Is JetPack meant for production use?

Hi all,

I’m currently doing some discovery for a potential new product and there was a discussion about using JetPack as the production OS instead of a custom OS using yocto. There were a few limitations to TX2 NX where the eMMC storage is only 16GB and Jetpack by default is 14GB.

I’m not sure what the industry standard are in terms of embedded OS, is it acceptable to use JetPack for production, or is custom OS with a custom updater (something like Mender or RAUC) the best way forward? I know JetPack provides an easier upgrade path to different version of L4T using apt, but is apt designed for embedded update?

Would love to connect with senior embedded engineer to learn and discuss a thing or 2 around this as well.

Regards,
Nicholas.

Hi @nicholas.leong, do you need all of the JetPack components? You could choose not to install some of those if you don’t actually need them.

If you just flash L4T and then install only the JetPack packages you want/need, that can significantly reduce the size. For example, instead of installing the entire CUDA Toolkit with all the samples/ect into your production image, just install the CUDA runtime libraries.

Hey Dusty,

That makes sense, will give it a try but what about OTA updates for embedded devices? I know Nvidia has an upgrade path for the OS over apt, would that be a good solution for IOT device OTA? Is it meant to be reliable to power fail and corruption with option to rollback to known good version?

I would check this section of the docs for more info about A/B redundancy:

https://docs.nvidia.com/jetson/archives/r34.1/DeveloperGuide/text/SD/Bootloader/UpdateAndRedundancy.html?highlight=redundant

I also think that using containers to manage your application updates makes a lot of sense for distributed edge/IoT devices.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.