NUMA balancing hampering performance of spiking network simulations

2026-07-24

AI Summary

This study addresses the performance instability and increased energy consumption caused by automatic NUMA load balancing in large-scale spiking neural network simulations, which undermines energy efficiency in high-performance computing. By integrating distributed performance profiling, fine-grained time tracing, and coordinated monitoring with the jemalloc memory allocator, this work reveals—for the first time—the transient interference mechanism induced by automatic NUMA balancing. Building on these insights, the authors propose an optimization strategy based on user-level control of NUMA balancing. Experimental results demonstrate that disabling automatic NUMA balancing reduces energy consumption by 30% and significantly enhances simulation stability, establishing a reproducible energy-efficiency tuning paradigm for scientific computing applications.

Abstract

Computing centers today mostly operate conventional CPU- and GPU-based systems, where the direct way of decreasing energy consumption is a reduction in the applications' runtime. Neuromorphic computing promises an alternative architecture with improved energy efficiency for artificial intelligence. In this endeavor, code for the simulation of large-scale spiking networks on conventional supercomputers is the reference. We show that turning off automatic NUMA balancing may reduce energy consumption by 30%. This dwarfs other attempts of increasing the energy efficiency of a computing center with respect to cost effectiveness. The memory access pattern of spiking network simulation code dynamically interacts with automatic NUMA balancing. This does not affect the correctness of simulation results and thus goes unnoticed in day-to-day neuroscience research. In performance analysis, however, time measurements fluctuate obstructing attempts to optimize simulation technology. A new time- and compute-node resolved performance display exposes the fine-grained temporal variability of distributed spiking network simulations. The analysis uncovers that automatic NUMA balancing is of disadvantage and affects the jemalloc library for thread-aware memory allocation in a transient manner. The method also allows developers to detect perturbations of the HPC system and target specific improvements to simulation technology. As a consequence, we have equipped our supercomputers with an option to turn on or off automatic NUMA balancing on a per-job basis on the user level. This gives researchers the opportunity to find the best setting for the application at hand. There are indications in the literature that the effect has been observed before, yet it does not seem common knowledge in scientific computing. It remains to be investigated how widespread the phenomenon is among scientific codes.

Problem

• NUMA balancing
• spiking network simulation
• energy efficiency
• performance variability
• HPC

Innovation

• NUMA balancing
• spiking neural networks
• energy efficiency
• performance analysis
• jemalloc