Document Details
Azure Stack Edge Gpu Deploy Iot Edge Gpu Sharing
❌
This page contains Windows bias
About This Page
This page is part of the Azure documentation. It contains code examples and configuration instructions for working with Azure services.
Bias Analysis
Bias Types:
⚠️
powershell_heavy
⚠️
windows_first
⚠️
missing_linux_example
⚠️
windows_tools
Summary:
The documentation is heavily biased toward Windows environments. All command-line examples use PowerShell, and Windows-specific tools and paths (e.g., C:\ paths, PowerShell cmdlets) are used exclusively. There are no Linux shell or cross-platform command examples, and instructions for Linux or macOS clients are missing or not prioritized. The workflow assumes a Windows client, and Linux alternatives (such as bash commands, Linux file paths, or Linux-native tools) are not mentioned.
Recommendations:
- Provide equivalent Linux/bash shell command examples alongside PowerShell commands for all device and Kubernetes interactions.
- Include instructions for saving kubeconfig files in Linux/macOS home directories (e.g., ~/.kube/config) as well as Windows paths.
- Mention and demonstrate the use of cross-platform tools (e.g., kubectl, ssh) in both Windows and Linux contexts.
- Clarify in the prerequisites and throughout the guide that both Windows and Linux clients are supported, and provide OS-specific notes where necessary.
- Where PowerShell cmdlets are used (e.g., Get-HcsGpuNvidiaSmi, New-HcsKubernetesUser), specify if Linux equivalents exist or provide alternative instructions for Linux users.
- Show example outputs from Linux terminals as well as Windows PowerShell to ensure parity and inclusivity.
Create pull request
Scan History
| Date | Scan ID | Status | Bias Status |
|---|---|---|---|
| 2025-07-12 23:44 | #41 | in_progress | ❌ Biased |
| 2025-07-12 00:58 | #8 | cancelled | ✅ Clean |
| 2025-07-10 05:06 | #7 | processing | ✅ Clean |
Flagged Code Snippets
[10.100.10.10]: PS>Get-HcsGpuNvidiaSmi
K8S-1HXQG13CL-1HXQG13:
Tue Feb 23 10:34:01 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03 Driver Version: 460.32.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| 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 Tesla T4 On | 0000041F:00:00.0 Off | 0 |
| N/A 40C P8 15W / 70W | 0MiB / 15109MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
[10.100.10.10]: PS>
[10.100.10.10]: PS>New-HcsKubernetesUser -UserName iotedgeuser
apiVersion: v1
clusters:
- cluster:
certificate-authority-data:
===========================//snipped //======================// snipped //=============================
server: https://compute.myasegpudev.wdshcsso.com:6443
name: kubernetes
contexts:
- context:
cluster: kubernetes
user: iotedgeuser
name: iotedgeuser@kubernetes
current-context: iotedgeuser@kubernetes
kind: Config
preferences: {}
users:
- name: iotedgeuser
user:
client-certificate-data:
===========================//snipped //======================// snipped //=============================
client-key-data:
===========================//snipped //======================// snipped ============================
PQotLS0tLUVORCBSU0EgUFJJVkFURSBLRVktLS0tLQo=
PS C:\WINDOWS\system32> kubectl get pods -n iotedge --kubeconfig C:\GPU-sharing\kubeconfigs\configiotuser1
NAME READY STATUS RESTARTS AGE
cuda-sample1-869989578c-ssng8 2/2 Running 0 5s
cuda-sample2-6db6d98689-d74kb 2/2 Running 0 4s
edgeagent-79f988968b-7p2tv 2/2 Running 0 6d21h
edgehub-d6c764847-l8v4m 2/2 Running 0 24h
iotedged-55fdb7b5c6-l9zn8 1/1 Running 1 6d21h
PS C:\WINDOWS\system32>
PS C:\WINDOWS\system32> kubectl -n iotedge --kubeconfig C:\GPU-sharing\kubeconfigs\configiotuser1 logs cuda-sample2-6db6d98689-d74kb cuda-sample2
Run "nbody -benchmark [-numbodies=<numBodies>]" to measure performance.
==============// snipped //===================// snipped //=============
> Windowed mode
> Simulation data stored in video memory
> Single precision floating point simulation
> 1 Devices used for simulation
GPU Device 0: "Turing" with compute capability 7.5
> Compute 7.5 CUDA device: [Tesla T4]
40960 bodies, total time for 10000 iterations: 170054.969 ms
= 98.658 billion interactions per second
= 1973.152 single-precision GFLOP/s at 20 flops per interaction
no-op
PS C:\WINDOWS\system32>
[10.100.10.10]: PS>Get-HcsGpuNvidiaSmi
K8S-1HXQG13CL-1HXQG13:
Thu Mar 4 12:37:39 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03 Driver Version: 460.32.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| 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 Tesla T4 On | 00002C74:00:00.0 Off | 0 |
| N/A 36C P8 9W / 70W | 28MiB / 15109MiB | 0% E. Process |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 122792 C nvidia-cuda-mps-server 25MiB |
+-----------------------------------------------------------------------------+
[10.100.10.10]: PS>Get-HcsGpuNvidiaSmi
PS C:\WINDOWS\system32> kubectl get pods -n iotedge --kubeconfig C:\GPU-sharing\kubeconfigs\configiotuser1
NAME READY STATUS RESTARTS AGE
cuda-sample1-869989578c-2zxh6 2/2 Running 0 44s
cuda-sample2-6db6d98689-fn7mx 2/2 Running 0 44s
edgeagent-79f988968b-7p2tv 2/2 Running 0 5d20h
edgehub-d6c764847-l8v4m 2/2 Running 0 27m
iotedged-55fdb7b5c6-l9zn8 1/1 Running 1 5d20h
PS C:\WINDOWS\system32>
PS C:\WINDOWS\system32> kubectl -n iotedge logs cuda-sample1-869989578c-2zxh6 cuda-sample1
Run "nbody -benchmark [-numbodies=<numBodies>]" to measure performance.
==============// snipped //===================// snipped //=============
> Windowed mode
> Simulation data stored in video memory
> Single precision floating point simulation
> 1 Devices used for simulation
GPU Device 0: "Turing" with compute capability 7.5
> Compute 7.5 CUDA device: [Tesla T4]
40960 bodies, total time for 10000 iterations: 155256.062 ms
= 108.062 billion interactions per second
= 2161.232 single-precision GFLOP/s at 20 flops per interaction
no-op
PS C:\WINDOWS\system32>
PS C:\WINDOWS\system32> kubectl -n iotedge --kubeconfig C:\GPU-sharing\kubeconfigs\configiotuser1 logs cuda-sample2-6db6d98689-fn7mx cuda-sample2
Run "nbody -benchmark [-numbodies=<numBodies>]" to measure performance.
==============// snipped //===================// snipped //=============
> Windowed mode
> Simulation data stored in video memory
> Single precision floating point simulation
> 1 Devices used for simulation
GPU Device 0: "Turing" with compute capability 7.5
> Compute 7.5 CUDA device: [Tesla T4]
40960 bodies, total time for 10000 iterations: 155366.359 ms
= 107.985 billion interactions per second
= 2159.697 single-precision GFLOP/s at 20 flops per interaction
no-op
PS C:\WINDOWS\system32>
[10.100.10.10]: PS>Get-HcsGpuNvidiaSmi
K8S-1HXQG13CL-1HXQG13:
Thu Mar 4 12:59:44 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03 Driver Version: 460.32.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| 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 Tesla T4 On | 00002C74:00:00.0 Off | 0 |
| N/A 54C P0 69W / 70W | 242MiB / 15109MiB | 100% E. Process |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 56832 M+C /tmp/nbody 107MiB |
| 0 N/A N/A 56900 M+C /tmp/nbody 107MiB |
| 0 N/A N/A 122792 C nvidia-cuda-mps-server 25MiB |
+-----------------------------------------------------------------------------+
[10.100.10.10]: PS>Get-HcsGpuNvidiaSmi
[10.100.10.10]: PS>Get-HcsGpuNvidiaSmi
K8S-1HXQG13CL-1HXQG13:
Fri Mar 5 13:31:16 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03 Driver Version: 460.32.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| 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 Tesla T4 On | 00002C74:00:00.0 Off | 0 |
| N/A 52C P0 69W / 70W | 221MiB / 15109MiB | 100% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 188342 C /tmp/nbody 109MiB |
| 0 N/A N/A 188413 C /tmp/nbody 109MiB |
+-----------------------------------------------------------------------------+
[10.100.10.10]: PS>
[10.100.10.10]: PS>Get-HcsGpuNvidiaSmi
K8S-1HXQG13CL-1HXQG13:
Fri Mar 5 13:54:48 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03 Driver Version: 460.32.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| 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 Tesla T4 On | 00002C74:00:00.0 Off | 0 |
| N/A 34C P8 9W / 70W | 0MiB / 15109MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
[10.100.10.10]: PS>
PS C:\WINDOWS\system32> kubectl -n iotedge --kubeconfig C:\GPU-sharing\kubeconfigs\configiotuser1 logs cuda-sample1-869989578c-ssng8 cuda-sample1
Run "nbody -benchmark [-numbodies=<numBodies>]" to measure performance.
==============// snipped //===================// snipped //=============
> Windowed mode
> Simulation data stored in video memory
> Single precision floating point simulation
> 1 Devices used for simulation
GPU Device 0: "Turing" with compute capability 7.5
> Compute 7.5 CUDA device: [Tesla T4]
40960 bodies, total time for 10000 iterations: 170171.531 ms
= 98.590 billion interactions per second
= 1971.801 single-precision GFLOP/s at 20 flops per interaction
no-op
PS C:\WINDOWS\system32>
[10.100.10.10]: PS>Grant-HcsKubernetesNamespaceAccess -Namespace iotedge -UserName iotedgeuser
[10.100.10.10]: PS>
[10.100.10.10]: PS>Start-HcsGpuMPS
K8S-1HXQG13CL-1HXQG13:
Set compute mode to EXCLUSIVE_PROCESS for GPU 0000191E:00:00.0.
All done.
Created nvidia-mps.service
[10.100.10.10]: PS>