Thursday, January 2, 2020

Dropped frames and shoddy motion detection with 14 IP cameras

Hello,

I have been tasked with deploying a security system for a small business. I have made efforts to do research and avoid bottlenecks.

Here's what I'm working with:

14 Amcrest IP2m-851E cameras (1920X1080p) @ 9FPS and VBR H.264, streaming limit set to 4,096Kbs

Netgear unmanaged 16 port gigabit network switch (GS316). This powers and handles connections for all the cameras.

Fios G1100 gigabit router

Security server/PC running an i5-8400, 16GB RAM and 11TB RAID with 2 12TB WD purple drives and Blue Iris.

My target is to have them at 9FPS with motion detection recording, in an attempt to keep recordings for as long as possible. However, after installation, I'm having issues with the camera recordings being extremely choppy, skipping entire seconds of footage, and motion detection happening only after motion has happened for quite a while, and not seconds before detection as I've set in Blue Iris. I've tried iSpy, which seems to work a bit more reliably, but with much lower framerates, higher CPU usage, and still the same issues with seconds of missing footage and dropped frames.

So I thought, perhaps the write speed of the hard drives is too low. Task manager shows quite minimal usage overall of system resources with Blue Iris. Then I thought the security PC network connection was not fast enough (but it is able to achieve 150mbs to the internet). I thought the network switch is overwhelmed, but I have a gigabit switch and router, which I thought would be overkill for the cameras. I plugged the PC into the network switch and was able to achieve 150mbs download/upload, with all 14 cameras connected and recording. So I've assumed that the switch and router are not overloaded.

At this rate, I'm not sure what's bottlenecking my setup, or if I've made a stupid mistake somewhere. Any suggestions as to what could be a potential bottleneck would be greatly appreciated as I really need this system up and running, and I can't think of what might be wrong. Thank you!



No comments:

Post a Comment