BioDynaMo project aims at a general platform for biological computer simulations, which should be executable on hybrid cloud computing systems. This paper describes challenges and lessons learnt during the early stages of the software development process, in the context of implementation issues and the international nature of the collaboration.
The BioDynaMo project is a long term effort in the field of biological simulation to build a scalable and flexible platform. The purpose is to give life scientists access to increasing amounts of computational resources and provide a framework that hides the computational complexity, allows them to focus on their research, and promotes reusability and reproducibility of the results from shared open access data. In order to have an impact on the community, the system must be flexible enough to execute simulations from different specialities with possibly quite distinct requirements.
The project started as a code modernization initiative,
inspired by the scientific principles underlying the simulation
software Cx3D [
In general, software modernization is a collective term that subsumes a variety of activities. In our case it means transforming the application from Java to C++ and changing the architecture in a way to utilize multiple levels of parallelism offered by today’s hardware and modern distributed computing models.
Although Cx3D has a very compact code base (15 kLOC), it is able to perform complex simulations like “cortical lamination”. However, the absence of modern software development practices such as automated tests, continuous integration, coding standards compliance, and code reviews hinders a sustainable development process.
This work is licensed under a CC-BY-4.0 license.
Our first step was to introduce development techniques and
infrastructure aimed at improving code quality and
maintainability, which are essential for our long time effort. Based on
available effort, we opted for testing the whole application
rather than writing unit tests for the entire codebase. Existing
demo simulations were taken and transformed into test cases.
The resulting simulation state is then transformed into JSON
format and compared to a ground truth obtained from Cx3D
0.03. A public code repository was created on Github [
III. MODERNIZING LEGACY CODE: EXAMPLES OF THE
METHODOLOGIES APPLIED
High performance and high scalability are the prerequisites to address ambitious research questions like modeling epilepsy. Our efforts in code modernization were driven by the goal to remove unnecessary overhead and update the software design to tap the unused potential enabled by the paradigm shift to multi and many-core distributed systems.
The Intel Modern Code Development Challenge organized in 2015 with CERN and Newcastle focused on optimizing sequential C++ brain simulation code provided by the Newcastle University in the UK. The contest followed a gamification approach where participating students competed against each other to win an internship at CERN. The ranking was based on the runtime of the provided simulation. Using data layout transformations (array of structures to structure of arrays), parallelization with OpenMP, a custom memory allocator and Intel Cilk Plus array notation, the winner was able to decrease the runtime by a factor of 320. This clearly shows the economic potential of code modernization efforts coupled with gamification and encourages to repeat the challenge.
Furthermore, we ported the Java code base to C++. This
language is better suited for high performance computing as it
is compiled to native machine code removing the overhead of
running in a virtual machine and provides the right ecosystem
for parallelization and optimization. The following iterative
porting approach has been chosen. First, a Java class is selected
and replaced by its C++ translation. In the second step, this
C++ class is connected to the remaining Java application.
Finally, the Java/C++ hybrid is compiled and used to execute
a number of tests. If all tests pass, the developer can proceed
with the next iteration by selecting another Java class. On the
other hand this means that errors, indicated by test failures,
must have been introduced by code changes since the last
iteration. Therefore, this procedure significantly simplifies
debugging. Although this approach is associated with additional
development overhead in connecting classes in C++ to Java,
it gives the benefit of obtaining a runnable system after each
iteration. Without that additional effort, the first time the C++
version would be able to execute tests, would be at the very
end, after all classes have been ported. Porting would have
been a lot easier if every class / function had sufficient unit
tests. In this scenario connecting both languages would no
longer be required since tests could be executed for each
function independently. Testing the whole simulation software
had another drawback: floating point differences on diverse
systems amplified over many iterations and were responsible
for test failures although the code was correct. We fixed that
issue using the math-library crlibm to obtain reproducible
results across different environments as suggested in [
The field of computational biology covers a wide range
of scientific topics, each producing many different scientific
models, such as for instance described by [
In addition to these more scientifically-centered aspects, also considerable challenges arise from a computational/technological point of view. First steps towards such efficient software implementation have been made in the context of the “Intel Modern Code Developer Challenge” competition. Overall, we believe we have created a collaborative foundation for the efficient continuation of the very ambitious software development project of BioDynaMo.
However, considerable challenges remain in the current
software development process. The verification and validation
of the software is paramount. The recent study of [
This work was possible thanks to the support by CERN and CERN openlab Code Modernization program in cooperation with Intel; by the Human Green Brain Project (www. greenbrainproject.org) through the Engineering and Physical Sciences Research Council (EP/K026992/1); by Innopolis University, and its Service Science and Engineering lab (SSE); and by SCImPULSE Foundation. The funding institutions had no role in the design of the project, decision to publish, or preparation of the manuscript.