According to that link I provided (which was posted in April), Facebook was dealing with more than 50,000 database requests per second and they claim a 95% hit rate on their cache. That’s indicative of a lot of concurrent users (whether or not it’s above or below 4 million users is anyone’s guess).

Neither of us knows how many servers Facebook could cut down to but under various scenarios but even though I’m not a fan of the company, I’m sure they’ve hired some smart engineers and I doubt that they’ve missed an obvious means to scale effectively using 1/100th the number of servers and at a fraction of the cost they’ve already incurred.

Empirically I mean 5 servers for 1m signups (account creations). Assuming the site we run has similar attrition rates as FB (and it does), it’s the empirical number that you can get, if you highly optimize. We started from 1 webserver per 100,000 signups, and were forced tp optimize to this level, since we had no money. To be honest, I do not know about TBytes of photo storage: we thought we could not afford to store more than a couple downsized pictures per user from the start.
I can tell you more about the technical details if you are interested. Not really interested in talking to their HR person, since we have our own site to work with.

Regarding that comment with advertizing numbers, what she says makes a lot of sense to me, like the ecpms she quoted- home (or portal) pages versus profiles- I have some similar data here.

If you’re talking about 1 million *concurrent* users (the figure that reallly matters), I think you’re off by orders of magnitude.

Consider this: when it comes to photo hosting (which is not even the most resource-intensive part of Facebook’s service in terms of processing power), Facebook hosts over 10 billion photos that require over a petabyte of storage. It adds 2-3TB of photos each day and serves over 300,000 photos per second.

I certainly wouldn’t be surprised to learn that Facebook’s architecture has some significant inefficiencies but realistically, an application like theirs that’s used by millions of people daily is not going to be 100% efficient. It’s just not realistic.

Earlier this year, GigaOm cited reports that Facebook was running 10,000 web servers. When it entered into a $100 million deal with TriplePoint Capital, BusinessWeek reported that it may be purchasing up to 50,000 more.

The bottom line is that Facebook is not getting by with 500 servers. If you think you can get them down that number, you’re talking to the wrong person. You should be in touch with Facebook HR, although I certainly wouldn’t be too excited about the stock options.

For what it’s worth, Facebook is already using horizontal partitioning, along with the rest of the kitchen sink:

http://venublog.com/2008/04/17/notes-from-social-graph-and-the-database/

As for “Arianna’s” comment, she doesn’t know what the fuck she’s talking about. She’s pulling the revenue breakdown for MySpace out of thin air. She then goes on to assume that since Facebook’s US metrics are approximately half of MySpace’s, we can just cut in half her estimates for MySpace revenues to arrive at Facebook’s revenues. Uninformed estimates on top of uninformed estimates.

Facebook isn’t in a position to cut 80% of its expenses. At this stage of the game, that’s kind of like saying that if Henry Paulson could get rid of 80% of bad mortgage debt by next week, the global economy would start to perk up. Nice idea, not going to happen.

I think this person knows what she is talking about. So I still stand by my statement that if they cut everything by 80%, they can still grow and operate. But that’s a big IF of course.

As for revenue, it’s worth keeping in mind the following:

• From what I understand, a fairly substantial chunk of Facebook’s revenue is guaranteed by Microsoft. A deal like this cannot be relied upon over the long term.
• The fact that ads on social networks don’t produce an ROI for the vast majority of advertisers is well known and after four years, Facebook hasn’t been able to find a solution to the social network ad puzzle. Its potential in this area is therefore decaying every day, especially now that many advertisers are fleeing to quality due to the economy.

I suppose the real problem is that companies like Facebook reach a point where they don’t have anywhere to go. Investors have “dressed” them for “jobs” that they’re not in the running for.