Container-based framework for large-scale spiking network simulation.

High-performance computing (HPC) is critical for simulating large-scale neural networks with detailed biophysics, yet deploying these simulations across heterogeneous computing infrastructures remains a significant challenge due to software dependencies and hardware variability. This study introduces a framework utilizing containerization technology to ensure scalable and reproducible simulations by encapsulating the entire software stack, including compilers, MPI libraries, and GPU toolchains, within a portable image. We created a spiking network model with millions of neurons and synapses by replicating the cerebellar neural circuit module and performed benchmarking with it. Our results demonstrate that containerized simulations introduce little performance penalty compared to native installations while delivering identical results across multiple systems with minimal effort, providing a robust solution for portability. This framework can accelerate the scalable development of spiking network models, facilitate the precise reproduction of complex workflows and simplify collaboration, addressing major barriers in modern computational neuroscience research.