Is there any benefit to using HPC for scaling Reolink NVR setups?

[user_863771412115668_863771412115668]

Hi all,

I’ve been working with Reolink NVRs (mostly the 16 and 32 channel models), and I'm starting to hit performance and storage bottlenecks as the number of high-resolution IP cameras grows. I was wondering—would there be any real benefit in integrating HPC Computing resources into the setup?

Has anyone experimented with offloading video processing, motion detection, or storage management to HPC clusters or more powerful servers, instead of relying solely on the NVR units?

I'm curious if it’s overkill or if there’s a practical way to leverage HPC-like setups (e.g., GPU nodes or distributed storage) for large-scale surveillance networks.

Thanks in advance!

[joseph_1979]

@user_863771412115668_863771412115668 You mean RLN36 which has 36 channels. This can have storage up to 48TB. This caters for 15 days recording from 36 8K cameras.

You may opt to have Frigate installed on a server with higher storage capacity. There are a lot of customers who have integrated Reolink cams with Frigate. They use the AI detections implemented on Frigate. Frigate supports both license plate recognition (LPR) and face recognition natively in version 0.16 and later, allowing you to identify known faces and vehicles by assigning custom labels and triggering Home Assistant automations based on these detections

[KimchiGUN]

@user_863771412115668_863771412115668

I have 2 36 channel NVRs in 1 site... both almost full. Almost 70 cams. To offload this to a HPC cluster or GPU cluster would be too much $$ compared to 2 NVRs.