I'm pretty certain that I store almost a billion objects into S3 every month (yes: the requests for that cost a lot)... does that make me a noticeable percentage of S3's object count? Seriously now?
I wish it was interesting, or even sane... one day my website got like a million more users /that day/ and every database mechanism I had to do some of the user statistics I was keeping totally failed: the website ground to a halt. So, rather than decide what I really wanted to keep, I frantically developed a little library that stored all of that data into S3 as a giant key/value store, with my own little makeshift inverted indexes. Needless to say, this was stupid, and when I next checked my S3 bills I noticed I was doing a billion puts per month (which costs like $10,000). Lukily, this has only been a few months, and I'm now fixing the situation during a giant cost-cotti g endeavor... but yeah: omg that was stupid.
(typed on my iPhone and therefore a little shortened: hacker news comments suck on the iPhone)
Interesting, thanks for sharing. I'm a packrat myself, but I guess something like S3 where you actively pay for storage and transport shows you very clearly the cost of being a packrat--there may be value in what you keep around, but it's not free to keep around, and it's not free to move it to storage.
