The Login Enterprise load testing function is typically used to determine the maximum number of desktops that can be hosted on a server (the VDI density). For this testing, the number and profile of vGPUs that could be assigned from the physical GPU to the pool of VDI VMs limited the VDI desktop density.
Login Enterprise provides a scripting function that allows users to performance test almost any application. The MicroDicom MRI and X-ray viewer application had several typical end user actions, including image rotation and zooming, scripted and the application was then added to the standard Login Enterprise Knowledge worker workload. This increased the workload on the L40 GPU over the standard Knowledge worker workload.
The team used the following login and boot conditions:
- Users were logged in every ten seconds
- All desktops were started before users were logged in
The training and inferencing activities generated the workload on the NVIDIA L40 GPU that was assigned to the AI virtual machine.
The following table describes the applications that make up the Login Enterprise workload and AI dataset that was used for all testing:
| Workload Name | Application Set |
| Modified Knowledge Worker | Office 365 Apps (Edge, Outlook, Microsoft PowerPoint, Word, Excel) + MicroDicom X-ray viewer and 361 MRI DICOM image library |
| AI Training, AI Inferencing | Dataset: ChestXRay14 Images: 112,120 png images Description: Chest X-ray dataset, containing over 100,000 frontal view X-ray images with 14 labeled diseases categories. |