メインコンテンツにスキップ
バージョン: 2.5

AIアクセラレータ(GPU)

CubeCOS GPUパススルーでAIの真の力を解き放つ

CubeCOSのGPUデバイスパススルーに対する堅牢なサポートを活用し、AIアクセラレータの性能を最大限に引き出しましょう。 現代のAIおよび機械学習ワークロードの厳しい要件を満たすよう設計されたこの機能により、企業は仮想マシン内で専用GPUの純粋な演算能力を直接活用できるようになり、比類のないパフォーマンス、効率性、および柔軟性を確保できます。

ターミナルまたはPuTTYを使用してCLIに接続する

$ ssh admin@IPADDRESS
Warning: Permanently added '192.168.X.X' (ECDSA) to the list of known hosts.
Password:
Welcome to the Cube Appliance
Enter "help" for a list of available commands
controller>

デバイスプロファイルを作成

controller:gpu> device_profile_create
Creating device profile for NVIDIA Corporation GA106 [RTX A2000] (resource unit: 1)
+-------------+------------------------------------------------------------------------------------------------------------------+
| Field | Value |
+-------------+------------------------------------------------------------------------------------------------------------------+
| created_at | 2024-07-18 02:34:24+00:00 |
| updated_at | None |
| uuid | bbe5ea28-deed-4131-9579-7fd9b06de9b4 |
| name | rtx_a2000_1 |
| groups | [{'resources:PGPU': '1', 'trait:CUSTOM_GPU_PRODUCT_ID_2531': 'required', 'trait:CUSTOM_GPU_NVIDIA': 'required'}] |
| description | None |
+-------------+------------------------------------------------------------------------------------------------------------------+

ステータスを確認

controller:gpu> status
+-----------+
| Node View |
+-----------+---------------------------------------

IOMMU: on

GPU IOMMU Group List
IOMMU Group 14: 18:00.0 VGA compatible controller [0300]: NVIDIA Corporation GA106 [RTX A2000] [10de:2531] (rev a1) (driver: nvidia)
IOMMU Group 14: 18:00.1 Audio device [0403]: NVIDIA Corporation GA106 High Definition Audio Controller [10de:228e] (rev a1) (driver: snd_hda_intel)

Thu Jul 18 10:33:50 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.129.03 Driver Version: 535.129.03 CUDA Version: N/A |
|-----------------------------------------+----------------------+----------------------+
| 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 NVIDIA RTX A2000 On | 00000000:18:00.0 Off | Off |
| 30% 42C P8 14W / 70W | 0MiB / 6138MiB | 0% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| No running processes found |
+---------------------------------------------------------------------------------------+

Thu Jul 18 10:33:50 2024
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 535.129.03 Driver Version: 535.129.03 |
|---------------------------------+------------------------------+------------+
| GPU Name | Bus-Id | GPU-Util |
| vGPU ID Name | VM ID VM Name | vGPU-Util |
|=================================+==============================+============|
| 0 NVIDIA RTX A2000 | 00000000:18:00.0 | 0% |
+---------------------------------+------------------------------+------------+

GPU 00000000:18:00.0
Migration capability : No

Virtualization:
No vGPU type configured

Supported vGPU types:

+--------------+
| Cluster View |
+--------------+------------------------------------

Accelerator:
+--------------------------------------+------+--------+----------+-------------------------------------------------------------------+
| uuid | type | vendor | hostname | std_board_info |
+--------------------------------------+------+--------+----------+-------------------------------------------------------------------+
| e24eeaa6-fabd-4262-ae3c-6957b66f5785 | GPU | 10de | cc2 | {"product_id": "2531", "controller": "VGA compatible controller"} |
| 45eeb6b3-02f6-42df-93bd-c278599bc163 | GPU | 10de | cc3 | {"product_id": "2531", "controller": "VGA compatible controller"} |
+--------------------------------------+------+--------+----------+-------------------------------------------------------------------+
GPU Resource Provider:
+--------------------------------------+---------------------+
| uuid | name |
+--------------------------------------+---------------------+
| c419cc67-fbac-3c6c-81c3-cd5188f7a119 | cc2_0000:18:00.0 |
| 00cce127-e7df-3071-99d2-07e62767a049 | cc3_0000:18:00.0 |
+--------------------------------------+---------------------+
Accelerator Request:

Accelerator Device Profile:
+--------------------------------------+-------------+------------------------------------------------------------------------------------------------------------------+-------------+
| uuid | name | groups | description |
+--------------------------------------+-------------+------------------------------------------------------------------------------------------------------------------+-------------+
| bbe5ea28-deed-4131-9579-7fd9b06de9b4 | rtx_a2000_1 | [{'resources:PGPU': '1', 'trait:CUSTOM_GPU_PRODUCT_ID_2531': 'required', 'trait:CUSTOM_GPU_NVIDIA': 'required'}] | None |
+--------------------------------------+-------------+------------------------------------------------------------------------------------------------------------------+-------------+

仮想関数を無効にしてデバイスのパススルーを有効にする

cc2:gpu> virtual_function_disable
set device 10de:2531 at 0000:18:00.0 driver to vfio-pci
set device 10de:228e at 0000:18:00.1 driver to vfio-pci

GPU IOMMU Group List
IOMMU Group 14: 18:00.0 VGA compatible controller [0300]: NVIDIA Corporation GA106 [RTX A2000] [10de:2531] (rev a1) (driver: pcieport)
IOMMU Group 14: 18:00.1 Audio device [0403]: NVIDIA Corporation GA106 High Definition Audio Controller [10de:228e] (rev a1) (driver: pcieport)

追加設定