Home > Workload Solutions > Safety & Security > White Papers > Vix Vizion Imagus Video Analytics Validation with Dell EMC PowerEdge Server and NVIDIA-T4 GPU > Project summary
The project was undertaken with the goal of characterizing the performance of the Vix Vizion Imagus Video Analytics application on a Dell EMC PowerEdge server platform with hardware acceleration using the NVIDIA-T4 GPUs, NVIDIA GRID Virtual GPU software on a VMWare Virtual Machine platform.
The primary objective was to evaluate performance using a single VM with one or more GPUs and multiple VMs sharing multiple GPUs configured as vGPU using NVIDIA GRID. Another objective was to evaluate the ability to migrate a running application from one or more VMs and GPUs to another host with equivalent resources (CPU, Memory, GPUs) and test the HA failover while running GPU-based applications on vGPUs.
The System under test (SUT) was receiving multiple camera streams from a camera simulator. The application was performing facial detection and recognition using test data camera streams with multiple faces in a high-density street scene moving camera setting. The primary measurement was how many camera streams could the SUT process without significant frame loss, measured as processed frames per second. The application was using a target processed frame rate of 12 frames/sec (fps) with an input frame rate of 30 fps using 1080p high-resolution camera streams at 4 Mb/s.