Hacker News Reader

S3 or glacier? Glacier is cost competitive with local disk but not very practical for the sorts of things people usually need lots of local disk for (media & disk images). Interested in how you use this!

20TB which u can keep in a 2-bay cute little nas will cost you $4k USD / year on S3 infrequent access tier in APAC (where I am). So "payback time" of local hardware is just 6 months vs S3 IA. That's before you pay for any data transfers.

Did you factor in the resilience and redundancy S3 gives you and you cannot opt out from? I have my NAS, and it is cheaper than S3 if I ignore these, but having to run 2 offsite backups would make it much less compelling.

Agree, they are not the same thing. Yes, S3 provides much better durability. I just can't afford it.

For my use-case I'm OK with un-hedged risk and dollars staying in my pocket.

I backup my nas to rsync.net, it’s very cost effective using borg backup.

$0.01/GB/mo, that does not seem better than Glacier, is it?

Except it’s not glacier speeds, there are no bandwidth costs, support is on a completely different level than aws (you can actually reach an actual knowledgeable human), and you can use anything that speaks ssh. They also have an expert price at 0.008$/gb/mo here https://www.rsync.net/products/borg.html

Yes and no. I have been using NAS for a long time, and I use older drives as offline/offsite backups. So the cost is mostly amortized already. Those machines are off except once a week (local)/ once a month (offsite), to do an incremental backup. So this is a good use of some older drives.

They probably factored RAID1 into that price, which you can skip if you're setting up three copies. (At least I hope they did, their hardware prices must be dire if $2000 only gets you a tiny NAS and two 10TB drives.) If I do napkin math based on US prices, a mini PC and a 20TB external drive are a bit under $500 total, and a 2 bay NAS and a 20TB internal drive are a bit over $500 total, so that's about $1500 for the triple-NAS option and $3000/year for the S3 infrequent access option. Still extremely compelling.

> S3 or glacier

This is the same product.

> 20TB

I think we might be pushing the 1% case here.

Just because we can shove 20TB of data into a cute little nas does not mean we should.

For me, knowledge that the data will definitely be there is way more important than having "free" access to a large pool of bytes.

20TB isn't all that much anymore, especially if you do anything like filming, streaming, photography, etc. Even a handful of HQ TV shows can reach several TB rather quickly.

Additionally, 20TB is only going to run you $300-400 for a consumer drive.

Yes. 20TB isn't a NAS. It's the HDD acting as bulk storage in your desktop.

It's wild consumer drives are so much more expensive than enterprise Exos drives with better performance, reliability and warranty

20 TB isn't that out of reach when you're running your media server and taking high resolution photos or video (modern cameras push a LOT of bits).

I'm the last person I know who buys DVDs, and they're 2/3s of the reason I need more space. The last third is photography. 45.7 megapixels x 20 FPS adds up quick.

S3's cost is extreme when you're talking in the tens of terabytes range. I don't have the upstream to seed the backup, and if I'm going outside of my internal network it's too slow to use as primary storage. Just the NAS on gigabit ethernet is barely adequate to the task.

hardly 1%, i’m sure anyone that works in the film industry or media in general has terabytes of video footage. Maybe even professional photographers who have many clients.

> knowledge that the data will definitely be there is way more important than having "free" access to a large pool of bytes

Until Amazon inexplicably deletes your AWS account because your Amazon.com account had an expired credit card and was trying and failing to renew a subscription.

Ask me how I know

>This is the same product.

Confusingly "Glacier" is both its own product, which stores data in "vaults", and a family of storage tiers on Amazon S3, which stores data in "buckets". I think Glacier the product is deprecated though, since accessing the Glacier dashboard immediately recommends using Glacier the S3 storage tiers instead.

20TB is a single drive

I currently store 10 TB on my NAS, and growing. The data is live, I access some of it every day, sometimes remotely. I have 3 rotating "independent" backups in addition to the NAS (by independent I mean they're made with rsync and don't depend on any specific NAS OS feature), stored in an old safe that would probably not be very effective against thieves but should protect the drives in case of fire.

There are no recurring costs to this setup except electricity. I don't think S3 can beat that.

> Just because we can shove 20TB of data into a cute little nas does not mean we should.

Okay, I'm curious now. When you were talking about "a bunch of local disks", what size disk did you have in mind?

Right now the best price per TB is found on disks in the 14-24TB range.

> If no one has a clue what is inside your blobs, they might as well not exist from their perspective.

This is not the perspective of actors working on longer timescale. For a number is agencies, preserving some encrypted data is beneficial, because it will be possible to recover in N years, whether any classic improvements, bugs found in key generators, or advances in quantum.

Very few people here will be that interesting, but... worth keeping in mind.

The point of encryption in this context is to defeat content fingerprinting techniques, not the focused resources of a nation state.

The only thing making S3 a no-go for me is their outbound traffic costs at $90/TB, which at 100TB restore would make out $9K just to transfer the data back once

> You'll never beat the QoS and resilience of the cloud storage product with something at home.

You can’t be serious.

How much does S3 cost these days? I've been burned by their hidden surge pricing before and hesitate to rely on them for personal storage when self hosting is fairly cheap.

Disk hosting cost isn't much

Bandwidth to get all of that back down to your system is much pricier, depending on how much you use that data.

Storage cost is nice, data transfer costs are often prohibitively high.

For large datasets, the cost of that is like rebuying your NAS every 9-12 months.

3-2-1 says you want both. Can be convenient to centralize backups on a local NAS and then publish to the cloud from there.