QNAP TS-433 Making For A Nice Open-Source & Mainline Linux NAS Experience

[phoronix]

Phoronix: QNAP TS-433 Making For A Nice Open-Source & Mainline Linux NAS Experience

While there is no shortage of consumer network attached storage (NAS) devices these days, those able to run a mainline Linux kernel, open bootloader, and other open/mainline software components is a bit more challenging. Thanks to the work of open-source developer Heiko Stuebner, the QNAP TS-433 is looking to be an interesting candidate for those wanting a nice 4-bay NAS while being able to load it with a mainline Linux kernel build and other upstream open-source software...

https://www.phoronix.com/news/QNAP-TS-433-Mainline-Linux

[kylew77]

Very cool news, but I remember writing in my storage class that a do it yourself NAS is almost always better than a store bought one. The do it yourself you get more powerful CPUs, RAM Options like ECC, more IO ports, etc. Since mass produced motherboards and CPUs exist for rolling one's own NAS, it is comparatively no harder than building a gaming rig, just with different parts. That is the great thing about x86 and x86_64, the interchangeable parts and how one can cobble together a system.

[flower]

tbh while it is technically true that it is linux i wouldn't call it that. they are just as closed as android and chrome os.
imho homemade is the way to go. i prefer nixos for that but it doesnt really matter

[Chugworth]

It's unfortunate that there are very few ZFS-based NAS devices since ZFS is really the ideal filesystem for a NAS. Consumer NAS devices often take Btrfs and set it on top of md, which is the last thing I want. It's time for md to retire. At this point there should be no new implementations based on it. If you just want a simple drive mirror then either Btrfs or ZFS are fine for that. And if you want a RAID-5 or 6 array then nothing beats ZFS.

[Lexx]

It's unfortunate that there are very few ZFS-based NAS devices since ZFS is really the ideal filesystem for a NAS. Consumer NAS devices often take Btrfs and set it on top of md, which is the last thing I want. It's time for md to retire. At this point there should be no new implementations based on it. If you just want a simple drive mirror then either Btrfs or ZFS are fine for that. And if you want a RAID-5 or 6 array then nothing beats ZFS.

But it works fine when implemented correctly (Synology was the original, netgear readynas, asustor and teramaster have btrfs on top of md raid 1/5/10/6 with self healing supporting working perfectly fine, the btrfs can talk to the MD layer so it can get the correct data block)

I wouldn't say zfs is ideal use case for a consumer nas (definitely should stick with raid6/z2 mode when using zfs)

qnap butchered it enough to get it working so it's consumer friendly

and we will soon have hexos that hopefully it has better general use usability or "it just works" as we expect it to work (most of this is around Samba) and not need to drop into console to change basic options and fix security permissions

[anarki2]

Way too expensive.

[pong]

Way too expensive.

Yep, for what it is.

And when a midrange non-flagship smart phone has more RAM than this NAS, and comparable CPU power, I have to think that a modern NAS very likely dealing with N*20TB range drives really deserves at least 0.1% of that much RAM and a solid 8 core CPU if one is going to be at all reasonably capable of doing any sane amount of caching, indexing, databasing, etc. even dealing with JUST the metadata will take that much resources.

And to have 4 big SLOW HDDs with NO 2-4x NVME mirror for cache / journal / log / index / metadata database / system logs etc. etc. is borderline insane.

How long is it going to take to just "find" a given file or look at what the inventory is and the ages / versions / etc. of all the stored stuff.

And then you get to the crypto, checksumming / hashing, compression, deduplication, better have a pretty decent CPU for all that.

Then running a few containers with a backup client, file server protocols, intranet web file / storage server, etc. needs more than these cores / RAM to work well.

No 10-40Gb/s NIC capability (the motherboard / CPU couldn't even deal with it if there was one) that's going to be slow.


Someone ought to just make a nice chassis / auxiliary hardware with like 6-8 nice HDD hot swap bays, 4x M.2 NVME slots somehow facilitated, spot to drop in any user provided ATX motherboard, spaces to put in a couple ATX power supplies and optional LiFePO4 battery attachment with a redundancy / UPS setup, spots for 4x240mm or so good fans and one would have a very nice basis for
a household / SMB storage system that would actually be flexible, reliable, perform well, and be usefully expandable.

[billyswong]

Someone ought to just make a nice chassis / auxiliary hardware with like 6-8 nice HDD hot swap bays, 4x M.2 NVME slots somehow facilitated, spot to drop in any user provided ATX motherboard, spaces to put in a couple ATX power supplies and optional LiFePO4 battery attachment with a redundancy / UPS setup, spots for 4x240mm or so good fans and one would have a very nice basis for
a household / SMB storage system that would actually be flexible, reliable, perform well, and be usefully expandable.

Because any companies that manufacture/assemble something close to that will add the CPU/motherboard/RAM etc by themselves and sell the product as a premium server?