Hacker News Reader

Comment by simonw | Hacker News Reader

My mental model of this is that Snowmobile was for companies with their own data centers looking to transfer petabytes of data to the cloud so they could stop self-hosting it.

The number of companies with that problem is going to naturally fall over time. Companies that are already using the cloud don't need Snowmobile to move their data.

Illustrative point from the article: "A spokesperson for satellite operator Maxar said the company used Snowmobile once in 2017 to move more than 100 petabytes to AWS from its own servers. Snowmobile was an “ideal solution for that moment in time,” the spokesperson said."

You're not going to get many repeat customers for that service!

View on HN

I'm not sure how much I can safely say, given I was employed by AWS, and provided some element of support to the snowball team from time to time, but Snowmobile was originally targeted towards one or two particular customers that had particular needs. Getting their data shifted was more than going to turn a profit on the whole exercise. Any other customers using the service was a bonus.

The amount of the data being shifted was one of the particular needs, but it was more the _other_ aspects of it being a huge truck that entered in to the equation.

I wonder how much they pay for 100PB in AWS, depending on the storage service/tier. Also it will be quite expensive if they ever want to move out of AWS. Or is a project like this considered to be "eternal"?

It’s actually hard to say. The prices these companies pay can be very different than list prices for us mortals. It’s one of my biggest complaints about AWS; the prices seem high, but in reality they are very opaque.

they probably pay around 40% of list, depending on how much their ARR is.

This really needs to be made illegal already. All prices should be public and you should not be allowed to charge different customers different prices for the same automated service.

I should be able to store my data at the exact same $/TB price as Epic Games, for example. They once mentioned it costs them less than a penny to serve several GB from AWS - why the fuck is it more expensive for me?

If you are willing to sign a contract and commit to spend $ millions per year on AWS services for 5+ years like Epic Games did, then sure, you can get the same $/TB price they got.

At some point of very large usage, you also have to work with AWS teams to give input on your demand forecast.

Unless you have such large usage you get list price, that also gives you the illusion of infinite elasticity.

This is not exclusive to AWS though, all hyperscalers have this policy.

Because you’re a more expensive customer per dollar of revenue, statistically - there are constant costs per customer for support, billing, etc.

Billing is automated. What support? S3 doesn't care whether it serves a PB of Epic traffic or from thousands of small customers, it's all automated. There is no actually reasonable explanation for the service being 10-100 times more expensive than greed.

Cards get declined. Support tickets get created. Customers die and you have to hold all their stuff open for 90 days while trying to reach them, without a chance of getting paid.

You might not be stupid, but there are people who look _just like you_ who will email support over a six cent billing issue which turns out to be correct anyway - but just burned a couple hours of support people’s time.

Epic probably doesn’t do stuff like that.

Would it also be illegal to have 2 very similar services, with no small tier on one of them?

Every business negotiates volume discounts. Do you think consumer goods companies charge Walmart for products at the same rate that they charge a mom and pop shop?

I’m sure if you had the same volume as Netflix you would pay less than the retail price to.

If you ever flew on a plane, I can guarantee you that there are probably 20+ different prices being charged for the same seat.

Many jurisdictions do require prices to be clearly displayed for goods, but services are a different beast. You have limited resources to deliver these services, customer needs can be complex, and differ across regions.

So what you are saying is that if I buy one Tesla, and you buy one hundred Tesla cars, they are not authorized to give you a discount?

> They once mentioned it costs them less than a penny to serve several GB from AWS - why the fuck is it more expensive for me?

You spend far less money than a corporation, and there are customer costs beyond the actual bandwidth/compute charges.

> This really needs to be made illegal already. All prices should be public and you should not be allowed to charge different customers different prices for the same automated service.

Why? Different people get different prices for products all the time. It's fundamental to a functioning free market.

Why should it be illegal for service providers to negotiate prices?

The EU has forced the providers to allow a “final egress on account closing” at no cost (or i should say, they are offering it at no cost). GCP was first in January [0], the others have since matched.

0: https://www.theregister.com/AMP/2024/01/11/google_cloud_egre...

Presumably doesn't apply to the cold archive tiers where a great deal of the cost is retrieval/egress?

It's also worth noting that the migration service only applies to data stored in one of the following services: BigQuery, Cloud Bigtable, Cloud SQL, Cloud Storage, Datastore, Filestore, Spanner, and Persistent Disk.

Still seems you would save a bundle if you had to pay to promote the data from archival to hot and then do the free exodus.

the article lists the following for gcp: BigQuery, Cloud Bigtable, Cloud SQL, Cloud Storage, Datastore, Filestore, Spanner, and Persistent Disk.

Using the calculator (https://calculator.aws/#/createCalculator/S3), AWS S3 "infrequent access" is $1.3M per month for 100PB. Not sure if that's single zone.

If you put it in Glacier Deep Archive it’s closer to $1.3M per year, just to set a lower bound.

Deep Archive has retrieval times up to 12 hours, so for on-prem/colo that is then competing with buying a few tape robots (you can get a single tape archive that handles >1 exabyte, but you'll of course want redundancy without having to have staff shuffling tape in an archive that will run into the thousands). You can buy management software that tries to replicate a glacier-like experience for things like SpectraLogic tape libraries but they may not be the cheapest option.

It ends up becoming a very different beast than the storage classes w/fast enough retrieval guarantees to need disk, in any case.

Unlikely that they want it hot or even warm. That data almost definitely went to Gkacier while they worked out what to do with it.

You overestimate the enterprise.

I would have expected we are close to steady state. The number of companies cyclically migrating too/from the cloud seems unlikely to decrease.

Reply on news.ycombinator.com