vCUDA: GPU-Accelerated High-Performance Computing in Virtual Machines

Shi, Lin; Chen, Hao; Sun, Jianhua; Li, Kenli

doi:10.1109/tc.2011.112

Cited by 211 publications

(121 citation statements)

References 20 publications

Supporting

Mentioning

117

Contrasting

Unclassified

Order By: Relevance

“…GridCuda [21] supports CUDA 3.2, although it is not publicly available. vCUDA [22] supports the CUDA 3.2 and implements an unspecified subset of the CUDA runtime API. The communication protocol between the node that executes the application and the remote GPU has a considerable overhead, because of the costs incurred during encoding and decoding, which results in a noticeable drop of overall performance.…”

Section: Related Workmentioning

confidence: 99%

Multi-tenant virtual GPUs for optimising performance of a financial risk application

Prades

Varghese

Reaño

et al. 2017

Journal of Parallel and Distributed Computing

View full text Add to dashboard Cite

Graphics Processing Units (GPUs) are becoming popular accelerators in modern High-Performance Computing (HPC) clusters. Installing GPUs on each node of the cluster is not efficient resulting in high costs and power consumption as well as underutilisation of the accelerator. The research reported in this paper is motivated towards the use of few physical GPUs by providing cluster nodes access to remote GPUs on-demand for a financial risk application. We hypothesise that sharing GPUs between several nodes, referred to as multi-tenancy, reduces the execution time and energy consumed by an application. Two data transfer modes between the CPU and the GPUs, namely concurrent and sequential, are explored. The key result from the experiments is that multi-tenancy with few physical GPUs using sequential data transfers lowers the execution time and the energy consumed, thereby improving the overall performance of the application.

show abstract

Section: Related Workmentioning

confidence: 99%

Multi-tenant virtual GPUs for optimising performance of a financial risk application

Prades

Varghese

Reaño

et al. 2017

Journal of Parallel and Distributed Computing

View full text Add to dashboard Cite

show abstract

“…For GPU virtualisation, the multiplexing work is performed in user space, owing to the lack of standard interfaces at the hardware level and driver layer. Precisely, the multiplexer and scheduler are put on the top of CUDA runtime or driver APIs (Gupta et al, 2009;Shi et al, 2012;Giunta et al, 2010). In our case, the coprovisor can perform multiplexing directly at the accelerator driver layer in Dom0.…”

Section: Fpga Virtualisation 221 Coprovisormentioning

confidence: 99%

“…FPGAs have been found to outperform GPUs in many specific applications (Che et al, 2008;Cope et al, 2005). Since 2007, many researchers (Dowty and Sugerman, 2009;Gupta et al, 2009;Shi et al, 2012;Giunta et al, 2010;Ravi et al, 2011;Lagar-Cavilla et al, 2007) have been focusing on making GPUs a shared resource within a virtualised environment, which would allow for adding GPUs to the infrastructure level of cloud computing. But the idea of adding FPGA accelerators to cloud computing (El-Araby et al, 2008;Gonzalez et al, 2012;Huang et al, 2010;Huang and Hsiung, 2013;Lübbers, 2010;Sabeghi and Bertels, 2009;Jain et al, 2014;Byma et al, 2014;Wang et al, 2013) still stays at an exploration stage.…”

Section: Introductionmentioning

confidence: 99%

pvFPGA: paravirtualising an FPGA-based hardware accelerator towards general purpose computing

Wang

Bolić

Parri

2017

IJHPCN

View full text Add to dashboard Cite

This paper presents an ameliorated design of pvFPGA, which is a novel system design solution for virtualising an FPGA-based hardware accelerator by a virtual machine monitor (VMM). The accelerator design on the FPGA can be used for accelerating various applications, regardless of the application computation latencies. In the implementation, we adopt the Xen VMM to build a paravirtualised environment, and a Xilinx Virtex-6 as an FPGA accelerator. The data transferred between the x86 server and the FPGA accelerator through direct memory access (DMA), and a streaming pipeline technique is adopted to improve the efficiency of data transfer. Several solutions to solve streaming pipeline hazards are discussed in this paper. In addition, we propose a technique, hyper-requesting, which enables portions of two requests bidding to different accelerator applications to be processed on the FPGA accelerator simultaneously through DMA context switches, to achieve request level parallelism. The experimental results show that hyper-requesting reduces request turnaround time by up to 80%.Keywords: field-programmable gate array; FPGA; hardware accelerator; virtualisation; hyperrequesting; streaming pipeline; DMA context switch.Reference to this paper should be made as follows: Wang, W., Bolic, M. and Parri, J. (2017) 'pvFPGA: paravirtualising an FPGA-based hardware accelerator towards general purpose computing', Int.

show abstract

“…One example is vCUDA [16]. In this approach, all API function calls on the target OS need to be redirected to the host OS when migration happens.…”

Section: Related Workmentioning

confidence: 99%

Transparent Accelerator Migration in a Virtualized GPU Environment

Xiao

Balaji

Dinan

et al. 2012

2012 12th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (Ccgrid 2012)

View full text Add to dashboard Cite

Abstract-This paper presents a framework to support transparent, live migration of virtual GPU accelerators in a virtualized execution environment. Migration is a critical capability in such environments because it provides support for fault tolerance, ondemand system maintenance, resource management, and load balancing in the mapping of virtual to physical GPUs. Techniques to increase responsiveness and reduce migration overhead are explored. The system is evaluated by using four application kernels and is demonstrated to provide low migration overheads. Through transparent load balancing, our system provides a speedup of 1.7 to 1.9 for three of the four application kernels.

show abstract

vCUDA: GPU-Accelerated High-Performance Computing in Virtual Machines

Cited by 211 publications

References 20 publications

Multi-tenant virtual GPUs for optimising performance of a financial risk application

Multi-tenant virtual GPUs for optimising performance of a financial risk application

pvFPGA: paravirtualising an FPGA-based hardware accelerator towards general purpose computing

Transparent Accelerator Migration in a Virtualized GPU Environment

Contact Info

Product

Resources

About