TrueNAS Scale on Arm (2024 Thread)

[geerlingguy]

It's been a few years since the last serious thread about TrueNAS on Arm passed us by. See:

• ARM port of FreeNAS planned??
• FreeNAS with Raspberry Pi. Is It Possible?
• Looking for: FreeNAS on ARM solution

Many of the old excuses for not porting to / maintaining an Arm release (arm64 only here) included:

• No Arm systems exist with server-grade hardware (ECC, fast, lots of PCIe)
• None of the Arm hobby boards have enough RAM to be a serious storage platform
• Arm CPUs that can run native Linux are slow
• Arm is not a mainstream platform target for small business or homelabbers who are the target audience for TrueNAS

I... think after a few more years here, seeing the introduction of the latest generation of RK3588-based boards on the low-end consumer side (example), and readily-available Ampere CPU + Motherboard combos available at the high-end (example), with a number of options in between, it would be good to revisit the discussion.

There is still a lack of universal support for things like UEFI, though the Ampere options have full UEFI support to the point they can run Windows on Arm without any hacking necessary. Just flash the installer to a USB drive and go. Same with Ubuntu, Fedora, Rocky Linux, etc. No special builds necessary.

I've been running multiple Arm NASes for a few months now, from a Pi 5 running 4 SATA drives over PCIe with 8GB of RAM to an Ampere system with 128 GB of ECC DDR4 RAM, U.2 NVMe + SATA storage, and 10 GbE networking (easily capable of much more). All my NASes are running ZFS with no stability issues, and they have been a joy to use.

I would prefer to use TrueNAS Scale, but have so far tested ZFS + Ansible (no fancy UI there) on one and openmediavault 7 on the other. The LTT YouTube channel just reviewed the FriendlyELEC RK3588-based board I linked earlier and mentioned in their video they also chose OMV because TrueNAS isn't available on Arm.

I even asked on Twitter last week, and @Kris Moore responded "Just hasn't been a big enough business case for us to go build and then maintain an ARM variant. As much as we'd love to do so from a pure geek standpoint :)"

I guess my question is: what else needs to happen to make a business case for Arm support? Because the hardware is here, at least judging by my experience. It's not even painful to use anymore!

 

[Ericloewe]

None of the Arm hobby boards have enough RAM to be a serious storage platform

Or connectivity. Never mind the fact that hobbyists won't pay the bills, serious enterprise systems will... And there's very little impetus to spend a lot of engineering hours on something that will bring - at best, with asterisks attached - a minor improvement to your overall product.

 

[Stux]

what else needs to happen to make a business case for Arm support?

A demand for ARM from enterprise I’m guessing. Which doesn’t really make sense, unless you can get a silly amount of lanes cheap.

 

[geerlingguy]

Or connectivity. Never mind the fact that hobbyists won't pay the bills, serious enterprise systems will... And there's very little impetus to spend a lot of engineering hours on something that will bring - at best, with asterisks attached - a minor improvement to your overall product.

I've seen some Ampere deployments running $100k+ ceph deployments. Connectivity is definitely not an issue in that space, nor is funding.
There is a distinct lack of 'middle of the road' options, like an Apple M-class chip with PCIe and IO that would rival an enthusiast PC, but I keep pestering Ampere about that.
On the other side, there are also more Qualcomm-based options coming on the scene with pretty good single-core performance, but lacking in robust IO options (even the SBC makers seem to be doing better on that front than the tiny PC industry).

 

[Arwen]

One issue that exists with ARM64, is simple, common boot. It's mentioned above, but let us be clear, each board is different and needs it's own boot code or setup. It's a solvable problem, since many ARM64 either supply a small flash drive for BIOS / UEFI. Or can chain boot from something.

Another would be feature set.
For example, would it be completely like TrueNAS SCALE, except on ARM?
If so, then Apps would have to be supported. Then more than minimal amount of memory to support those Apps. Then potentially 10Gbit/ps, (or higher), Ethernet, (even if by PCIe card), so those Apps can supply clients with data via the network.

In the case of Qualcomm, they tend not to be Open Source about drivers. Or have limited support. This can be problematic for developing all in one software like TrueNAS.

On the other hand, if iX was going after the smaller, cheaper consumer market, a TrueNAS Mini ARM could be designed and made. But, it would have a specific, limited feature set, like many other packaged, cheap NASes.

 

[Ericloewe]

I've seen some Ampere deployments running $100k+ ceph deployments.

And those are hardly hobby boards. As I said, their benefits, insofar as they may be real, have not merited serious consideration being given to working on seriously and credibly supporting them.

 

[Nicki]

I think there is an opportunity for long term success here. Even if the TrueNAS solution someone ends up buying from iX in the future isn't ARM, maybe they chose TrueNAS because of their familiarity with it? There should also be a market for cheap, energy efficient off-site backup solutions that isn't a cloud provider.