2024-07-10 –, Function (4.1)
The poster studies the optimal distributed architecture to run simulations with Julia-based models in the field of Systems Biology. Julia parallel framework of distributing the code across remote workers is applied to different cluster configurations: managed by HTCondor and Kubernetes through Julia interfaces ClusterManagers.jl and K8ClusterManagers.jl. The poster summarizes InSysBio experience and presents pros and cons of building such computational environment for modeling and simulations.
One of the main computational challenges of Systems Biology and Quantitative Systems Pharmacology is solving large differential equation systems with thousands sets of input parameters. The task becomes even more challenging when multiple therapies and clinical conditions are being modeled. The software and hardware environment to address this problem and speed-up the simulations is of utmost importance for the field. The poster proposes several approaches to address this problem in Julia. Julia ability to easily parallelize the computations and distribute workloads across workers is applied to the following architectures:
- On-site distributed computational environment managed by HTCondor through ClusterManagers.jl interface.
- Extension of the local computational resources by adding cloud based compute nodes.
- Fully cloud-based environment managed by Kubernetes through K8ClusterManager.jl package.
The advantages and limitations of the studied approaches are presented in the poster.
Ivan earned his Master’s degree in Mechanical Engineering from the Department of Mathematics and Mechanics of Lomonosov Moscow State University. Ivan started his career at IBM East Europe/Asia providing customers with technical support and «proof of concept» implementation of IBM software products. In 2017, Ivan joined InSysBio as mathematician and software developer, where he continues to work until present as senior software developer. Ivan participates in the development of key InSysBio software tools for simulations and analysis of QSP models, namely Julia-based package for simulations and parameters estimation HetaSimulator.jl, practical identifiability toolkit LikelihoodProfiler.jl, packages for Virtual Patients generation, etc. Beyond software development, Ivan is engaged in the enhancement of mathematical and computational methods for QSP modeling.