Crazy idea of the day: scrap-tastic flash raid

So I have an asus flashstor NAS which holds all my files.
Unfortunately, it is full.

So, now I have to decide if I:

  • buy another one
  • upgrade to 8tb drives
  • go back to using spinning drives
  • something else

The biggest problem with the asus flashstore is that it’s difficult to expand. It has room for 12 ssds, and that is it. It got me thinking though. Speed isn’t really that much of an issue when you have lots of drives, so maybe I can build a raid using USB hubs?

USB can go up to 120Gbit nowadays, but it’s expensive and not really required for a NAS.

A nas is generally limited to 10Gbit, unless you put some extra fancy networking hardware in there. Filesystems and stuff have a fair bit of overhead, so let’s say that we need to have 20Gbit/s of bandwidth to the filesystem in order to more or less fill the 10Gbit/s network connection. That’s still only two USB 3.2 connections.

So maybe, all I need is a computer with a 10Gbe ethernet connection, some usb hubs, and some of these:

Put it all in a box and put a fan in it to get the hot air out, and then do some software raid stuff to make it all run. More and larger hubs means it’s easy to add more drives.

A computer with four USB ports, connected to four 7-port hubs would allow for 28 drives and 40Gbit/s speed.

The speed for each drive becomes very limited this way, but the overall speed for the array is still very respectable, at least on paper.

Replacing the hubs with 10-port hubs allows for more drives, but does not speed anything up.
On the other hand, adding more USB ports on the host computer does increase the speed, and can be done at any time by buying PCIe expansion cards.

The whole point of doing it this way is that smaller SSDs (~2Tb) are cheaper per Tb than larger one, but computers with large numbers of NVMe ports are fairly expensive. Large number of usb ports however, is fairly easy to come by…

If it’s something new to try, the RAID concept is pretty fascinating. Which would you use? Are you going for speed or redundancy for safety?

Good rabbit hole.

Also Lego Star Wars is on in the background and the Landolorian wins for best mashup character.

My existing NAS uses software raid6 with a btrfs file system on top for snapshots.
A USB-based raid might need a little extra redundancy, which might mean multiple raid6 setups, or hot-standby drives.

You can’t use any external HDs?

They go up to at least 5tbs I know of (I got one) & connect through USB 3.0

Spinning drives are too slow.

2 Likes

…and too easily damaged.

Loving the topic even though I’m on a Mac since I gotta reevaluate where I’m gonna take the next round of backups.

It all depends on how much data you need to backup.

If your needs are relatively low, say ~2Tb or less, then online backups probably makes the most sense. These cost money per month, but for small sizes they are relatively good and will keep your data safe even if your home were to catch on fire or flood or something.

If you have more data, hard drives is probably best for backup. They last a long time, can be written multiple times, and doesn’t require special hardware to read. (Like tapes do.) The biggest problem is how to keep them safe. You can’t keep them connected to your computer all the time, because then a virus or malware could wipe or encrypt the data. Basically, you need at least two backup drives, and one should always be kept in a safe place. I keep mine in a fireproof safe in my house, but taking them to work or something might be better.

In the past, I used to build a mini-server, then I would ask a friend to hook it up to a network in their house, and it would do automatic backups every day. Eventually I gave up on it, because the data is too much to fit on a single drive, and the mini server would very rarely be stable enough that I could trust it without having to go poke it at my friends house every now and then.

My current NAS is a 12x4TB Asustor FlashStor. This has worked quite well, and I’m pretty happy with it. Before that I had 14x4Tb spinning drive setup in a standard desktop case. It also worked reasonably well, but it was really really really slow until I added a 2Tb flash drive as a cache. The cache was complicated and somewhat brittle though. It would often need repairs after a reboot.

For backups, I currently use a 5-bay USB drive case. Each backup is a set of 2 or 3 hard drives, which I mount as one large btrfs volume, and then I do the backup there. I have three sets of backups, which I cycle into a fireproof safe once a week or so. Not perfect, but it works.

Now, if you have a lot of data to backup (like hundreds of Tb) the only solution that makes any kind of sense is tapes. Tape is king when it comes to huge data backups. The hardware is expensive, but it’s cheaper per TB than all other solutions. As I’m writing this, tape prices can go as low as $4.50 per Tb, while hard drives (which is the second cheapest) is ~$11/Tb.

I’d like to give an honorable mention for M-DISK in the backup game though. M-DISK only store 100Gb per disk, so they aren’t practical for large-scale backups. However, they are guaranteed to be readable for ~100 years or so, which is pretty impressive. (Assuming of course that blue-ray drives still exists in ~100 years.)

I have multiple 2+tb SSD’s with the data split across them to spread the storage out that actually are in a fire-safe and another 4tb one stored elsewhere, all sync’d every two to three months. Honestly I keep and prefer to have stuff stored on my own because I’ve seen what can happen with “The Cloud”. I’ll keep tabs on this topic, thanks for the feedback.

Funny you mention the blu-ray bit because I just bought a cd reader/writer as well as a blu-ray specific one, just to back up aging DVD’s, CDs, and Blu-Rays in our collections.