What question did this study set out to answer?

This research aims to enhance GPU sharing efficiency for deep learning tasks using a hybrid scheduling framework.

MMK: A Hybrid Scheduling Framework for Fine-Grained GPU Sharing for Deep Learning Applications

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial shows improved GPU sharing performance in deep learning applications, indicating better resource utilization.

This research aims to enhance GPU sharing efficiency for deep learning tasks using a hybrid scheduling framework.
Developed a hybrid scheduler to integrate Multi-Instance GPU (MIG) and Multi-Process Service (MPS) resources.
Implemented a kernel scheduler for fine-grained scheduling of online and offline jobs.
Conducted extensive experiments to evaluate the proposed framework's performance against state-of-the-art solutions.
Reduced average job completion time by 28% compared to the state-of-the-art framework.
Decreased makespan by 32%, demonstrating enhanced scheduling capability.
Increased system throughput by 35%, indicating significant performance improvements.

Jiang et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark