PRISM: Probabilistic Runtime Insights and Scalable Performance Modeling for Large-Scale Distributed Training
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917400764481536 |
|---|---|
| author | Golden, Alicia Kuchnik, Michael Hsia, Samuel DeVito, Zachary Wei, Gu-Yeon Brooks, David Wu, Carole-Jean |
| author_facet | Golden, Alicia Kuchnik, Michael Hsia, Samuel DeVito, Zachary Wei, Gu-Yeon Brooks, David Wu, Carole-Jean |
| contents | Large model training beyond tens of thousands of GPUs is an uncharted territory. At such scales, disruptions to the training process are not a matter of if, but a matter of when -- a stochastic process degrading training productivity. Dynamic runtime variation will become increasingly more frequent as training scales up and as GPUs are operated in increasingly power-limited and thermally-stressed environments. At the 64,000+ GPU scale, we already observe 9% GPU time variability for frontier foundation model training. Motivated by our analysis and the large design space around performance variability, we present PRISM -- a performance modeling framework that captures the stochastic nature of large-scale distributed training. The core of PRISM is a statistical method that quantifies probabilistic guarantees on training time. Using PRISM, we explore the design and optimization space of distributed training, enabling principled, variability-aware decisions that improve performance and system efficiency at scale. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_15596 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | PRISM: Probabilistic Runtime Insights and Scalable Performance Modeling for Large-Scale Distributed Training Golden, Alicia Kuchnik, Michael Hsia, Samuel DeVito, Zachary Wei, Gu-Yeon Brooks, David Wu, Carole-Jean Distributed, Parallel, and Cluster Computing Large model training beyond tens of thousands of GPUs is an uncharted territory. At such scales, disruptions to the training process are not a matter of if, but a matter of when -- a stochastic process degrading training productivity. Dynamic runtime variation will become increasingly more frequent as training scales up and as GPUs are operated in increasingly power-limited and thermally-stressed environments. At the 64,000+ GPU scale, we already observe 9% GPU time variability for frontier foundation model training. Motivated by our analysis and the large design space around performance variability, we present PRISM -- a performance modeling framework that captures the stochastic nature of large-scale distributed training. The core of PRISM is a statistical method that quantifies probabilistic guarantees on training time. Using PRISM, we explore the design and optimization space of distributed training, enabling principled, variability-aware decisions that improve performance and system efficiency at scale. |
| title | PRISM: Probabilistic Runtime Insights and Scalable Performance Modeling for Large-Scale Distributed Training |
| topic | Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2510.15596 |