A comprehensive review of GPU virtualization and sharing methods
Abstract
We present a survey of of GPU virtualization and sharing methods for environment with containers and virtual machines. Vendors-enabled technologies as well as approaches exploited by researchers are covered. Extra capabilities enabled by virtualization layers such as remoting, overcommitment, checkpoints, scheduling, and preemption are disused. Specific solutions are classified and summarized in tables. Future research directions for accelerator sharing in cloud environment are discussed.
Fields of Study & Subjects
Cloud Computing and Resource Management
Parallel Computing and Optimization Techniques
Cloud Computing and Remote Desktop Technologies
Publication details
- Journal / source
- HAL (Le Centre pour la Communication Scientifique Directe)
- Publication date
- 1 January 2026
- Language
- EN
- Document type
- Preprint