No relevant resource is found in the selected language.

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>Search

Reminder

To have a better experience, please upgrade your IE browser.

upgrade

FusionCloud 6.3.1.1 Solution Description 04

Rate and give feedback:
Huawei uses machine translation combined with human proofreading to translate this document to different languages in order to help you better understand the content of this document. Note: Even the most advanced machine translation cannot match the quality of professional translators. Huawei shall not bear any responsibility for translation accuracy and it is recommended that you refer to the English document (a link for which has been provided).
GPU-accelerated ECSs

GPU-accelerated ECSs

GPU ECSs provide outstanding floating-point computing capabilities. They are suitable for scenarios that require real-time, highly concurrent massive computing. GPU-type ECSs are divided into graphics-accelerated ECSs (G series, whose virtualization type is Xen) and computing-accelerated ECSs (P series, whose virtualization type is KVM). In the preceding information:

  • G series ECSs are suitable for 3D animation rendering and CAD.

    G series ECSs are divided into G1 ECSs, which use GPU virtualization, and G2 ECSs, which use GPU SR-IOV.

  • P series ECSs are designed for deep learning, scientific computing, and CAE.

Application Scenarios

  • Applications

    Deep learning, scientific computing, 3D animation rendering, CAD, and CAE

  • Scenario characteristics

    Highly real-time, highly concurrent, and massive computing

  • Application environments
    • P series ECSs are suitable for artificial intelligence: Each GPU contains thousands of computing units, providing outstanding parallel computing capabilities. P series ECSs have been optimized for deep learning, supporting massive computing within a short period of time.
    • P series ECSs are suitable for scientific computing: Scientific computing has strict requirements on double-precision computing. During computing emulation, a large number of compute resources are used, and large volumes of data are generated. Therefore, scientific computing also has strict requirements on storage bandwidth and latency. P series ECSs meet these requirements.
    • G series ECSs are suitable for graphic workstation: G series ECSs provide outstanding computing capabilities for professional CAD, video rendering, and graphics processing.

Model

The GPU vendor is NVIDIA whose vendor_id is 0x10de. Table 18-6 describes the GPU models that support GPU ECSs.

Table 18-6 GPU models

GPU ECS

GPU Model

product_id

API Type

Alias Name

Description

G1/G2 ECSs (graphics-accelerated)

NVIDIA Tesla M40

0x17fd

PCIe 3.0 x16

nvidia-m40

12 GB video RAM

NVIDIA Tesla M60

0x13f2

PCIe 3.0 Dual Slot

nvidia-m60

Two cores with 8 GB video RAM for each, two slots

PI1 ECSs (computing-accelerated)

NVIDIA Tesla P4

0x1bb3

PCIe 3.0

nvidia-p4

8 GB video RAM

PI2 ECSs (computing-accelerated)

NVIDIA Tesla P40

0x1b38

PCIe 3.0 Dual Slot

nvidia-p40

24 GB video RAM, two slots

P1 ECSs (computing-accelerated)

Tesla P100 PCIe 12GB

0x15f7

PCIe 3.0

nvidia-p100-pcie-12

12 GB video RAM, two slots

Tesla P100 PCIe 16GB

0x15f8

PCIe 3.0

nvidia-p100-pcie-16

16 GB video RAM, two slots

P2 ECSs (computing-accelerated)

Tesla V100 PCIe 16GB

0x1db4

PCIe 3.0

nvidia-v100-pcie-16

16 GB video RAM, two slots

Tesla V100 PCIe 32GB

0x1db6

PCIe 3.0

nvidia-v100-pcie-32

32 GB video RAM, two slots

NOTE:
  • If FCD is used to set up the environment during installation and GPU ECSs were not planned previously but you need to use them after the environment is set up, log in to the physical servers to enable intel_iommu. For details, see Product Management > FusionSphere OpenStack > Configuring GPU and USB Parameters of KVM-Based Compute Nodes in FusionCloud 6.3.1.1 O&M Guide.
  • To use GPU ECSs, log in to DMK and check whether the value of is_supported_gpu_ecs is true.

Graphics-accelerated ECSs

Graphics-accelerated ECSs (G series ECSs, whose virtualization type is Xen) are divided into G1 ECSs, which use GPU virtualization, and G2 ECSs, which use GPU SR-IOV.

  • G1 ECSs are based on NVIDIA GRID virtual GPUs and provide economical graphics acceleration. G1 ECSs use the NVIDIA Tesla M60 GPU and support DirectX and OpenGL. The ECSs have a maximum of 8 GB video RAM and 4,096 x 2,160 resolution, and are suitable for applications that require high performance in graphics rendering.
  • G2 ECSs are based on NVIDIA Tesla M60 hardware passthrough and provide graphics acceleration and single-precision computing with a maximum of 8 GB GPU memory and 4,096 x 2,160 resolution. They support DirectX, OpenGL, CUDA, and OpenCL, provide 2,048 CUDA cores, and are suitable for media editing, 3D rendering, and transcoding.

Scope of Support

  • G1 ECSs support:
    • NVIDIA M60 GPUs
    • Graphics acceleration applications
    • GPU hardware virtualization (vGPUs)
    • Application flow identical to common ECSs
    • A maximum of 8 GB GPU memory and 4,096 x 2,160 resolution for processing videos
  • G2 ECSs support:
    • NVIDIA M60 GPUs
    • Graphics acceleration applications
    • GPU hardware passthrough
    • Enhanced SR-IOV network performance and high bandwidths
    • A maximum of 8 GB GPU memory and 4,096 x 2,160 resolution for processing videos
    • DirectX, OpenGL, CUDA, and OpenCL
    • Up to 2048 CUDA cores

Notes

  • G1 ECSs
    • G1 ECSs do not support flavor change.
    • G1 ECSs support the following OSs:
      • Windows Server 2008 R2 Enterprise SP1 64bit
      • Windows Server 2012 R2 Standard 64bit
      • Windows Server 2016 DataCenter
      • Windows Server 2008
      • Windows Server 2012
      • Windows Server 2016
    • If a G1 ECS is created using a private image, install a GPU driver on the ECS after the ECS creation. To download the driver, log in at http://www.nvidia.com/grid-eval, set the NVIDIA GRID version to 4.1, and select the GRID for UVP software package. The operations are as follows:
      1. Check whether NVIDIA is used for the first time:
        1. If yes, go to 2.
        2. If no, go to 4.
      2. Obtain the Product Activation Key (PAK) from the email indicating successful registration with NVIDIA, as shown in Figure 18-1.
        Figure 18-1 PAK
      1. Enter the PAK obtained in 1.b on the Redeem Product Activation Keys page and click Redeem.
        Figure 18-2 Redeem Product Activation Keys
      1. Click the Archived Versions tab.
        Figure 18-3 Logging in to the official NVIDIA website
      2. Log in at the official NVIDIA website as prompted and choose Software & Services > Product Information.

      3. Click the Archived Versions tab.
      4. Click NVIDIA GRID of version 4.1.
      5. On the Product Download page, click GRID for UVP.
    • If you log in to a G1 ECS using MSTSC, graphics acceleration will fail. This is because MSTSC replaces the WDDM GPU driver with a non-accelerated remote desktop display driver. In such an event, you must use other methods to log in to the ECS, such as VNC.

      If the remote login function available on the management console fails to meet your service requirements, you must install a suitable remote login tool on the ECS.

  • G2 ECSs
    • G2 ECSs do not support flavor change.
    • G2 ECSs support the following OSs:
      • Windows Server 2008 R2 Enterprise SP1 64bit
      • Windows Server 2012 R2 Standard 64bit
      • Windows Server 2008
      • Windows Server 2012
    • If a G2 ECS is created using a private image, install a GPU driver during the private image creation. Otherwise, install the GPU driver after creating the ECS.

      To download the GPU driver, log in at http://www.nvidia.com/Download/index.aspx?lang=en-us. You are advised to select the latest CUDA toolkit version.

    After the GPU driver is installed, run the following command to switch the GPU working mode and restart the ECS (for example, the GPU driver is installed in C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe):

    "C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe" -dm 0

    • If a G2 ECS is created using a private image, install an SR-IOV driver during the private image creation. Otherwise, install it after the ECS is created.

      To download the SR-IOV driver, log in at https://downloadcenter.intel.com/search?keyword=Intel++Ethernet+Connections+CD. You are advised to select version 20.4.1 or later.

    • If you log in to a G2 ECS using MSTSC, graphics acceleration will fail. This is because MSTSC replaces the WDDM GPU driver with a non-accelerated remote desktop display driver. In such an event, you must use other methods to log in to the ECS, such as VNC.
    • G2 ECSs do not support remote login. If you need to log in to a G2 ECS using VNC, install a VNC server on the ECS before login.

Flavors

Table 18-7 Xen-based GPU ECSs

Type

vCPU

Memory (GB)

Flavor Name

Basic graphics processing G1

4

8

g1.xlarge

8

16

g1.2xlarge

16

32

g1.4xlarge

Accelerated graphics processing G2

8

64

g2.2xlarge

Computing-accelerated ECSs

Computing-accelerated ECSs are divided into PI1, PI2, P1, and P2 ECSs, whose details are as follows:

  • PI1 ECSs use NVIDIA Tesla P4 GPUs dedicated for AI inference. Working with INT8 calculators of NVIDIA Tesla P4 GPUs, PI1 ECSs have shortened the inference latency by 15 times.
  • PI2 ECSs use NVIDIA Tesla P40 GPUs dedicated for ultra-high AI inference throughput, providing outstanding real-time inference. The single-precision floating-point computing performance and INT8 integer computing performance of NVIDIA Tesla P40 GPUs are twice times faster than those of NVIDIA Tesla P4 GPUs. When processing deep learning workloads, a server configured with eight NVIDIA Tesla P40 GPUs equals over 100 servers configured only with CPUs.
  • P1 ECSs use NVIDIA Tesla P100 GPUs and provide flexibility, high performance, and cost-effectiveness. These ECSs support GPU Direct for direct communication between GPUs, improving data transmission efficiency. P1 ECSs provide outstanding universal computing capabilities and have strengths in deep learning, graphic databases, high-performance databases, Computational Fluid Dynamics (CFD), computing finance, seismic analysis, molecular modeling, and genomics. They are designed for scientific computing.
  • Compared with P1 ECSs, P2 ECSs use NVIDIA Tesla V100 GPUs, which have improved both single- and double-precision computing capabilities by 50% and offer 112 TFLOPS of deep learning.

PI1 ECSs

Features

PI1 ECSs have the following features:
  • NVIDIA Tesla P4 GPUs
  • Up to 5.5 TFLOPS by a single GPU for single precision
  • Up to 22 TOPS INT8 by a single GPU
  • 8 GB ECC GPU memory with a bandwidth of 192 Gbit/s by a single GPU
  • Hardware video encoding and decoding engines embedded in GPUs for concurrent real-time 35-channel HD video transcoding and inference
Notes
  • Ensure that the NVIDIA driver has been installed on the image used to create a PI1 ECS. If not, install the driver after the PI1 ECS is created for computing acceleration. To download the driver, log in at http://www.nvidia.com/Download/Find.aspx?lang=en. Figure 18-4 shows parameter settings.
    Figure 18-4 NVIDIA Driver Downloads
  • PI1 ECSs do not support flavor change.
  • PI1 ECSs support the following OSs:
    • Ubuntu Server 14.04 64bit
    • CentOS 7.3 64bit
Flavors
Table 18-8 Flavors of PI1 computing-accelerated ECSs

Type

vCPU

Memory (GB)

Flavor Name

GPU

PI1 computing-accelerated ECSs

8

64

PI1.2xlarge.8

1×P4

16

128

PI1.4xlarge.8

2×P4

32

256

PI1.8xlarge.8

4×P4

PI2 ECSs

Notes
  • Ensure that the NVIDIA driver has been installed on the image used to create a PI2 ECS. If not, install the driver after the PI2 ECS is created for computing acceleration.
  • PI2 ECSs support the following OSs:
    • Ubuntu Server 16.04 64bit
    • Debian 9.0 64bit
    • CentOS 7.4 64bit
    • EulerOS 2.2
    • Window Server 2012 R2 64-bit (Enterprise Edition)
Flavors
Table 18-9 Flavors of PI2 computing-accelerated ECSs

Type

vCPU

Memory (GB)

Flavor Name

GPU

PI2 computing-accelerated

8

64

PI2.2xlarge.8

1×P40

16

128

PI2.4xlarge.8

2×P40

32

256

PI2.8xlarge.8

4×P40

P1 ECSs

Features

P1 ECSs have the following features:
  • NVIDIA Tesla P100 GPUs
  • 9.3 TFLOPS for single precision and 4.7 TFLOPS for double precision
  • Comprehensive basic capabilities

    Networks are user-defined, subnets can be divided, and network access policies can be configured as needed. Mass storage is used, and elastic capacity expansion as well as backup and restoration is supported to make data more secure. Auto Scaling allows you to add or reduce the number of ECSs quickly.

  • Excellent supercomputing ecosystem

    The supercomputing ecosystem allows you to build up a flexible, high-performance, cost-effective computing platform. A large number of HPC applications and deep-learning frameworks can run on P1 ECSs.

Notes
  • Ensure that the NVIDIA driver has been installed on the image used to create a P1 ECS. If not, install the driver after the P1 ECS is created for computing acceleration.
  • P1 ECSs support the following OSs:
    • Debian 9.0 64bit
    • Ubuntu Server 16.04 64bit
    • CentOS 7.4 64bit
    • Debian 8.0 64bit
    • CentOS 7.3 64bit
    • EulerOS 2.2 64bit
    • Windows Server 2012 R2 Standard 64bit
Flavors
Table 18-10 Flavors of P1 computing-accelerated ECSs

Type

vCPU

Memory (GB)

Flavor Name

GPU

P1 computing-accelerated

8

64

P1.2xlarge.8

1×P100

16

128

P1.4xlarge.8

2×P100

32

256

P1.8xlarge.8

4×P100

P2 ECSs

Features

P2 ECSs have the following features:
  • NVIDIA Tesla V100 GPUs
  • 14 TFLOPS of single-precision computing, 7 TFLOPS of double-precision computing, and 112 TFLOPS of deep learning
  • Comprehensive basic capabilities

    Networks are user-defined, subnets can be divided, and network access policies can be configured as needed. Mass storage is used, and elastic capacity expansion as well as backup and restoration is supported to make data more secure. Auto Scaling allows you to add or reduce the number of ECSs quickly.

  • Flexibility

    Similar to other types of ECSs, P2 ECSs can be provisioned in a few minutes.

  • Excellent supercomputing ecosystem

    The supercomputing ecosystem allows you to build up a flexible, high-performance, cost-effective computing platform. A large number of HPC applications and deep-learning frameworks can run on P2 ECSs.

Notes
  • Ensure that the NVIDIA driver has been installed on the image used to create a P2 ECS. If not, install the driver after the P2 ECS is created for computing acceleration.
  • P2 ECSs support the following OSs:
    • Ubuntu Server 16.04 64bit
    • EulerOS 2.2 64bit
Flavors
Table 18-11 Flavors of P2 computing-accelerated ECSs

Type

vCPU

Memory (GB)

Flavor Name

GPU

P2 computing-accelerated

8

64

P2.2xlarge.8

1×V100

16

128

P2.4xlarge.8

2×V100

32

256

P2.8xlarge.8

4×V100

Translation
Download
Updated: 2019-10-23

Document ID: EDOC1100063247

Views: 73707

Downloads: 192

Average rating:
This Document Applies to these Products
Related Version
Related Documents
Share
Previous Next