What question did this study set out to answer?

The study aims to develop a scheduling framework to mitigate fragmentation and resource contention in GPU workloads.

December 22, 2025Open Access

An Online Fragmentation-Aware Scheduler for Managing GPU-Sharing Workloads on Multi-Instance GPUs

View Full Paper

Ask AI

Bookmark

Authors

An online scheduler improves job placement in GPU-sharing environments, suggesting better performance with conditional load balancing.

The study aims to develop a scheduling framework to mitigate fragmentation and resource contention in GPU workloads.
Proposed an online scheduling framework that utilizes conditional load balancing and job migration
Dynamic adaptation of job placement to minimize contention
Integrates GPU resource partitioning to address fragmentation
Improved system efficiency achieved with the proposed framework
Makespan reduced by up to 35% when all techniques are applied

Ting et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark