arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.03561cs.DCcs.PF
收藏

在大规模上提升性能洞察:一种异构框架用于exascale诊断

Enhancing Performance Insight at Scale: A Heterogeneous Framework for Exascale Diagnostics

Rice University(莱斯大学)

Dragana Grbic

AI总结:

本文提出一种异构框架,利用高性能C++API和GPU并行计算,实现高吞吐量诊断,展示了在Aurora上对10万个MPI进程的快速处理及314倍的加速效果。

摘要:

随着exascale系统达到前所未有的并发性,传统性能分析工具难以应对大规模遥测的开销。我们提出了一种加速基础设施,利用高性能C++API和GPU并行计算,实现高吞吐量诊断。我们的C++API在Aurora上实现了对10万个MPI进程的9.69秒数据摄入时间。此外,我们的GPU加速层在分析10万个执行轨迹时,比CPU处理快314倍。最后,我们实现了一个拓扑感知的工作流,将逻辑性能异常映射到物理Slingshot互连坐标,定位了Aurora上22个不同机架的网络拥堵。我们还展示了框架高级接口如何与外部工具无缝集成,提供复杂的分析模型。我们引入了新的三维性能模型,从执行轨迹中“重新生成”迭代行为;使用该模型,我们发现在Frontier上GAMESS工作负载有32.28%的潜在加速。

英文摘要:

As exascale systems reach unprecedented concurrency, traditional performance analysis tools struggle with the overhead of massive-scale telemetry. We present an accelerated infrastructure for the hpcanalysis framework that leverages a high-performance C++ API and GPU parallelism to enable high-throughput diagnostics. Our C++ API achieves a 9.69-second ingestion time for 100,000 MPI ranks on Aurora. Furthermore, our GPU-accelerated layer achieves up to 314x speedup over CPU-based processing when analyzing 100,000 execution traces. Finally, we implement a topology-aware workflow that maps logical performance outliers to physical Slingshot interconnect coordinates, localizing network congestion across 22 distinct racks on Aurora. We also demonstrate how the framework's advanced interface seamlessly integrates with external tools to provide sophisticated analytical models. We introduce a novel tri-dimensional performance model that "re-materializes" iterative behavior from execution traces; using this model, we identified a 32.28% potential speedup for a GAMESS workload on Frontier.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑