Exploiting GPUs in Virtual Machine for BioCloud

Heeseung Jo,Jinkyu Jeong,Dong Hoon Choi,Myoungho Lee

doi:10.1155/2013/939460

Abstract

Recently, biological applications start to be reimplemented into the applications which exploit many cores of GPUs for better computation performance. Therefore, by providing virtualized GPUs to VMs in cloud computing environment, many biological applications will willingly move into cloud environment to enhance their computation performance and utilize infinite cloud computing resource while reducing expenses for computations. In this paper, we propose a BioCloud system architecture that enables VMs to use GPUs in cloud environment. Because much of the previous research has focused on the sharing mechanism of GPUs among VMs, they cannot achieve enough performance for biological applications of which computation throughput is more crucial rather than sharing. The proposed system exploits the pass-through mode of PCI express (PCI-E) channel. By making each VM be able to access underlying GPUs directly, applications can show almost the same performance as when those are in native environment. In addition, our scheme multiplexes GPUs by using hot plug-in/out device features of PCI-E channel. By adding or removing GPUs in each VM in on-demand manner, VMs in the same physical host can time-share their GPUs. We implemented the proposed system using the Xen VMM and NVIDIA GPUs and showed that our prototype is highly effective for biological GPU applications in cloud environment.

Highlights

Virtualization technology has been widely adopted into computing systems to increase hardware resource utilization and reduce total cost of ownership (TCO)
(i) Reimplementation and low flexibility of graphic processing units (GPUs) application programming interfaces (APIs): for sharing of GPU among virtual machines (VMs), the management of GPU is concentrated on host operating system or virtual machine monitor (VMM), and the communication between virtual device drivers is highly dependent on the implementation of them
The GPU-Manager consists of three parts: the AdminListener thread which handles the request of the PoolChecker, the WrapCUDA library hooking the API calls of the native CUDA library to support the implicit allocation or deallocation, and the RequestSender to provide the explicit user interface

Summary

Introduction

Virtualization technology has been widely adopted into computing systems to increase hardware resource utilization and reduce total cost of ownership (TCO). Virtualization technology enables multiple computing environments to be consolidated in a single physical machine This consolidation brings efficient use of hardware resources and flexible resource provisioning to each computing environment [1]. Biological applications, which require high performance computing environment, are moving into cloud computing environment due to the narrowed performance gap and the advantage of flexible resource provisioning [6,7,8]. Data copies between host memory and internal memory of GPU could be overhead of GPU computing. In the application processing unit (APU) of AMD, CPU and GPU are integrated in a single chipset, and the two computing units share the same memory controller so that GPU can access the host memory. From the high performance computing power of GPUs, biological applications hosted in cloud can show high performance while minimizing TCO of their computing infrastructure.

Background and Related Work

GPU Virtualization

Evaluation

Sharing Effect

Discussion

Conclusions