After I lost a hard drive to (possibly) a lightning storm, I re-evaluated my backup strategy. I chose which cloud backup software I'd use (winner: Duplicacy), but where should I upload my backups to? There are lots of blob storage providers.
The pricing is fairly tricky to work out. There's usually storage fees (per GB per month), sometimes 'restore fees' if it's in deep storage, and sometimes network egress fees when you want to get your data back out onto the public internet. Some providers have 'minimum retentions' where you have to pay for at least 1, 3, 6, or 12 months of storage.
What I care about: I have about 1TB to back up. I want backups to be quick, so they should probably have an Australian region. I'm happy to keep the backups for at least 90 days. But I probably want to start thinning backups out after 90 days, and delete altogether after a year. I don't really want minimum retention period of a year because then I'll feel very locked-in.
Egress fees: I don't care so much about network egress fees because I think the risk of me losing this data locally is pretty low, about <10% each year. If my local data fails, I'll probably be happy to pay a bit to restore. Or maybe I can poke around the data on a VM in their cloud, and only egress the really interesting stuff. So my expected value of egress fees is maybe one tenth what's quoted. Egress fees are generally high, so you really don't want to use a backup software that has to read back data as part of normal operation.
Your mileage may vary: This is a pretty rough, not very rigorous, somewhat "vibes-based" comparison. I didn't bother considering restore fees. You should probably read Anthony Agius' Cloud Archive Storage Comparison in addition to mine, because he goes deeper into niche players that I couldn't be bothered evaluating, and considers restore fees. Also this great comparison on the Duplicacy Forum.
I chose Google Cloud Storage Coldline. This was frankly a bit disappointing, as I was hoping to find something better than what I was already using!
Here's the spreadsheet I made to evaluate (original in Google Sheets):
Google Cloud Storage
Disclosure: I'm employed by Google, and use Google Cloud Storage at work.
I was already using Google Cloud Storage Coldline for Arq backups from my laptop. It's a good fit for backups, with decent storage pricing ($72/TB/year) and still instant retrieval when needed. There's a high-speed Sydney region, the security posture is great, I'm very familiar with the platform, the tool ecosystem is large, but the egress pricing, particularly in Australia, is extremely expensive.
Amazon Web Services S3
The market leader and inventor of this segment. Amazon has 5(!) classes of blob storage, from "Standard" to "Glacier Deep Archive".
Amazon has a Sydney region, which is probably fast to upload to. Their security posture is great. But their instant-access 90-day retention offering, "Glacier Instant Retrieval" was much more expensive ($166/TB/year storage) than Google's. And I didn't really feel like wrestling with the tradeoffs of "Glacier Flexible Retrieval" ($43/TB/year storage). Would each API request take an hour to return? What if I need to restore many files, do I need to rewrite my backup client to work in parallel? Or do I copy to other storage first? There's probably good answers to these questions but I don't have them.
AWS' egress fees are famously expensive and opaque, though it looks like it's cheaper than Google in Australia. It was pretty hard to find the egress pricing and I'm not totally sure I'm reading the right price?
Microsoft Azure Blob Storage
Azure is the #2 cloud. They have a Sydney region, and 30-day ("Cool") and 180-day retention options ("Archive"). They have lots of different pricing depending how much level of replication you want. I just looked at their cheapest option. Their 30-day retention plan "Cold" costs $132/TB/year storage. The 180-day "Archive" storage is cheaper than Google, but might take hours to restore, and I don't want to figure that out.
A lot of people on backup forums recommend Backblaze B2. The price is great: $60/TB/year storage, the egress fees are <10x cheaper than Google/Amazon/Microsoft, and it's backed by a big backup company. I thought it was promising, so I tried it.
B2 was slow. I'm not sure exactly why; maybe it's the latency from Australia to their closest datacenter (US-West)? Or maybe their computers or disks just can't sink the data fast enough. My home internet is 40Mbps up, but with Duplicacy backing up to Backblaze B2, I struggled to maintain 10Mbps, even when I cranked it up to 20 parallel upload threads.
This made the backup take around 3x as long as it should (around a week, rather than two-ish days), so I cancelled the upload midway and moved to Google Cloud Storage. It's a shame! I would have loved a cheaper option with no egress/retention fees.
Backup forum people also recommended Wasabi. Their price is the same as Google Cloud Storage Coldline ($72/TB/year), and they have no egress fees. I signed up to try them.
I like that they have a Sydney datacenter, which might make them faster. But it turned out you can only list buckets by sending requests to the US?
What? So I couldn't set up Duplicacy to use the local datacenter. This seems to have been broken since 2019, so I don't think it'll be fixed anytime soon.
I find it shady that Wasabi don't explain on their pricing page that they will charge you for minimum 90 days storage, while all other competitors I surveyed were up-front about this, Wasabi buries this information in their FAQ. I only found out about this because of people warning about it on backup forums! It would have been very annoying if I'd deleted some data and charged for 90 days of storage.
On Wasabi's pricing page, they directly compare their offering (90day min retention) against Google Cloud Storage Standard (0 day retention), when Coldline would be the comparable offering, and cost the same price.
This seems, at best, misinformed about their competitors, and dishonest at worst.
With data storage, trust is so important! You're paying for the trust that you'll be able to read your data back. Underhanded-at-best price comparisons, and hiding retention fees, hardly inspires trust. So I closed my Wasabi account and moved on.
I was backing up my Synology Network-Attached-Storage device. And Synology has an integrated cloud storage offering. They have no retention fees, but the closest datacenter they have is Taiwan, which is still a while away. And the price was pretty similar to Google Cloud Storage Coldline. In the end, I was pretty unimpressed with their software's speed, so I moved on.
Cloudflare R2 made a big splash recently with object storage with $0 egress fees from a large CDN. It's unclear whether they have storage close to Sydney: they only support the
auto zone, where they decide for you where your data lives. The pricing didn't really make sense for me though, at $180/TB/year. So I didn't look further into it.
iDrive e2 has rock-bottom pricing ($48/TB/year), no egress fees. Their storage is "S3-compatible". No Australian region however, the closest is San Jose, CA.
But the vibes seem off – their pricing lights up the "too good to be true" scam-detection part of my brain. iDrive offer 90% off the first year, which seems... steeper of a discount than a reputable company would do.
Obviously it's a customer acquisition strategy and they hope to lock you in. But I'm skeptical of storage businesses that sell unsustainable products.
There are a lot of bargain-basement storage providers that offer "lifetime deals" or "unlimited storage". These are unsustainable and smell like pyramid schemes. iDrive isn't quite doing this, but it's close. I want to pay them a sustainable price to store my data. If not, why wouldn't they cut corners, and lose my data?
Some backup forum people talked of Storj, which claims to "Store data securely on over 16k nodes worldwide instead of just a few vulnerable data centers with privacy and CDN-like performance by design". The price is really low: $48/TB/year, with very cheap egress: $7/TB/year.
Random people get paid for their disk space on Ethereum. Regular audits are carried out to ensure nodes are actually storing the data.
It's interesting. But, again! Backup storage is about trust, and I don't trust blockchain grifters with my backups.
I went with Google Cloud Storage Coldline. The price is transparent and decent, the data is stored in Sydney, the upload speed is good, and the security posture is strong. I will pay a bunch in egress fees if I have to restore, but there's low probability of my disks failing, so I'm discounting that.
Hope this helps if you're trying to set up your own backups.