A comprehensive evaluation of spatial co-execution on GPUs using MPS and MIG technologies
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866909000521482240 |
|---|---|
| author | Villarrubia, Jorge Costero, Luis Igual, Francisco D. Olcoz, Katzalin |
| author_facet | Villarrubia, Jorge Costero, Luis Igual, Francisco D. Olcoz, Katzalin |
| contents | To mitigate the increasingly common underutilization of computational resources in modern GPUs, spatial sharing methods enable multiple applications to use them simultaneously. This work presents a comprehensive evaluation of NVIDIA's primary technologies to achieve that goal: Multi-Process Service (MPS) and Multi-Instance GPU (MIG). Our findings reveal a crucial trade-off between MPS's flexibility and MIG's isolation, and provide many key insights for improving the co-execution strategy according to job profiles. In the most favorable scenarios, MPS improves performance by up to 30% and reduces energy by about 20%, using its provisioning option to avoid resource monopolization. However, under memory contention, it suffers severe degradation, worsening performance by around 30%. Conversely, MIG's full hardware isolation resolves memory contention, leading to more consistent improvements, but these gains are tempered by higher overhead, and its rigid scheme can degrade performance in certain cases. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_22430 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | A comprehensive evaluation of spatial co-execution on GPUs using MPS and MIG technologies Villarrubia, Jorge Costero, Luis Igual, Francisco D. Olcoz, Katzalin Distributed, Parallel, and Cluster Computing Hardware Architecture C.1.3; B.8 To mitigate the increasingly common underutilization of computational resources in modern GPUs, spatial sharing methods enable multiple applications to use them simultaneously. This work presents a comprehensive evaluation of NVIDIA's primary technologies to achieve that goal: Multi-Process Service (MPS) and Multi-Instance GPU (MIG). Our findings reveal a crucial trade-off between MPS's flexibility and MIG's isolation, and provide many key insights for improving the co-execution strategy according to job profiles. In the most favorable scenarios, MPS improves performance by up to 30% and reduces energy by about 20%, using its provisioning option to avoid resource monopolization. However, under memory contention, it suffers severe degradation, worsening performance by around 30%. Conversely, MIG's full hardware isolation resolves memory contention, leading to more consistent improvements, but these gains are tempered by higher overhead, and its rigid scheme can degrade performance in certain cases. |
| title | A comprehensive evaluation of spatial co-execution on GPUs using MPS and MIG technologies |
| topic | Distributed, Parallel, and Cluster Computing Hardware Architecture C.1.3; B.8 |
| url | https://arxiv.org/abs/2604.22430 |