Is an Object Storage solution more reliable long term than Google Drive or the likes?
Recently I got to review what's taking space on my Google Drive, as I only use it for very important documents and so on (I don't use it for media, obviously) and it's taking about 100GB of data (I know, rocky numbers), so I pay for the $3 Google One addon…
Recently I got to review what's taking space on my Google Drive, as I only use it for very important documents and so on (I don't use it for media, obviously) and it's taking about 100GB of data (I know, rocky numbers), so I pay for the $3 Google One addon for space. This data is daily sync'd with rclone to my home server as to have a local offline copy. But it got me thinking: maybe I could just zip all the data (studies, historical financials...) and upload it to a Object Storage solution like BackBlaze B2 or maybe OVH Object Storage 3AZ (Paris region, with infrequent/cold tiers). That same data would be about $1.20 in OVH AZ3 Infrequent or $0.25 in OVH Cold Archive (they seem to use 4DCs and a 8+4 erasure coding with 11 nines of durability?). Retrieval fee would cost about $1.25 worst case scenario, but it would mean I already lost the local copy I got. So.. do you think it would be a good idea? I always used Google Drive as a universal storage solution where those kind of important docs were saved, but maybe this change would be better long term as a backup solution, while being also slightly cheaper? Or do you think it's not really worth it? My idea would be having my small local server with all the data (the one that also have all my media), also write it to a BD or HDD for local "cold backup", and upload the offsite to that S3 service. So if I need to check a file, I can go and take a peak through the home server instead of needing a restore of the ZIP on S3. Therefore, filling all the 3-2-1 backup steps, but now taking the local server as the root instead of Google Drive. But again... I don't know if it's a good idea or it reduces/increases durability/reliability and so on. Thanks
Collected discussion
Expand the replies to this comment to learn how AI was used in this post/project.
Yep, that's the question. Would you trust an object storage service (BackBlaze, OVH with 3AZ...) more than Google Drive to hold data long term and as your offsite backup service?
Great. So I understand you have a similar setup to mine now, Google Drive holds the files, you sync to local, and then ship it elsewhere for backup, so 3 copies total of the data taking Google Drive as the root? I was wondering if cutting Google Drive of the equation (therefore, local data only + second local disk + S3 like OVH or Hetzner) would be better than the current flow
Yep, any sensitive data is encrypted (cryptomator, 7z encryption...) once it touches any cloud, no matter what The idea to change the flow was about long term reliability really, and to DeGoogle a bit, but if it ends up being the best approach, then I value more the easiness and protection of data than being DeGoogled
Right, object storage is infrastructure and Google Drive is a service. Google Drive is certainly using object storage underneath.
Then I suppose you would keep a cloud copy as the root, then local sync (mirror) as to keep an offline copy, and S3 as a backup? In your case, how do you do it or what cloud(s) do you trust? Thanks
I’d avoid making it about the root copy. Keep the daily-use copy wherever you actually use it, then make the backups versioned and test a restore once in a while. And I’d trust that more than any specific cloud. A sync folder can still mirror a bad delete.
I think you are looking for magnetic tape drives.
* AWS S3 's Object Storage: Yes. * Other Cloud S3 Providers: Maybe. Depends on their SLAs. * Minio /Garage/RustFS/etc on your NAS: No, but it's a good idea to store a copy there anyways The real risk with Google Drive is your account getting banned and you losing all your data. I use RClone to download my google drive, then I use Restic to ship off backups elsewhere on the internet.
the concrete edge s3 has over drive here is object lock, you can flip the bucket immutable so a compromised key or a fat fingered rclone delete cant nuke the backup. restic and borg both support append only repos for that, and b2 and wasabi do object lock even on the cheap tiers so imo thats the real reliability win over drive
They’re almost certainly the same thing, object storage is just the cost effective way of doing large scale storage. Google drive just doesn’t expose the object storage directly to you.
Look into actual backup solutions - Restic, Borg ... Those things actually have retention policies (e.g. take hourly, daily, weekly, monthly... snapshots and keep the last 3 of each). This is something neither storage providers you've mentioned offers a good option for (yes, you can set lifecycle policies on S3 buckets, but that's per object, you can't get a consistent view of a folder a week ago for example). If you add encryption at rest, the privacy aspect of whatever cloud provider you go with doesn't matter. And as long as the tool you use can talk to that storage layer (e.g. Restic can leverage anything S3 compatible), it doesn't really matter where your data is, as long as you know it's there (i.e. test your backups regularly). With something like Borgmatic, you can definite multiple targets, and all your backups are automatically pushed to all these targets - e.g. remote S3 and local NFS. In terms of reliability, short of a datacentre going up in flames (I guess it might be relevant for OVH...), it's all more or less the same.
First of all I would not recommend storing anything, much less very important documents on google as it’s one of the lowest privacy mainstream cloud solutions out there. It’s not clear what your motivation is for changing but I would at least recommend using a more privacy respecting cloud service at minimum. If you self host and do object storage, make sure you have a recovery plan that you’ve tested, otherwise you could lose the data.
Absolutely nothing is more reliable than the other simply because it could disappear from under you for various reasons because the company said so. Ideally you'd have two object storage backups + local nas backup. That could get expensive fast so I backup to object storage only vital data that's not too large. Anything else is backed up at home.
More reliable than Google Drive?
I would not make the home server the root copy unless you are also testing restores. Object storage is fine as a cold backup, but the risky part is sync behavior. If rclone mirrors a bad delete or corrupted file, the backup can faithfully preserve the mistake. I’d want versioning or object lock turned on, plus a small restore test every so often.
Imo the reliability is not in the tech, its in the plan. Google drive is "reliable" until the account gets flagged by some robot and you discover support doesnt exist. My setup is boring on purpose : local copy + a second machine + offsite, tested restores once in a while, and mail notifications on every backup job so silence means something is wrong. Any decent object storage fits in that plan, none of them replaces it.
I keep my backups with Scaleway Object Storage (S3 compatible). Multi-AZ with Object Lock enabled (means once uploaded, files can't be deleted up to n days) which is extra protection from losing data by mistake etc. 100 GB would cost a couple bucks a month.
Object storage is more durable than Drive, but durability and reliability aren't the same thing and it's worth being clear which you're buying. B2 or OVH cold with 11 nines means the bytes won't rot. It says nothing about you fat-fingering an rclone sync and propagating a delete, or losing access to the account. That's why 3-2-1 is about independent copies, not one super-durable copy.
since you daily sync things automatically, you don't have a backup. A scheduled restic backup to an s3 storage would be a good idea, and solve your problem too. It is an incremental backup tool with retention settings. You can keep monthly backups for 2 years without the cost of 24x your data. As for syncing stuff, it's pretty hard to do cheaper than big tech. It's easier to amortize the cost if you already host other stuff. The main issue is that you need a server, storage, and a client. Nextcloud to S3 with encrypted storage would give you privacy, but not reliability nor cost.
The mechanism doesn't matter really. What matters is the skill and reputation of the company providing it. I would be surprised if you lost data off either Azure Blob Store or Microsoft OneDrive.