Caliper improves accuracy in high performance computing network simulators
CALIPER: Calibrating HPC/AI Network Simulators
Networking and Internet ArchitectureDistributed, Parallel, and Cluster Computing
Summary
Simulating networks in high-performance computing and artificial intelligence systems helps researchers test ideas without building costly hardware. But these simulators need to be carefully tuned to match real-world systems accurately. The authors introduce CALIPER, a way to systematically adjust simulator settings using real measurements and logical constraints. This technique improves predictions across different network sizes, making simulations more reliable for designing future systems.
What this means in practice
- •For hpc system designers: Use CALIPER to tune network simulators for accurate performance predictions across large-scale HPC configurations.
- •For ai infrastructure engineers: Improve deployment planning of AI clusters by calibrating simulators to represent network communication accurately at various scales.
- •For cloud service providers: Refine virtual network performance estimates in cloud data centers by applying CALIPER's calibration methodology to simulation tools.$Commercial implications: Enables selling enhanced network simulation software that predicts data center performance more reliably to cloud providers.
Authors
Seyyidahmed Lahmer, Daniel Seibel, Srijeet Basu, Arno Troch, Jayesh Badwaik, Subhomoy Bhattacharyya, Abubakr Nada, Debjyoti Bhattacharjee, Vinay Kumar Baapanapalli Yadaiah, Michael Peeters, Andreas Herten, Darayus Adil Patel
Abstract
Network simulation plays an important role in HPC/AI system research by enabling exploration of network architectures, communication strategies, and system configurations at scales that may be impractical to study experimentally. Existing approaches to network simulator calibration and validation lack a principled methodology that combines parameter alignment, constrained inference, and independent validation for collective-centric HPC/AI networks. We present CALIPER, a methodology for systematic calibration and validation of HPC/AI network simulators. CALIPER distinguishes directly observable system parameters from auxiliary parameters which are configured via calibration and constrains the latter to feasible domains consistent with the target system. We evaluate CALIPER's hierarchical, scale-aware calibration and its effects on parameter identifiability and generalization using collective communication benchmarks on a reference system with up to 64-nodes at the Jülich Supercomputing Centre. Validation results show that the calibrated parameter configurations identified by CALIPER predict system performance with less than 5% mean errors across scale-up and scale-out network domains. CALIPER demonstrates that hierarchical, constrained calibration can produce simulator configurations whose predictive accuracy persists beyond the observations used for fitting.