Show HN: A simple storage pricing calculator for AWS | Hacker News Reader

Show HN: A simple storage pricing calculator for AWS

duckbillgroup.com

169 pointsQuinnyPig78 comments

Given my experience with AWS, I doubt this is reliable in any way... Every bill I've ever seen(and paid personally for that matter) felt a bit like someone played darts blindfolded. One that really struck me (around 3 years ago when I decided to move over to gcp) was two consecutive months: my usage was 99.2% identical. First month my bill was around 151.01 euros. Second month it was 235.32. Everything was absolutely identical, API's usage, storage, traffic, everything. I compiled a report sent it over to their support and I got a convoluted response which could roughly be summarized with ¯\_(ツ)_/¯

> my usage was 99.2% identical

What does that mean? If you're talking about S3, things like PUTs are way more expensive than GETs for example (by two magnitudes). So your usage can be "nearly identical", you can still spend money on precious GETs.

Example: Last month I spent around 300 EUR on food. The month before, I spent 150 EUR, yet my food was "97% identical", the 3% being those two times I ordered 75 EUR of sushi takeout.

PS: A lot of people get frustrated/confused by AWS pricing and, fair enough, but for a lot of businesses it's often a feature that they allow developers to architect their applications with costs in mind. It's a knife that cuts three different ways.

Yeah, I know how that works. And no, none of that. In my case it was renting a GPU for training a model. No put traffic involved in either of the cases, the data was in a s3 prior to that. It was simply tweaking hyperparameters in the code, nothing more. On both occasions it took the exact same time, exact same to train, exact same resources. There were apparently no changes to the pricing but still. I'm referring to aws's quantum bills as a whole. Never had such issues with gcp... Yet...

Does cost explorer show no difference in your top costs?

If you're willing to share the reports you wrote up, I'd be curious to take a look at them.

Cost explorer should show exactly what was going on.

In that case, that sounds like Amazon is charging you differently based on how other people are using their resources. In the times of surge pricing, I wouldn't be surprised if they are doing that to make it so they make the same amount of profit, regardless of how much you are directly using, but depending on whether they are getting enough money from all tenants of a particular shared host. So if other people are paying less, they jack up your pricing to equalize things out so at the end of the day, their systems are still hitting their profit target.

All of AWS' billing criteria are very clearly published, and IMO, reliable. The variable cost in most scenarios is active usage (ie, API calls). The inactive usage (storage) is very cheap and predictable.

I am not sure why this is the highest-voted comment. The OP has shared a tool which performs a calculation for a pricing estimation. This comment does not address the metrics, or the accuracy, of the calculator. It makes a complaint about AWS pricing, but does not attempt to evaluate or comment on the post, or even provide evidence for the unrelated claim about AWS.

HN comment sorting is weird. It’s not based solely on upvotes, as far as I can guess. Probably a mixture of upvotes/downvotes/flags much like the homepage is.

The post is commenting on the pricing calculator; if AWS's pricing varies month-on-month for the same usage then there is no way a calculator for it can be accurate even with the caveat that it's "only an estimate." An estimate that's out by 50% is effectively useless.

AWS is deprecating their trusty pricing calculator in favor of one that makes you describe what you're building first, and this makes me sad.

This one only works internally to one region at a time, but it compares various costs in different storage mechanisms on an apples-to-apples basis as best I can.

> AWS is deprecating their trusty pricing calculator...

Is AWS deprecating the simple-calculator [0]? Frustratingly, it has been neglected and has never supported many aws services.

> ...in favor of one that makes you describe what you're building first, and this makes me sad.

https://calculator.aws is pretty good, to be honest.

Or, are you referring to the TCO calculator [1]?

---

[0] https://calculator.s3.amazonaws.com/index.html

[1] https://awstcocalculator.com/

https://calculator.aws is good but it takes extra clicks. Having something lighter (if all you care about is storage) is pretty nice.

1TB in GCP Cloud Storage's Archive storage class is just $1.20 per month. https://cloud.google.com/products/calculator/#id=8404213d-ef...

Retrieval is within seconds, not hours like in Glacier.

Impressive to me but $50 to retrieve 1TB vs cheaper tiers on AWS for slower. I don't know what GCP's retrieval throughput is though.

There's also B2, with 5$/TB for storage & 10$/TB for retrieval [0].

---

[0] https://www.backblaze.com/b2/cloud-storage-pricing.html

Disclaimer: I work for AWS.

A counter point to think about. Archival storage needs to inform engineering behavior as well. If I'm archiving something knowing that retrieval will take longer or is expensive, I will make smart decisions about what I should be archiving. If I archive knowing that retrieval is in seconds, I'm probably dumping everything in the archive.

I don't know how popular GCP's archival storage is, but when it grows and customers start using it simply as a cheaper hot storage (because it is fast and cheap), they'll have to rethink their pricing.

I like the design, but is anyone choosing storage based on price? I can see comparing cloud-to-cloud, but within a single cloud the type of data (blob, block, database, etc...) and the frequency of r/w prescribe the storage you should be using regardless of price.

I just don't see a scenario where I say I'm just going to move from EFS to EBS or such because its 8% of the price. I chose EFS for a reason in the first place.

Not meaning offense, I'm just unsure of usefulness. Appreciate all the content you provide though! Didn't realize who you were :)

>I just don't see a scenario where I say I'm just going to move from EFS to EBS or such. I chose EFS for a reason in the first place.

How about when your team has terabytes of data stored in EFS that is only used sporadically and now you want to move it off? (maybe you started with EFS because it makes the most sense for your application but now you have a lot of older data that you still want to keep but don't necessarily need it to be accessed by the application anymore). Something like this helps justify how much you could save by starting that migration project.

I do see your point though. IME, calculators like this are used more for management who are trying to estimate their budget for planning purposes (as in, I have a team that tells me they will need 5 TB of S3 storage, how much is that going to cost me?), not necessarily by devs making architecture decisions.

If you have the latitude of moving from file storage to block storage you should already be doing it. I'd say you chose the wrong use case from the start, which also backs up his point.

The point is that you may not be able to use block storage at the start, but as time goes on your needs may change, giving you more latitude.

Price can affect design decisions - i.e. the number of times you access backups for analysis can depend on takeout cost.

Would anyone choose glacier over normal S3 if they were both priced the same? I think it's literally the price difference that drives you to choose Glacier for archival data, no?

I agree wholeheartedly.

Personally, I use this to reinforce architectural change suggestions along the lines of “teach that application to speak to S3 instead of thrice-replicated EBS and you’ll save X.” It’s helpful as a demonstration point.

It's clearly missing storing data uuencoded in Route53 TXT records.

Or you can get 75GB free by using Lambda....

https://www.lastweekinaws.com/podcast/screaming-in-the-cloud...

(Yes this is a joke)

It's because of thoughts like this that the number of tags per item is limited.

Initially they were going to allow unlimited tags until someone pointed out that you could use tags on your items as a poor mans key/value store for free.

I'm reminded of people who use S3 as an eventually-consistent database. Encode your rows in CSV and use them as names of empty objects. Query using paginated LIST requests for a prefix of the columns.

AWS already offers a free key value store - SSM Parameter Store. It’s severely rate limited if you don’t pay for it - but it’s still cheap.

The price for Glacier Deep Archive is wrong.

I entered {50GB, US-East-<either>} and got a price of $1.05/mo. I think the actual price is $0.0495/mo (just under a nickel).

(Less worrying, I don't see Glacier [normal] at all.)

You found two bugs. Awesome! Fixing.

I also never use normal Glacier at all (which is how this got missed in testing), but the pricing's all there on the backend. It's a frontend thing...

...aaaand it's an AWS renaming thing in their pricing API. Lovely!

AWS is so bad at naming things that it apparently now breaks other things too.

Everything is relative: compared to Azure virtual machine family naming, AWS is not so bad!

Maybe related: Reduced Redundancy is showing as more expensive than Standard (for same 50G/US-East sample data)

That’s a real pricing facet! See https://www.lastweekinaws.com/blog/s3-reduced-redundancy-sto...

So my cost is $69. But it doesn't say per what!

I assume it's per month, but it really should say.

Not being in a position where I'm billed for storage (work does that), I didn't really know either. Guessing per month as well.

It's back up in the top right:

"Your monthly payment will be:"

Not on the page I get: https://www.duckbillgroup.com/aws-super-simple-storage-calcu...

bit odd but only visible after selecting options for both:

"Where is your data currently?"

"Which tier are you using?"

I was also wondering, and my initial guess was per year. Otherwise I can buy myself a 2TB HDD for the cost of 1-2 months of cheapest equvalent storage at AWS.

Sounds like a gold mine for Amazon at any rate.

Glacier hasn't really seemed like it's worth it for me and looking over this calculator it's almost as expensive as Standard mode? Why would I use it over Standard, let alone Intelligent Tiering - Infrequent Access.

I have no earthly idea, personally.

This calculator is either giving completely wrong outputs for the Glacier pricing, or it's secretly throwing in some of the other costs like data transfer/API request costs.

For example, for 50 GB it's showing a cost per month of $1.15 for Standard, which is correct (Standard is priced at ~$0.023/GB). For Deep Glacier, it's giving pricing of $1.05 (which is wildly incorrect, as the price for Deep Glacier is ~$0.00099/GB, meaning the monthly cost should say $0.05, or less than 5% of the Standard cost).

Yeah, we're investigating what's going on with this. Seems there's a bug in the public price list we're trying to work out.

The fact that a 3rd party feels the need to publish a calculator suggests it is not ‘simple’

The mindblown platypus when you enter a high petabyte value is cute. I like it :)

This doesn't seem quite right at least for Glacier Deep Archive. I put 12TB in Ohio in Deep Archive and it is saying it should cost $250. It should cost ~$12. Maybe just some math typos in the calculator?

Did you consider to replace the AWS' names and technical terms by generic terms and their explanations, so it's easier to understand for people not familiar with AWS' storage offers?

Yes, but that feels a bit out of scope for this calculator. Including all of the trade-offs between different storage services and when to use each results in basically rewriting the AWS marketing pages.

That's a different project of mine. :-)

Related: AWS in plain english (re-write of just the product names), https://news.ycombinator.com/item?id=10202286

S3 transition costs rely on knowing the number of objects being stored and are independent of storage size. If you’re moving between S3 tiers, this tool provides no input for the number of objects being transitioned. These costs can be non-trivial. Predicted savings involving S3 transitions also need to consider minimum storage duration. Consequently, this feels like a misleading tool. It would probably benefit from showing proof of working.

What... "Standard - Infrequent Access" & "One Zone - Infrequent Access" is cheaper than Glacier Deep Archive??

5TB in ap-southeast-1 is what I plugged in.

Either the webpage is incorrect or uses a non-obvious formula to calculate inclusive costs (Data Retrieval, Data scanned, API charges along with Data storage)?

Re: Storage: Deep Archive's ~$0.001 per GiB whilst IA buckets are ~$0.01; a 10x difference.

API charges for data retrieval from Deep Archive is 100x (?) (~$0.1004 per thousand reqs) expensive than IA buckets ($0.001).

Data retrieval for Deep Archive is 2x ($0.02 per GiB) the cost of IA ($0.01).

https://aws.amazon.com/s3/pricing/

This is nice, but is it the price per month? Per year?

It's monthly, but yes they should add this information.

It does say "Your monthly payment will be"

Oof. Fixing. Good catch!

This is neat.

https://www.duckbillgroup.com/aws-super-simple-storage-calcu... is a way to show "how much more expensive is X than S3"?

This is cool, but the EFS infrequent access pricing for Australia looks wrong. Shouldn't 100GB cost $2.72 (not $1.20)?

With the cloud, I have learn about computers and I have to learn about AWS/GCP/Azure's version of computers. The cognitive overload just barely equals not having to maintain my own boxes.

Pick you poison, maintain your own boxes or maintain your own YAML files and spend a lot of time trying to figure out billing.

Newbie question: which Amazon storage services can be used to add storage to a VPS ? Storage that would be used like a secondary disk. My VPS provider is asking way too much to my taste ($50 a year for 40Gb disk I think) when comparing with Amazon.

Is it possible at all to mount some amazon storage like I would for a block device ?

There are FUSE adapters to mount S3 and similar services... some have various caching options, and ymmv a lot.

I would test your latency from your VPS to the various S3, Azure etc locations... May also want to consider Digital Ocean's storage which is API compatible to S3.

> May also want to consider Digital Ocean's storage which is API compatible to S3.

Don't forget to rclone to second provider though.

If you can move your VPS over to Amazon, EBS (Elastic Block Store) is what you are looking for, but it's likely not much cheaper than you're currently paying (EBS is priced at $0.10/month per gb, so $4/mo for 40gb of general purpose).

You can mount S3 buckets as a hard disk on a VPS, but that isn't the main use case for S3.

$4/mo, not $40/mo

Lol yep, thanks. Fixed. I think I was an order of magnitude off because the parent comment was stating his costs in years. D'oh.

You'd have to use EC2 or Storage Gateway (to expose S3 as a pseudo-block device)

`QuinnyPig - I love the extra mind blown platypus touch when you enter 40000 PB

https://www.dropbox.com/s/eqdpkiv0ikevsm4/IMG_1937.jpg?dl=0

The mindblown graphic that comes up when you enter a rather large data size is delightful :)

This list doesn't showcase FSx[0] as an option.

---

[0] https://aws.amazon.com/fsx/

Would be nice to see other cloud providers for comparison

Generally I find that whenever I'm doing storage calculations it's for existing workloads inside of a given cloud provider. "We're moving the whole thing to GCP because it's going to be X% cheaper" sounds compelling, but... I just don't see it happening at anything other than small scale.

Excellent design :)

Thanks! It didn't fall flatypus.

0 for dynamodb ?

is this just an ad?

Reply on news.ycombinator.com