-
- The test consisted of starting the video analytics process with a known number of video streams and logging the performance of the application (processed fps) using the Grafana based data logging capabilities of Imagus and the Graphics and host processors (received fps, processed fps, Frame Processing Time) and GPU and host memory utilization using the GPU profiling application during the different test scenario. Each scenario was run for a minimum of 30 minutes.
- While the test is running, log the performance values of the GPUs using the GPU Profiler application logging capability.
- Validate the results to show that the system was able to perform the video analytics face recognition task at the 12 fps frame rate for all camera streams.
- Validate the NVIDIA GPUs do not exceed the thermal cut-off temperature (70° C).
- Validate that the NVIDIA GPU does not exceed 90% processor or memory utilization.
- Validate that the system board CPU and memory utilization remain in an acceptable range (below 80%)
- Validate that a running system using GPU resources can be migrated while running, and a running application using vGPUs can be failed over to a partner system and measure the time from power fail to running system.
- Validate that a system can utilize multiple VMs to share GPU resources using vGPU technology provided by NVIDIA GRID.
- By observing resource utilization at the various camera stream numbers, it should be possible to use this information to determine the resources required for a system size based on similar use case requirements.