Open CESARDELATORRE opened 5 months ago
Interestingly, if the camera provisioned is a real camera (i.e. 5G camera or Wi-Fi camera), it doesn't impact in performance like it does right after provisioning the RTSP-container..., why?
This is with the new version (v2) but once the cameras were un-provisioned and stream is being processed for the last 5 minutes pending:
But as soon as I provision a new camera, it starts converting format from RTSP to the other format for HTML and timing goes up (The time of the camera I don't think it matters, though, below is just coincidence that the last stream was coming from the 5G camera):
Now, only RTSP container simulatort stream, but BEFORE and AFTER un-provisioning the camera, so after un-provisioning, the video stream converter is not working, only the AI model and the alerts, for 5 minutes:
In any case, please compare the deployment of v1.8 vs. v2.0 in the same environment (i.e. an UBUNTU VM) and let's see timing/perf comparison.
There's is a big difference in performance 400 mlSecs vs. 2 secs (ALERT's total time).
I think this is due to the process/compute needed by the stream converter from RTSP to the format needed by the HTML app.
As soon as I remove the container-simulator, the time needed per Alert drops significantly. Note that stream from the RTSP continues to come for 5 minutes after provisioning because AKRI has that "time-out", so the real RTSP container continues to work for 5 minutes but the RTSP format converter is removed since you don't see it anymore in the camera's dashboard HTML video app, but you only get it as ALERTS.
Please, identify why this significant difference in time.