ASCN (Asynchronous Scalable Computing Node) is a high-performance infrastructure designed to automate and scale gprMax physics simulations. By decoupling simulation requests from execution using a message-broker architecture, it ensures that the research workflow remains robust, scalable, and reproducible.
The system is built on a Producer-Consumer model to eliminate bottlenecks in high-intensity electromagnetic wave modelling.
Figure 1: Distributed Workflow using RabbitMQ, Redis, and Dockerized Workers.
- Orchestration: Kubernetes (K8s) with HPA for 10x to 100x elastic scaling.
- Messaging: RabbitMQ for fault-tolerant task queuing.
- Storage: PostgreSQL for metadata and Redis for real-time state tracking.
- Environment: Docker for bit-for-bit identical physics execution.
We don't just scale; we validate. Every simulation undergoes a rigorous Physics Regression check.
Figure 2: Perfect alignment between 'Current' and 'Golden Reference' waveforms.
| Metric | Value | Logic |
|---|---|---|
| NRMSE | 0.0 |
Zero numerical drift across containerized nodes. |
| Signal Integrity | 100% |
Verified via automated A-scan accuracy checks. |
| Scaling Latency | Minimal |
Rapid pod spin-up via optimized Docker layers. |
- Ingestion: Simulation parameters are pushed to the RabbitMQ queue.
- Execution: Available ASCN Workers pull tasks and run
gprMaxin isolated Docker containers. - Validation: Results are compared against SQL Golden Truth benchmarks.
- Reporting: Automated Matplotlib plots are generated for visual verification.
Developed by Prateek Sharma
GSoC '26 Applicant | Cloud-Native Infrastructure Engineer