Remote GPU Virtualization: Is It Useful?

Silla, Federico; Prades, Javier; Iserte, Sergio; Reaño, Carlos

doi:10.1109/hipineb.2016.8

Cited by 26 publications

(10 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…GPU-as-a-Service Two other efforts, DS-CUDA [47] and rCUDA [48], have focused on providing access to a remote GPU for the purposes of GPU-as-a-Service [49][50][51][52][53][54][55]. They also rely on a proxy process.…”

Section: Related Workmentioning

confidence: 99%

CRUM: Checkpoint-Restart Support for CUDA's Unified Memory

Garg

Mohan

Sullivan

et al. 2018

2018 IEEE International Conference on Cluster Computing (CLUSTER)

View full text Add to dashboard Cite

Unified Virtual Memory (UVM) was recently introduced on recent NVIDIA GPUs. Through software and hardware support, UVM provides a coherent shared memory across the entire heterogeneous node, migrating data as appropriate. The older CUDA programming style is akin to older large-memory UNIX applications which used to directly load and unload memory segments. Newer CUDA programs have started taking advantage of UVM for the same reasons of superior programmability that UNIX applications long ago switched to assuming the presence of virtual memory. Therefore, checkpointing of UVM will become increasingly important, especially as NVIDIA CUDA continues to gain wider popularity: 87 of the top 500 supercomputers in the latest listings are GPU-accelerated, with a current trend of ten additional GPU-based supercomputers each year.A new scalable checkpointing mechanism, CRUM (Checkpoint-Restart for Unified Memory), is demonstrated for hybrid CUDA/MPI computations across multiple computer nodes. CRUM supports a fast, forked checkpointing, which mostly overlaps the CUDA computation with storage of the checkpoint image in stable storage. The runtime overhead of using CRUM is 6% on average, and the time for forked checkpointing is seen to be a factor of up to 40 times less than traditional, synchronous checkpointing.

show abstract

Section: Related Workmentioning

confidence: 99%

CRUM: Checkpoint-Restart Support for CUDA's Unified Memory

Garg

Mohan

Sullivan

et al. 2018

2018 IEEE International Conference on Cluster Computing (CLUSTER)

View full text Add to dashboard Cite

show abstract

“…One of the most prominent solutions related to concurrent remote usage of CUDA-enabled devices in a transparent way is rCUDA. 32 Thanks to the split-driver approach, there is no need to modify and recompile the CUDA-enabled application in order to use it with rCUDA. Indeed, the framework takes care of all the necessary details in order to execute the CUDA kernels on a remote or local GPGPU.…”

Section: Related Workmentioning

confidence: 99%

“…One of the most prominent solutions related to concurrent remote usage of CUDA‐enabled devices in a transparent way is rCUDA . Thanks to the split‐driver approach, there is no need to modify and recompile the CUDA‐enabled application in order to use it with rCUDA.…”

Section: Related Workmentioning

confidence: 99%

Accelerating Linux and Android applications on low‐power devices through remote GPGPU offloading

Montella

Kosta

Oro³

et al. 2017

Concurrency and Computation

View full text Add to dashboard Cite

Summary Low‐power devices are usually highly constrained in terms of CPU computing power, memory, and GPGPU resources for real‐time applications to run. In this paper, we describe RAPID, a complete framework suite for computation offloading to help low‐powered devices overcome these limitations. RAPID supports CPU and GPGPU computation offloading on Linux and Android devices. Moreover, the framework implements lightweight secure data transmission of the offloading operations. We present the architecture of the framework, showing the integration of the CPU and GPGPU offloading modules. We show by extensive experiments that the overhead introduced by the security layer is negligible. We present the first benchmark results showing that Java/Android GPGPU code offloading is possible. Finally, we show the adoption of the GPGPU offloading into BioSurveillance, a commercial real‐time face recognition application. The results show that, thanks to RAPID, BioSurveillance is being successfully adapted to run on low‐power devices. The proposed framework is highly modular and exposes a rich application programming interface to developers, making it highly versatile while hiding the complexity of the underlying networking layer.

show abstract

“…By virtualizing GPUs, their utilization is increased [21], similarly to what happens with traditional CPU virtualization. Furthermore, it is also possible to enhance GPU virtualization by adding a new feature to it: remote access [22]. In this way, it is possible to virtualize GPUs and provide them to applications that are being executed in other nodes of the cluster.…”

Section: Introductionmentioning

confidence: 99%

Improving the performance of physics applications in atom-based clusters with rCUDA

Silla

Prades

Baydal

et al. 2020

Journal of Parallel and Distributed Computing

Self Cite

View full text Add to dashboard Cite

Traditionally, High-Performance Computing (HPC) has been associated with large power requirements. The reason was that chip makers of the processors typically employed in HPC deployments have always focused on getting the highest performance from their designs, regardless of the energy their processors may consume. Actually, for many years only heat dissipation was the real barrier for achieving higher performance, at the cost of higher energy consumption. However, a new trend has recently appeared consisting on the use of low-power processors for HPC purposes. The MontBlanc and Isambard projects are good examples of this trend. These proposals, however, do not consider the use of GPUs. In this paper we propose to use GPUs in this kind of low-power processor based HPC deployments by making use of the remote GPU virtualization mechanism. To that end, we leverage the rCUDA middleware in a hybrid cluster composed of low-power Atom-based nodes and regular Xeon-based nodes equipped with GPUs. Our experiments show that, by making use of rCUDA, the execution time of applications belonging to the physics domain is noticeably reduced, achieving a speed up of up to 140x with just one remote NVIDIA V100 GPU with respect to the execution of the same applications using 8 Atom-based nodes. Additionally, a rough energy consumption estimation reports improvements in energy demands of up to 37x.

show abstract

Remote GPU Virtualization: Is It Useful?

Cited by 26 publications

References 25 publications

CRUM: Checkpoint-Restart Support for CUDA's Unified Memory

CRUM: Checkpoint-Restart Support for CUDA's Unified Memory

Accelerating Linux and Android applications on low‐power devices through remote GPGPU offloading

Improving the performance of physics applications in atom-based clusters with rCUDA

Contact Info

Product

Resources

About