Boosting GPU Virtualization Performance with Hybrid Shadow Page Tables

@inproceedings{Dong2015BoostingGV,
  title={Boosting GPU Virtualization Performance with Hybrid Shadow Page Tables},
  author={Yaozu Dong and Mochi Xue and Xiao Zheng and Jiajun Wang and Zhengwei Qi and Haibing Guan},
  booktitle={USENIX Annual Technical Conference},
  year={2015}
}
The increasing adoption of Graphic Process Unit (GPU) to computation-intensive workloads has stimulated a new computing paradigm called GPU cloud (e.g., Amazon's GPU Cloud), which necessitates the sharing of GPU resources to multiple tenants in a cloud. However, state-of-the-art GPU virtualization techniques such as gVirt still suffer from non-trivial performance overhead for graphics memory-intensive workloads involving frequent page table updates. To understand such overhead, this paper… CONTINUE READING

Results and Topics from this paper.

Key Quantitative Results

  • Evaluation using GMedia shows that gHyvi can achieve up to 13× performance improvement compared to gVirt, and up to 85% native performance for multithread media transcoding.
  • Evaluation using GMedia shows that gHyvi can achieve up to 13x performance improvement compared to gVirt, and up to 85% native performance for multithread media transcoding.
  • Experiments using GMedia on an Intel GPU card show that gHyvi can achieve up to 13x performance improvement compared to gVirt, and up to 85% native performance for multi-thread media transcoding.
  • An evaluation showing that gHyvi achieves up to 85% native performance for multi-thread media transcoding and a 13x speedup over gVirt.

Citations

Publications citing this paper.
SHOWING 1-10 OF 13 CITATIONS

Evolution of Cloud Operating System: From Technology to Ecosystem

  • Journal of Computer Science and Technology
  • 2017
VIEW 4 EXCERPTS
CITES METHODS
HIGHLY INFLUENCED

Experiments about the Performance of GPU Oversubscription Strategy

  • 2019 IEEE 3rd Information Technology, Networking, Electronic and Automation Control Conference (ITNEC)
  • 2019
VIEW 1 EXCERPT
CITES METHODS

Efficient Sharing and Fine-Grained Scheduling of Virtualized GPU Resources

  • 2018 IEEE 38th International Conference on Distributed Computing Systems (ICDCS)
  • 2018
VIEW 2 EXCERPTS
CITES BACKGROUND

Scalable GPU Virtualization with Dynamic Sharing of Graphics Memory Space

  • IEEE Transactions on Parallel and Distributed Systems
  • 2018
VIEW 1 EXCERPT
CITES METHODS

Building Media-Rich Cloud Services from Network-Attached I/O Devices

  • 2016 IEEE 4th International Conference on Future Internet of Things and Cloud (FiCloud)
  • 2016
VIEW 2 EXCERPTS
CITES BACKGROUND

References

Publications referenced by this paper.
SHOWING 1-10 OF 25 REFERENCES

LoGV: Low-Overhead GPGPU Virtualization

  • 2013 IEEE 10th International Conference on High Performance Computing and Communications & 2013 IEEE International Conference on Embedded and Ubiquitous Computing
  • 2013
VIEW 13 EXCERPTS
HIGHLY INFLUENTIAL

Rodinia: A benchmark suite for heterogeneous computing

  • 2009 IEEE International Symposium on Workload Characterization (IISWC)
  • 2009
VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL

GPUvm: Why Not Virtualizing GPUs at the Hypervisor?

  • USENIX Annual Technical Conference
  • 2014
VIEW 6 EXCERPTS
HIGHLY INFLUENTIAL

Gpu - accelerated high - performance computing in virtual machines

J. A. STRATTON, C. RODRIGUES, +4 authors G. D. LIU
  • Computers , IEEE Transactions on
  • 2012

Parboil: A Revised Benchmark Suite for Scientific and Commercial Throughput Computing

John A. Stratton, Christopher I. Rodrigues, +5 authors Wen-mei W. Hwu
  • 2012
VIEW 3 EXCERPTS

An efficient open - source gpu machine learning library

Z. QI, J. YAO, +3 authors H. AND GUAN
  • International Journal of Computer Information Systems and Industrial Management Applications
  • 2011