Hacker News Reader

Comment by statictype | Hacker News Reader

If I'm calculating this correctly, S3 costs about $12.50 per 100GB per month (and that's not including any http requests you make).

How is Dropbox (I assume they store data in S3) offering a lower price than that?

I know they de-dupe data but is there enough common data being stored for them to be able to charge so little?

View on HN

I pay for a 50gb account but use about 5% of it. I would imagine the vast majority of their users don't use anywhere near the full storage amount.

Why do you pay for it, then?

I like paying for products that I enjoy, I figure I'd rather pay them $9/month than try and get enough referrals to get the space I need. Now it seems I've lucked out because of that decision I suppose.

And even among those who do, a lot of that will be purchased music, movies etc that they can dedup.

edit: not to mention shared folders. 45 GB of my account is a shared folder from work, which is probably shared among 10 users who all have to pay for the same data

They are definitely going to be in the top tier of S3 (5000TB+). If you do reduced redundancy, it's $3.70/month, $5.50 with full redundancy. There's plenty of room for them to make money there.

1) Not everyone has their account at max 2) They get bulk discounts from Amazon that take the price down quite a bit 3) They only store one copy of a file across their user base, so even though they bill you for 100GB and you have it completely full, it's possible they are only storing a small fraction of that. A folder of music for example is likely to not actually take up any space at all, but they get to bill you for every bit.

Except that the metadata on your MP3's is likely different, whether you've purchased them or just edited the genre in ID3 data.

I've often wondered if backup services "parse" MP3's into data and metadata for backup purposes, just because it can enhance deduplication so much. I've never heard of them doing it, but it's the first thing I would do...

Perhaps, but not a lot of people rip their own anymore. There are a lot of identical songs out there thanks to Amazon MP3, iTunes Music Store and all the various scene releases.

Whilst I don't disagree that few people are ripping their own stuff anymore, im pretty sure iTMS still puts the purchasers name in the metadata even now theyre drm free.

I can't recall of it was Dropbox or one of the other services, but I read they blocked up the files into blocks of a certain size. As long as the MP3 header doesn't change size (they usually contain padding or are at the end of the file to avoid this) it would just be one block that changed.

You don't need to worry about the particular structure of the files if you break the files into fixed-size chunks (e.g. 512KB) and do the dedup based on the chunks. The difference in the metadata of MP3 will be just in the first chunk, and the rest will be exactly the same no matter how you change the ID3 data. This strategy works for many other types of media files, too.

It's probably the utilization rate.

The average person on the 50GB Dropbox tier might only actually have 25GB of stuff. Now they're on the 100GB tier, but they still only have 25GB of stuff.

This is almost certainly why they also don't offer smaller paid accounts (like 10-20GB) -- they know users of such tiers will probably keep their accounts maxed out. I know I almost always utilize nearly 100% of my 2GB free account, and I'd probably do the same with 10GB.

Volume discounts and long term contracts will bring the price down.

Aside from de-duping and underutilization, Dropbox might get better rates from S3 than the rest of us.

If we share files using dropbox it comes out of both of our allowances (+ what everyone else said)

I'd be very interested in hearing more S3/Dropbox comparisons..

Reply on news.ycombinator.com