Ask HN: Why is cloud storage so cheep? | Hacker News Reader

Ask HN: Why is cloud storage so cheep?

14 pointshenningschuster12 comments

Today Dropbox dropped prices. They now offer 1 TB for $10/month. Other providers like Google Drive have similar offers for quite some time.

I wonder how they do this. Dropbox uses Amazon S3 to save files. The lowest possible price on S3 is $0.0275/GB/month. If a Dropbox user utilizes all available space it would cost Dropbox $27.50/month. In addition most users are not paying for Dropbox but use there free offering adding files (=costs) to Dropbox. Running the company, development of the application, application server, download traffic etc. will come on top.

I see the following reasons why Dropbox might be able to make profit in the end:

- Most people are using much less space than the maximum of 1 TB. - Dropbox is doing deduplication. If two users have the same file Dropbox only needs to save it once. - Dropbox might have a better deal with Amazon and is not paying $0.0275/GB/month.

Do I overlook something? What is most important reason why Dropbox and others are able to make this prices?

Dropbox used to say they use Amazon S3 [1], but now they just say "storage servers are provided by a managed service provider" [2] which could mean they found someone cheaper than S3, or that another company is managing the infrastructure for Dropbox. I always imagined Dropbox would move on from S3, and Drew even said back in 2010 that "We may not be on s3 forever, and will build our own store in addition, but for now this lets us focus on both the client software and the performance of the layer we’ve developed on top of s3." [3]

1. https://web.archive.org/web/20140101095459/https://www.dropb...

2. https://www.dropbox.com/help/7

3. http://www.makeuseof.com/tag/dropbox-review-invites-and-7-qu...

A huge part of it is the practice of overselling. The basic idea is that you have n customers who all want k gigabytes of space, but on average they each use only j gigabytes of the space. If that holds, then rather than having nk gigabytes available, you just need n(k/j) gigabytes to satsify their needs. After amortizing usage, the people who overbuy and barely use their space more than makes up for the users that max out their space. I don't know for sure if Dropbox does this, but most hosting companies do, and its a sensible practice (especially when your backbone is scalable).

Pairing with that, nobody actually uses all of their space. If you see yourself getting within some percentage of your usage, one generally will buy more to keep from hitting that upper bound. Nobody likes being in the dangerous red zone.

I believe dropbox also compresses your data serverside, so you really use up n*p gigabytes (where n is your local usage, and p is the compression ratio), but don't quote me on that.

Pretty much nails it on the head. Overselling is a large part of online services overall, like web hosting.

You think HostGator can actually give you unlimited disk space for a few dollars a month? People don't use anywhere even remotely near their quotas, but it's a compelling marketing/sales tactic for sure.

Like banks do with money, they have an X capital and lend (or buy bonds) for 10000X. Only difference is if Dropbox fails, the tax payer isn't gonna pay the mess indirectly... He will lose the data.

The cost of the space is quickly approaching meaningless [1] and people are streaming more and more which reduces the need for space in some heavy areas - you don't need to store local copies of music, movies, tv shows etc anymore when you have on demand unlimited access to them.

Expenses also get eroded by constant progress towards automation. As one example it was only a few years ago that Heroku pioneered a way of deploying apps that removed a shitload of manual labor maintaining and scaling apps, which has now become widespread with competition and freely available through open source.

[1] http://www.pcworld.com/article/2599484/seagates-monstrous-8t...

Also in that article - "Right now on Amazon you can get a 4TB desktop drive for just under $200, with 5 and 6 TB drives edging closer to $300." Dropbox's $100/year for 1 TB is just closer to what it should be.

I think you have it there.

A terabyte is a nice round number, but I think getting a TB into cloud storage would be challenging, since most people have lower upload than download allowances.

At 1 Mbit/sec upload, for instance, it would take 8000 sec to upload a GB or 8 million sec to upload a TB, which amounts to a few months of continuous uploading.

That's a serious project for most people.

As for the other factors you'd better believe that they use compression, deduplication, special deals and every other trick in the book to bring down costs.

On top of all that, Dropbox has received over $1B in funding so it is very possible that they are losing money to build market share.

So far as cost analysis, I'd also look at the networking costs with AWS, since those are going to be significant. AWS doesn't charge for ingress, but it does charge about 5 cents to move a GB out at high volumes, and that can add up if a lot of stuff is being moved out.

So there's a couple things going on that would make this feasible.

1) People are unlikely to use that much actual storage in the first place. The tipping point, assuming their on S3 and haven't moved to a cheaper option like some other comments have suggested, for the $10/mon is ~363 GB. That's a lot of data for a single person to upload especially considering the poor state of US internet speeds.

2) There are technologies to push that break even point even higher from the customers perspective. Deduplication and compression will shrink many of a users files while still increasing their used count. It's really hard to estimate the effects of dedup and compression though because of the number of unknowns.

I think you've covered most of the reasons. Compression probably also reduces the actual storage size of files. They are probably also using some of their own storage outside of Amazon. Rarely accessed files could be moved to slower, cheaper storage facilities. But you wouldn't get the full answer without access to internal Dropbox documents.

Dropbox uses file deduplication, also I suspect the distribution of disk space used is nowhere near the maximum offered (this is, they offer 1TB but most users will take up only a small fraction of it)

I'm sure owning a large market share of a service storing consumer data is worth something, as unethical as that sounds... Abuse is fairly common as we've seen in recent years. When google scans a file into it's own format, or when OCR's run on your images to index information, who else can access these indexes?

*cheap

It's all the birds. Up in the clouds. Cheep cheep cheep.

Reply on news.ycombinator.com