Page History

...

The current pilot infrastructure at ECMWF features 2x5 NVIDIA Tesla V100 32 x A100 80 GB cards targeting Machine Learning workloads. They are exposed as Virtual GPUS to the instances on the cloud, which allows for multiple VMs to transparently share the same Physical GPU card.

How to provision a GPU-enabled instance

Once your tenant is granted access to the GPUs, creating a new VM with access to a virtual GPU is very straightforward. Follow the process on Provision a new instance - web paying special attention on the Configuration Step:

CentOSRocky

1. On the Library screen, choose CentOS. 
2. On Layout, select the item with "-gpu" suffix (e.g.: "centosrocky-79.92-gpu" )
3. On Plan, pick one of the plans with the "gpu" suffix, depending on how much resources are needed, including the amount of GPU memory:
   
   1. 8cpu-4gbmem64gbmem-20gbdisk-4gbgpu30gbdisk-a100.1g.10gbgpu
   2. 8cpu-8gbmem64gbmem-20gbdisk-4gbgpu
   3. 8cpu-32gbmem-40gbdisk-8gbgpu
   4. 16cpu-32gbmem-80gbdisk-16gbgpu

Image Removed

Ubuntu

1. On the Library screen, choose Ubuntu.
2. On the Configuration Options select Version 20.04
3. On Layout, select the item with "-gpu" suffix (i.e.: "ubuntu-20.04-gpu" )
4. On Plan, for the Ubuntu instance select the following plan :  16cpu-32gbmem-80gbdisk-16gbgpu  ( * )

Image Removed

( * ) Note:  the "16cpu-32gbmem-80gbdisk-16gbgpu" plan has limited availability in the current EWC infrastructure. Please contact EWC team via Support Portal if you encounter any problem during the deployment as it might be caused by the temporary lack of availability.

1. 1. 30gbdisk-a100.2g.20gbgpu
   2. 16cpu-128gbmem-30gbdisk-40gbgpu    
   3. 48cpu-384gbmem-30gbdisk-80gbgpu ( * )

( * ) The latest plan "48cpu-384gbmem-30gbdisk-80gbgpu" is only available upon request for a limited amount of time for justified use cases requirements.

Image Added

$> nvidia-smi 
WedMon MarOct 2901 09:17:13 2023       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450525.156105.0017   Driver Version: 450525.156105.0017   CUDA Version: 1112.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GRID V100A100D-16C40C       On   | 00000000:00:05.0 Off |                    0 |
| N/A   N/A    P0    N/A /  N/A |   1168MiB5MiB / 16384MiB40960MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

Related articles

...