Overwhelmed with options on cloud backup services for ~5 TB of data from TrueNAS

[testfire10]

As the title goes... I've been reading a lot about different cloud storage services. The purpose of the cloud storage for me is the '1' in the '3-2-1' strategy. I have TrueNAS on-site backing up to my synology NAS on-site. Currently, I've been using Backblaze personal backup, which is great, but now that I am moving my data from my workstation to a couple of NASs as I change my storage strategy and upgrade the homelab, I can't use the Backblaze personal backup option anymore (they don't backup from a NAS). I have to switch to B2 in order to do that.

I like BB, and have had a positive experience with them for a few years now, but I don't know about the idea of 'pay to play' the way B2 is structured. To store the ~2TB of data I have now with B2, it would be around $130 a year or so, which isn't that bad, but I'd always be feeling like I'd need to watch my uploads so I didn't really get out of control on the cost.

I don't want to use Amazon or Google because I don't trust them (well, I don't trust anyone with my data, but I really don't want them having it). I have looked through some of the other options like idrive, dropbox, etc. as well.

What are all of you guys using for cloud storage? Do you have any recommendations for something that is secure (very important for me), and somewhat cost effective? I don't need (hopefully) to access the data, only adding to it, until that fateful day that I may one day need to restore everything, if that matters.

Am I asking for too much?

 

[Mlovelace]

As the title goes... I've been reading a lot about different cloud storage services. The purpose of the cloud storage for me is the '1' in the '3-2-1' strategy. I have TrueNAS on-site backing up to my synology NAS on-site. Currently, I've been using Backblaze personal backup, which is great, but now that I am moving my data from my workstation to a couple of NASs as I change my storage strategy and upgrade the homelab, I can't use the Backblaze personal backup option anymore (they don't backup from a NAS). I have to switch to B2 in order to do that.

I like BB, and have had a positive experience with them for a few years now, but I don't know about the idea of 'pay to play' the way B2 is structured. To store the ~2TB of data I have now with B2, it would be around $130 a year or so, which isn't that bad, but I'd always be feeling like I'd need to watch my uploads so I didn't really get out of control on the cost.

I don't want to use Amazon or Google because I don't trust them (well, I don't trust anyone with my data, but I really don't want them having it). I have looked through some of the other options like idrive, dropbox, etc. as well.

What are all of you guys using for cloud storage? Do you have any recommendations for something that is secure (very important for me), and somewhat cost effective? I don't need (hopefully) to access the data, only adding to it, until that fateful day that I may one day need to restore everything, if that matters.

Am I asking for too much?

Tarsnap is a solid choice.

 

[jgreco]

The main problem with tarsnap would be that it is AWS-hosted, which the OP didn't want, though since only users have the keys, it isn't clear that this is a realistic issue to have a crisis over.

 

[Mlovelace]

To be fair if you want bulk cloud storage that doesn't cost you more than it would to just build and colo another server, it is very difficult to get away from the big three. My thought on it being, if the data is encrypted then sent, who cares if it sits in one of "their" datacenters, but to each their own.

 

[jgreco]

To be fair if you want bulk cloud storage that doesn't cost you more than it would to just build and colo another server, it is very difficult to get away from the big three. My thought on it being, if the data is encrypted then sent, who cares if it sits in one of "their" datacenters, but to each their own.

Well, looking at the pricing for tarsnap, $0.25/GB/month works out to $250/TB/month, and it doesn't take many of those to outrun the capex/opex cost of a cheap colo. Possibly as much as $1250/5TB/month for the OP's 5TB? They do some sort of nebulous deduplication that may lower that cost, and apparently they store it in multiple S3 areas.

Straight Amazon S3 is a bit more palatable at $118/5TB/month (US East per AWS calculator), but that's getting to the point where it could be tempting to find a 1U colo hosting provider, typically $30-$100/month, and put a 1U FreeNAS host in there with a pfSense gateway VM.

Basically, once you get into the terabyte range, it seems like cloud primarily wins on the convenience side of things -- you don't have to figure out any of the messy bits yourself, but you're paying for the convenience.

 

[testfire10]

@jgreco, I think you're missing a decimal place somewhere. Tarsnap is priced $10e-12 per byte/month (the 'picodollar', nerds), TB is 1e12 bytes, so I'm coming up with $10/TB/month, which is sort of on the order of Backblaze B2.

 

[jgreco]

@jgreco, I think you're missing a decimal place somewhere. Tarsnap is priced $10e-12 per byte/month (the 'picodollar', nerds), TB is 1e12 bytes, so I'm coming up with $10/TB/month, which is sort of on the order of Backblaze B2.

I don't care for cutesie-ism's or being made to do nonstandard math, so I take

"Tarsnap uses a prepaid model based on actual usage: 250 picodollars / byte-month of encoded data ($0.25 / GB-month)"

to mean 25 cents per GB per month, the bit (haha) they themselves specified in the parenthesis. That's off their main web page. It also seems to translate correctly to 250 picodollars / byte-month.

I don't know where you get $10e-12 per byte/month from.

So if we then assume that the number that they specify, and the number that I calculate, is correct, and you are mistaken, then a terabyte is one thousand gigabytes, so you multiply $0.25 by 1000 = $250 for cost per terabyte-month, and then times 5 for 5TB, and I come back up with my same $1250/month.

 

[testfire10]

You're right @jgreco, I misread their 'picodollars' page, mis-interpreting it to mean the pricing was 10e-12 picodollars/byte-month. Thanks for clarifying. This is much more expensive that I'd want to spend.

 

[Heracles]

Here, I did looked at these clouds at first but quickly discarded them. I considered that it would be cheaper and safer to run everything myself.

My first server grade box was the T-110 that now runs Thanatos. I had that for a long time. I upgraded it with the T-130 that became Hades few years after. Each one was in the 1,000$ range and before I built my private cloud.

The T-130 was not that powerful as an ESXi server, so I got myself the T-330. That one was my first significant ESXi server. It was back then that I built my private cloud, having only a 2 copies backup strategy instead of 3.

It was short after that that I had the opportunity to buy a used monster (32 cores ; 256G RAM for 2,500$). That one became my ESXi, rotating my servers down 1 step and letting me put the 3 copies rule in place.

Over all, considering the cost of the smaller servers and that I intend to have these for years, I am better by myself than with commercial clouds.

 

[testfire10]

My first server grade box was the T-110 that now runs Thanatos. I had that for a long time. I upgraded it with the T-130 that became Hades few years after. Each one was in the 1,000$ range and before I built my private cloud.

The T-130 was not that powerful as an ESXi server, so I got myself the T-330. That one was my first significant ESXi server. It was back then that I built my private cloud, having only a 2 copies backup strategy instead of 3.

Thank you @Heracles - so when you say private cloud, what does that mean? Like, where is Hades located? Not geographically I mean (I see you list it as 400km away), but what is the infrastructure you used to do that? You physically carried your server to a friends house or something? Then use the site-to-site VPN to replicate?

 

[Heracles]

Hi @testfire10 ,

My private cloud is based on Nextcloud. I run it from a Docker host in my ESXi server. It offers all the services a public cloud typically offer : file sharing, photo upload, contacts, calendars, bookmarks, Office suite, mail, ...

Storage is provided by TrueNAS. The Docker host mounts the NFS share and Docker re-maps it inside the container. Thanks to that, the container is not even aware that data are not local. I created a primary dataset that I called HAStore. It is in that one that I put everything I want replicated to the 2 other servers. The space for it is reserved on each of the 3 servers, so I am sure that if it fits in Atlas, it will get replicated to the 2 others. For other stuff like the recording from ZoneMinder, I put that in another dataset (LocalStore) that is not replicated.

Hades is at my father's place. For its first sync, I had it at my place. I did the sync over LAN. I then unplugged it and took it down to my father's place where I plugged it back. I updated the local DNS name to point to its new address in my father's network and voilà : the sync process sent the few missed snapshots and keeps it up-to-date every 15 minutes now. Both my father and I are running pfSense firewalls (a VM for me, an SG-1100 for him) and have IPSec configured between the two. Also, both of our networks are segregated and servers are running in an isolated segment that users can not reach. We both have unlimited Internet access plan.

Have fun designing your own solution,

 

[jgreco]

Cloud was originally a term that meant a multi tenant computing environment hosted on someone else's data center assets, but dilution of the term and frustration with idiots has led to other uses.

I've fought a few "why not do everything on cloud" battles over the years, and the trick is, while cloud is suuuuuper-awesome at being able to spin up a workload and do some work and then vanish back into the ether, with a "modest" cost, it is really horrible at maintaining baseline workloads and storage and stuff like that, because the cloud costs seem to work out to like 5x-20x other reasonable solutions such as running it on-prem or in a colocation facility.

So the way that this works out from a practical point of view is that you can pick up a full rack at Hurricane Electric, with power and a gigabit ethernet, for $400/month, and then load three of @Heracles $2500 ESXi servers in there ($7500 capex) plus three FreeNAS hosts based on some cheap Dell R510's and 14TB HDD's ($16000 capex) plus random other parts for a total capex of maybe $25K. Over a five year period, that works out to $416/month amortized capex plus $400/month opex, or $816/month. That would give you 300TB of RAIDZ3 storage plus 96 cores and 768GB of RAM.

Of course you could go much lighter on the hardware and split costs with a buddy as well.

Now the thing is, you add up the costs to do 300TB of storage with no compute and no bandwidth over on AWS and ... what's that number? $6,800/month?

So the thing is, those of us who actually maintain nontrivial baseline resources on the Internet find the prospect of public cloud to be relatively horrifying. I can actually more easily afford to throw a bunch of resources out in a colo and then spin them up as demand requires. I've got nodes at the data center that have never been powered on for use, in fact more nodes never used than the number actually carrying our workload. And it's still cheaper, by far, than some janky-ass public cloud.

We end up referring to this as "private cloud" because we do actually sell to clients, so it is technically multi-tenant, and it's private, since you cannot just go to some rando web page and sign up for it.

Other people are doing other meaningful things with their own resources. As the speeds of end-user Internet connections get better, you also get a lot of new options such as hosting gear with friends or family, so it stops mattering so much.

 

[adrianwi]

I'm still using CrashPlan running from a Linux VM, which has been remarkably stable for the last few years. The regular CrashPlan updates certainly cause less damage than trying to run the headless version in a Jail. They don't have the cheap $60/year with unlimited storage as an option anymore, but the small business subscription is only $120/year/device ($10/month) and I think is capped around 10TB? My VM mounts all of the TrueNAS datasets I want to backup, so I'm only using one 'device' and currently around ~3TB.

 

[testfire10]

Thanks for all of the very helpful replies, folks. I think I'm going to re-consider Backblaze B2 more carefully. From what I hear here, and my own research, it seems to be one of the more cost-effective solutions. I'm not yet technically capable of implementing some of the 'personal cloud' type solutions, nor do I think the effort/cost is necessary for what I need.

 

[john60]

... I'm going to re-consider Backblaze B2 more carefully. ...

Prior to Truenas, I was backing up my data to dropbox and google drive. The main issue I found was that the data rates were very so slow, my <2TB data would take about a week to transfer.

How is Backblaze working out and how is the transfer rates to these guys?

 

[testfire10]

Prior to Truenas, I was backing up my data to dropbox and google drive. The main issue I found was that the data rates were very so slow, my <2TB data would take about a week to transfer.

How is Backblaze working out and how is the transfer rates to these guys?

Hey friend. I've been using B2 for about 3 months now. I'm afraid I won't have good information on you re upload speeds. My upload speed from my ISP is quite slow - only 35 Mbps, but the upload for my ~1.4 TB of data ran at that speed pretty much continuously for about 5 days or so, so I think I at least did not saturate their limit at that rate. Hope that helps!

 

[ChrisRJ]

I am using OneDrive from Office 365 Family with encryption. Having purchased a voucher from Amazon on Black Friday, this gives me 6 * 1 TB for about 60 Euros per year.