Hacker News new | ask | show | jobs
by j45 901 days ago
I would never run a self-hosted nas when a synology/qnap are available as a dedicated appliance for around the same price.

The hardware is much more purpose equipped to store files long term and not the 2-3 years between consumer SSDs'

It's not to say self-hosting storage can't or shouldn't be done, its just about how many recoveries and transitions have you been through, because it's not an if, but a when.

3 comments

> The hardware is much more purpose equipped to store files long term

What hardware would that be, specifically? The low end embedded platforms that don't even support ECC?

> how many recoveries and transitions have you been through

3 or 4, at this point, using the same 2 disk zfs mirror upgraded from 1TB to 3TB to 10TB.

Qnap and synology among other have plenty of reasonably priced ecc enabled equipment.

I’m not sure where low end comes from, most people start with a Pi, old computer or something, and grow from there.

Storage like a home appliance is a good thing.

The hardware is basically the same as self-hosted NAS, the motherboard could even be of a lower quality. The software though is closed source and most consumer NAS only get support for 4-5 years which is outrageous.
You're not buying from the right brand.

Synology supports their hardware for about 10 years since release. They are the "Apple"-like of NAS.

I think the average # of years between buying a NAS could easily be 5-10 years.
I bought a QNAP about a decade ago under the same assumption, but my experiences [0] there means I'm unlikely to buy a SOHO-level storage appliance ever again.

The tl;dr of my rant was around shortcomings in NFS permission configuration, and a failure of the iSCSI feature (the appliance crashed when you sent it data).

Further, these appliances invariably use vanilla RAM sticks, so you're exposed to gentle memory-based file corruption you probably won't notice for years.

So I'd argue the hardware is 'better equipped', and I'd also argue the software as shipped matches the marketing promises accompanying same.

Things have doubtless changed - I'm sure those bugs are long gone now - but unless you're looking at an ECC appliance, I'd say you're better off building your own white box.

[0] https://jeddi.org/b/brief-rant-on-trying-to-use-iscsi-on-a-q...

> but unless you're looking at an ECC appliance, I'd say you're better off building your own white box.

Synology actually allows ECC DRAM and even sells it and list which models would accept them.

But yeah, at the price of a full featured model with an x86 CPU and SO/DIMM RAM and 4+ drives you are in the territory of building your own, with a lot more of control and without DSM (in Synology case) shenanigans.

EDIT: actually the biggest problem here is actually finding a good case, because even ATX cases now usually don't have more than 2-3 3.5" bays by default and often don't have 5.25" at all.

https://www.synology.com/en-us/products/DDR4

A decade ago might as well have been 40 years ago.

Both Qnap and synology wt the least offer ECC capable nas. So much so they are getting better established in the smb and enterprise space. It’s a good thing you’ve pointed out to include ECC in your appliance search.

My preference and statement for storage as appliance includes experience of building my own boxes and working with servers in my own rack in a data centre for over a decade, out of my pocket.

The cloud is someone else’s computer, but I want my own storage and cloud and something easily recommendable to others when it comes up, even when they aren’t techies.

Besides, if we’re going that back in time we could just run a scsi raid array with a fibre channel connected to a power hungry server :)

I get your point - though neither company was young a decade ago (QNAP founded 2004, Synology in 2000) - and my problems were not related to 'what technology was like in the old days', rather the device was advertised with a feature that did not work (and indeed it did not work in a fashion that reliably crashed the machine).

The other feature - NFS - I'd say is one of two core features for a NAS (the other being SMB/CIFS) and it was lacking basic granularity in its ACL.

I'll note that at the time the 'app store' associated with that product has a hundred or more rinky-dinky add-ons & plugins - php admin, helpdesk product, that kind of thing. I suppose some customers found those useful.

But I'd argue those were less useful features for a NAS than being able to reliably receive & send files, or conveniently secure access to those files.

So my concern is not around ephemeral tech / feature failures, but the marketing-driven design, complexity of troubleshooting where it met the opacity of the software, and the barely-there value-add / extra-cost of a stack like that (especially for people handy with screwdrivers and a CLI).

Technology not having a feature at all, or majorly in the way advertised is a massive pain. The WD hard drive fiasco was bad too.

Looking at your list, I ended up going with QNAP to be safe .. for flexibility to get anything I needed the way I wanted to. The choices at the time were a ARM based CPU or intel celeron with very low power. I went with the latter, to realize I probably would have been good as well with the Synology I was eyeing.

The bliss of set and forget, and multiple types and locations of backups happening with good reliability is thanks to one change in the NAS space.

Existing players were expensive, and to the degree that there is profit in large enterprise storage customers wanting a backup locally in each field office for example, the NAS industry has been able to provide enterprise class features down market relatively quick and affordably.

Your posts made me recall how much I had to research the enterprise features to try and find what I wanted, equivalently in a NAS, and it ended up definitely on the Prosumer/SMB models, but ultimately in the high throughput NAS' that focused on video.

I somehow ended up with 2.5 GbE so long ago. In my case I could have just bought used fibre channel scsi enclosures, but I wanted to be mindful of power needs, and in turn backup power. It wasn't an ideal time to buy then, but is much better now. In fact, I can probably buy a second or third used version on what I have and it will work together pretty OK, including ECC if I want.

I try and remember to follow a rule to spend and buy 10-15% more capacity than I need when buying, if not one level up to get true long term utility value. For example, always at least 2 more empty drive bays than you need. If you're buying a car, consider a hatchback. If your'e buying a hatchback, consider a small SUV, etc. Buy a little bigger and nicer to let your life grow into it, or buy thrice.