In fairness, while I think the previous comment is a bit clueless myself, I'm inclined to ask how many of those servers would be necessary if FB wasn't logging so very much data on everyone.

If FB's job was once again just to show me my fucking posts, instead of building the most complete profile of a human being in the history of mankind so that it can sell me out to governments and corporations, it seems to me the data load would be a lot lighter.

I don't have a source for this right now, but I remember reading that the vast majority of disk space used by Facebook is taken up by users' photos. They are probably saving hundreds of uploaded photos per person (on average) and while they do compress them quite a bit this still has to take up a lot of storage space.

Of course they do also collect a lot of behavioral data, but I imagine all the data Facebook has collected on me probably uses up less storage than my profile pictures alone...

Also I remember reading that they keep 7 copies of everything at all time a few years back in an article about how deleting your account didn't actually delete your data, which became obvious after data from accounts deleted several years before had reappeared after the introduction of a new feature.

> I'm inclined to ask how many of those servers would be necessary if FB wasn't logging so very much data on everyone.

Hmm, but that's how they make money. They log, they analyze, they show you ads and they sell your data.

I doubt they really sell your data. The only way advertisers can get info about you is if you click through an ad.

Edit: So downvoters, how do I buy data about Facebook users?

You cannot and you are right there. The only way to target individuals is by having specific lists of emails or ids that you can upload and create custom audiences (or via their notion of retargeting through pixel fires). Even then though you don't know WHO saw your ads. Its a walled garden and impossible to get per user data as you would in traditional retargeting campaigns. That is why the bidding system is impossible to game (despite some claims from some companies). You have no idea WHO is seeing your ads - only the DEMOGRAPHIC of the ad views/clicks and only if you segmented your campaigns in a way you can track that.

Right now, this would be insider stuff and asking for much trouble to reveal it. Meanwhile, you just have to wait for it to be obvious that facebook has failed and it will start selling user data directly, or that they get hacked and a dump is available on the black market.

In the meantime, find yourself a dodgy intelligence agent or sysadmin.

I wonder how many email servers are running today? Probably comparable within an order of magnitude. Yet those server are not owned by one corporation, they're multiple corporations cooperating to move data between each other.

It's too bad Facebook and Twitter are businesses, instead of multiple protocols and multiple businesses. Or in other words, it's too bad The Old Ones didn't think cat pictures and what they had for breakfast was important. https://en.wikipedia.org/wiki/List_of_Internet_pioneers

EDIT: Changed to intended link.

WhatsApp is of comparable scale and is doing this with 100x less servers.

As per [1], in March 2014 they were handling 20B messages/day (with 600M pics, 100M videos) with just about 550 servers.

My guess would be that, every day more photos are shared on WhatsApp than on FB.

Major difference: At present, WhatsApp does not "keep" your messages or sell "you" to advertisers.

[1] http://highscalability.com/blog/2014/3/31/how-whatsapp-grew-...

Its not comparable scale. WhatsApp are shifting messages from point A to point B and that's it. Even 100x less servers is still a LOT of servers just or that. They don't store data, just traffic it from one user to another.

Okay, 60,000 servers - for 1.23 billion active users. That's less than 0.0005 server per user. Facebook as such may cost millions to run of course, but the price they pay to keep your account active is so negligible that you're essentially selling your data for peanuts.