Skip to main content

Performance testing on Raspberry Pi 4

29.12.2020.

raspberry-pi-4-label

Not long ago, Yucca gained support for the ARM and ARM64 architectures, so we ran a test – how many cameras can be added on the popular Raspberry Pi single-board computer.

Testing

We took the entry-level Raspberry Pi 4 board with 2 gigabytes of RAM and a Geekworm x832 board for connecting a 3.5" HDD. A video review is available on our Youtube channel.

Installation on a Raspberry Pi and on a home PC running Linux doesn't differ; you just need to choose the correct architecture in the documentation section. For the Raspberry Pi 3 and 4, the ARM architecture is suitable, and for the Raspberry Pi 4 with 8 gigabytes of RAM, choose ARM64.

After building and installing Yucca, we added 2 cameras with archive recording to the HDD. Over 24 hours we got the following resource utilization figures: 01

Hard drive (HDD) metrics: 02

In terms of performance, running on a Raspberry Pi doesn't differ much from running on an AMD64 server in a data center:

But who's interested in watching performance with just 2 cameras? Let's try to squeeze the maximum out of the board – let's add 10 more streams. Overall figures: 03

Hard drive (HDD) metrics: 04

The LA grew significantly, but that's normal since 2 processes are started per stream – many cameras, many processes, many interrupts. Once LA goes past 1, a queue builds up and this affects responsiveness, but it keeps working. In total, I managed to add 21 cameras: 05

The archive keeps working, but watching LIVE was no longer possible. So what do the charts show? Things aren't good there – the processor hit its limit, LA is above 5. Overall figures: 06

Hard drive (HDD) metrics: 07

To sum up: We don't recommend adding more than 10-12 cameras to Yucca on a Raspberry Pi 4, since this will affect the software's performance and responsiveness, but even that result is a great showing for a $35 single-board computer. We're already working on an improved architecture and plan to further reduce the workload going forward. This will save resources and, as a result, allow adding even more cameras.

As a reminder, there is a paid version of Yucca, which may be of interest to small internet providers, video surveillance integrators, medium and large businesses, residential complexes, developers, and many others. Read about additional features and terms on the website. You can ask a question through the contact form, by writing to info@yucca.app, or in the Telegram chat.