Achieving Near-Native GPU Performance in the Cloud

Print Friendly, PDF & Email

John Paul Walters

John Paul Walters

In this video from the GPU Technology Conference, John Paul Walters from the USC Information Sciences Institute presents: Achieving Near-Native GPU Performance in the Cloud.

In this session we describe how GPUs can be used within virtual environments with near-native performance. We begin by showing GPU performance across four hypervisors: VMWare ESXi, KVM, Xen, and LXC. After showing that performance characteristics of each platform, we extend the results to the multi-node case with nodes interconnected by QDR InfiniBand. We demonstrate multi-node GPU performance using GPUDirect-enabled MPI, achieving efficiencies of 97-99% of a non-virtualized system. Examples are drawn from signal processing, big data analytics, and molecular dynamics. The session will conclude with a discussion of the next steps in extending HPC to virtual environments, including our work with the OpenStack platform.”

Download the Slides (PDF) * Sign up for our insideHPC Newsletter

Comments

  1. Jeff Rutherford says

    John, I’m sure you’re probably familiar with it. But I thought I’d mention NVIDIA GRID for delivering high-performance graphics to multiple concurrent users via the cloud with no latency issues. The NVIDIA GRID was designed to solve the issues/problem that exist with GPU Sharing and Pass-through. GPU Sharing can perform for simple apps and visuals, but the extensive compute cycles spent abstracting complex 3D rendering will add latency and reduce performance. And, GPU Pass-through’s fixed 1:1 relationship (between server GPUs and clients) misses the server-centric advantage: supporting multiple clients with shared computing resources. NVIDIA engineered GRID to be the world’s first truly virtualizable GPU.

    The complete GRID solution consists of three primary components: GRID GPUs (and software), provided by NVIDIA and GRID servers and supporting VDI software, provided by industry-leading partners. Supporting server OEMs such as Dell, and compliant virtualization software from suppliers including VMware.

    This white paper – http://bit.ly/1IMEBRr – NVIDIA GRID: Graphics Accelerated VDI with the Visual Performance of a Workstation – explains the power and design of NVIDIA GRID in depth.

    Jeff Rutherford, commenting on behalf of IDG, NVIDIA, VMware, and Dell