67 comments

[ 4.7 ms ] story [ 124 ms ] thread
this is much cheaper than S3 glacier. Any explanation on how to keep data safe?
Any explanation on how to keep data safe?

Start by not using an unknown provider with questionable reliability or stability where the only contact is an email address.

I think you're forgetting the cost parameter—obviously no provider can guarantee either reliability or stability
They can and they do - and if they fail, they cover damages.
Is there a provider that would cover the customer's damages like interruption of business? There are some insurance companies at least that do similar things.
A SLA would. I would just use multiple backups. One self-hosted on-premise and two off-site. Then the two off-site can have egress costs. Wouldn't matter. The self-hosted one would be the primary one to backup from. And it is going to be solid, usually. For example, a Synology NAS would suffice.
I think nearly all SLAs just give you discounts on the service's bill.

Eg if your service vendor costs you 1000/month and they have a 10-day outage and you lose 10000000, they will credit you 1000 best-case, or maybe 1/3 of that (since it was up most of the month).

> if they fail, they cover damages

Umm, most providers' standard T&Cs exclude liability for indirect or consequential loss following unavailability of their service.

I'm sorry, S3 insures your data in case of loss? How come they don't ask the value of what you're storing then?
I'd rather pay a little more money to a reputable vendor, honestly. If I depend on the backup, I want a reasonable guarantee they will deliver on their promise.
We keep 3 copies of your data - two on LTO drives (across two different locations) and one on enterprise grade hard drives.
In different locations?
Different data centers
Where are the data centers located?
Well now I know it's fake. LTO drives might be a few bucks per TB but enterprise grade hard drives sure aren't.
How is this much cheaper? S3 glacier deep archive costs 0,00099 € per GB and month.
I have a fairly optimized backup to glacier strategy involving a couple of days of soak time in S3 before moving to glacier with lifecycle rules to avoid the 180 day min storage for transient files. Net I end up paying about $1.20 to $1.50 per TB per month. Restoration of my data set would cost around $100/TB though. So a straight $1/TB would be a nice savings on a good month, and a great savings if my house ever burns down and I need to restore.
Backblaze is 6$/TB/month with no egress fee, and are a well established company with a good reputation:

https://www.backblaze.com/cloud-storage/pricing

I'm surprised the don't come up more in a discussion about backup storage.

It doesn’t? From my experience it comes up every time S3 is mentioned, except maybe recently Cloudflare took the spotlight.
That has retrieval fees as well as minimum storage times.
No egress fees make a huge cost difference if you ever need your archived data back.
Did it ever get revealed how glacier works? I had heard they were using DVD-Rs back in the day, but that was a decade ago and a big price drop.

https://www.backblaze.com/blog/hard-drive-cost-per-gigabyte/

I know the link is LTO based for some of it. Is LTO really cheaper than hard drives? Or is it just that LTO has automation equipment behind it?

It's interesting to me that hard drives have dropped 56% in cost since 2017, and yet I don't think storage prices have dropped anything like that. Per the article, it's because demand allows them to keep it high.

I welcome a price war! And thanks for the honesty Backblaze.

I dont believe aws have ever said.

My long held suspicion is they designed the product slos/apis to allow for the underlying storage technology to change without customers being aware, hence its clearly inspired by tape access patterns.

Possibly some of the very largest aws regions use tape libraries.

But I suspect aws co-mingles the data with S3 and just wears the cost. They've been known to price below cost (eg lambda implementation before firecracker) to corner the market.

Submit button on the learn more page doesn't work for me.

> Failed to load resource: the server responded with a status of 404 () _buildManifest.js:1

> Failed to load resource: the server responded with a status of 404 () _ssgManifest.js:1

> Failed to load resource: the server responded with a status of 404 ()

(comment deleted)
Is it common to write units this way? My pedantic physics loving side would prefer USD/(TB * month).
It is. Per TB per month, per seat per year, etc.
Isn’t a/b/c == a/(b*c), also in a pedantic mathematical sense? I at least would consider a/b/c to be common (enough) and understandable.
Yes you are right, I am just used to a different format
But on the actual website, it says “$0.00099 GB/month”, which is incorrect.
It checks out with all the usual precedence rules. USD/TB/month is the same as USD/(TB * month). Really pedantic physicists would probably write something like USD.TB⁻¹.month⁻¹ which is, again, the same thing. Unless it is TiB and not TB, that would make the plan about 9% cheaper ;)
Perhaps a$byte⁻¹s⁻¹

A physicist generally should use the metric system. SI units convert to s⁻¹ from month⁻¹. Usually put the metric prefix first so T⁻¹ becomes pico. 2.6298e+6 seconds per month so pico becomes atto. Can't use B because that is magnetic feild in Tesla in SI units. $ is perhaps not as clear as USD, but feels like a better symbol to me (ideally with USD subscript but hard you write that here). I'm not a physicist, but generally I expect a physicist to use whatever is least ambiguous: which in this case is what was given (not my terrible attempt to convert to SI units!!!)

I'll take 10TB perpetual for free in exchange for sorting out your web server configuration
Apart from having zero guarantee about safety and reliability (not that a guarantee is enough), and website layout being pretty broken on a tablet, clicking on "Start archiving" hijacks the back button. Not cool.
Usually those unlimited storage for cents, but only if you pay up for 10 years in advance, plans are a scam. They give like 10kb upload speed, good luck uploading terabytes with that.
Was this site developed by chatGpt?

The visual inconsistency of this webpage suggests it was not written "by hand".

No LLMs were involved in making of this website.
I'm just waiting for CloudFlare to add Archive tier to R2 at this point.
Try Backblaze B2

https://www.backblaze.com/cloud-storage

$6/TB/month

(Not affiliated, just happy user)

>Not affiliated, just happy user

Just a quick question, do you know if you will be charged 2TB / $12 per month if you have 1.1TB of Data?

I believe they charge per GB-month. The $6/TB is just a human relatable interpretation of that rate.
I am a huge fan and user of Backblaze B2. For archival (essentially, my “tape” backups), I like S3 Glacier Deep Archive. It drives the storage price to approximately $1/TB/mo, though access isn’t instant and there’s a six month minimum cost on the storage capacity used. (Some other caveats and important considerations also apply. For anyone interested, do yourself a favor and read the docs.)

If that’s within your use case it’s fantastic. I use B2 to store my backup repositories, and S3 Deep Archive to mirror “master” stuff like my optical media backups, family photo archives, etc.

Microsoft has an almost identical product in Azure at a similar price point.

My only issue with S3 is egregious egress expenses.
The other day there was a discussion on Quora’s demise and I thought about how a lot of good ideas aren’t profitable.

My university gives everyone some storage and hosting for whatever. As long as it’s legal and tolerates the built in rate limiter, go nuts.

What if governments did that? Just a bit of free hosting and a bit of compute for anyone to do whatever with as long as it’s legal and tolerates the rate limiter. Heck this feels like a public library type thing. Give people the raw workshop space for their ideas.

I know there’s a lot of “yeah but…” issues, but I’m feeling optimistic about the idea.

I could see this working as a community project potentially. Instead of a physical maker space, it would be cool idea to have collectives built around shared digital storage/compute resources.
Yeah but.. who needs this? I don’t think the government would/should invest money into something that few need, has varying needs and is already a commodity?

There are far better services a government could offer, like a nation-wide login system. Imagine asking for a passport by clicking a couple of buttons and taking a selfie.

I wish there was a better intro or some sort of short paragraph describing a bit more about what their premise might be...Sure, its possible that i am not the expected audience/customer segment...but still, i mean, business is business, and should have some selling point, or baseline explanation. Sorry if i'm coming off as critical...but what if i am a potential customer - that neither you or i know could be a great match for each other - and i have budget to spend...why won't you help me spend my rhetorical money on you/your service? Nevermind, i think i'll spend my rheotrical budget on another provider who can tell me what value and/or service they actually can bring me.
There are interesting failure cases on the website, like the login page kicking me out to the front page reliably a few seconds into typing my email.

I wouldn't trust it with my data, but I like the entrepreneurial spirit from OP.

> Show HN is for something you've made that other people can play with. HN users can try it out, give you feedback, and ask questions in the thread.

> Please make it easy for users to try your thing out, ideally without barriers such as signups or emails. You'll get more feedback that way.

This seems more like an ad to get customers than anything else, doesn't it?

Oh my god why would you override the back button on the "start archiving" page, once you get there you're stuck and the only way to leave is to close the entire tab and open a new one. You can of course press "homepage" which will bring you back to the homepage.....but then pressing back brings you to the login page again. Arghhhh.
Right click on the back button and you'll get the full navigation history. Or long tap if you're on a touchscreen.
While I don't doubt there's some good thinking behind this, there just isn't nearly enough information for me to even begin to trust this. If I'm storing data which I might need again later I need to know, at minimum:

* Where is the data stored, and what's done to keep it safe * How / why is it so cheap, and how can I be assured the business will keep going once I've loaded data into it * What is my upload and access method? Is this S3 compatible, maybe something else? * What sort of terms are we entering into, what's the companies support obligation to me if I need my data tomorrow and there's a problem? * Do you have any existing customers, and if not why should I trust you with my data over AWS et al?

I'm sure there's some great thoughts about the business model, and probably a lot of hard work on the technical side. As a prospect customer though, I'm just seeing a price and a sign up button. It doesn't inspire any confidence to give over my email address, if anything it feels quite shady. This sort of information needs to be at my finger tips to form some level of trust.

I hope they're able to expand the information if it is a long term business venture. For all I know it could be amazing, unfortunately they're just not telling us why.

Looking at Alipay I can find 256TB drives to the tune of 26 USD. If I needed 3 of them for some reliability and change them every ten months, I still have a cost of ~ USD 0.03/TB/month. Why is offline storage still so expensive nowadays? Is it all simply the network costs?

(Apologies for promoting this advertisement thread unintentionally; the real question is for S3, Backblaze and the like.)

I think you mean 256 Gb
No please check Alipay/aliexpress for prices. You can also get usb keys nowadays for 32TB or more for less than USD 10. Maybe they are mostly scams, but if I were to extrapolate from price trends of a few years back when I bought drives more regularly either these prices are possible soon or we hit tech barriers I’m curious about.
Those are almost certainly fake capacities.
Cheapest I can find hard drives is around $30 per TB.

Which means that the break even time for this product is at least 3 years. Assuming personnel and infrastructure costs are near zero.

I can't see how this is a genuine product.

`LTO-9 Cloud Storage` `Affordable long term archival.` `for safely storing large sequential data` `instead of S3 Glacier`

A 10-pack of LTO-9 tape cartridges, so 180TB, can be found for ~$400. Glacier storage is used for petabyte-scale datasets you might only add chunks to years then read sequentially one time to train your next LLM. It's not for random access like a hard drive.

I don't know man, the website was first online on 2023-11-19, as reported by the first known https certificate, no information about who you are and what you're doing. I'm supposed to give you my data and my money/payment information, but as others like @bluehatbrit pointed out i know nothing else about your infrastructure, my data safety or the process behind the scenes. It takes time to save to LTO, how often is this done? Is my data going to be at risk for days?

The website looks and feels like it was made by some undergrad that's following django+bootstrap class and has to deliver an entirely made-by-hand project the next day. Not that it would matter much normally, but there's really nothing inspiring confidence on it.

I see you're using Next+Tailwind+Supabase and the main Next instance is being diretly hosted on a bare metal server from Scaleway, https and all, which is a somewhat weird choice for a framework like Next that works very well with standard hosted/serverless options from your usual big providers.

Considering there's not much logic going on, since the DB and auth logic are being handled by a hosted version of supabase, the homepage links redirect to github and the storage request flow is akin to sending an email, this could explain the HN hug of death, since no proxy/cache/scalable option was used.

Furthermore, the email verification link was sent from a simple gmail alias, with no confirmation on the website part.

I wouldn't dox you publicly, but your full name, surname and YOB are visible in the email sender information (along with your private gmail address). This confirms my undergrad project hunch.

All of the above makes me even more concerned about what could happen with any data uploaded here.

Ultimately I don't trust this, and I would advise anyone else not to.

There's nothing wrong with experimenting, learning and building cool stuff, especially while young, but I would have considered a whole lot more things before posting in a community such as HN