As Facebook’s monster initial public offering nears, it is useful to remember: none of this was inevitable.
And not all the credit for its success belongs to Mark Zuckerberg, or even Facebook’s
disarmingly plain, deviously clever user design.
Performance mattered just as much. That has to do with the engineering few people see.
And critically important parts of that engineering came from a combination of good advice that
Mr. Zuckerberg and his Harvard dropout buddies got from an industry graybeard and a chance design
move that paid off big.
When Facebook started in 2004, the big name in social media was Friendster, a service much like Facebook.
Friendster was founded in 2002, a year before MySpace, and quickly had millions of users.
It had frequent service problems, including outages and very slow page loading during times of heavy usage.
Managing social traffic on the Web, which can be very heavy and require a lot of small interconnections,
was very new.
When the venture firm Accel Partners invested in Facebook in 2005, it brought to the company as
a consultant Jeff Rothschild, who had earlier founded Veritas Software, a data storage management
company. Mr. Rothschild knew how to build big, complex software systems that didn’t break under
heavy loads, and Mr. Zuckerberg convinced Mr. Rothschild to take a position at Facebook,
overseeing the servers and advising the young founders.
“Getting Jeff in was probably the most valuable thing Accel did for us,” says Dustin Moskovitz,
who cofounded Facebook with Mr. Zuckerberg and others, and until 2008 oversaw the back-end technology
at Facebook. “I can’t thank him enough. I was 20, and he trusted me.”
It paid off well for Mr. Rothschild too; according to the Web site Who Owns Facebook,
his stake in Facebook should be worth $680 million after the shares start trading.
The lucky break, according to an executive involved in setting up the technology,
was a choice to design the site simply, and keep the focus on just a couple of things.
“Friendster tried to figure out paths between people, how many degrees you were separated
from someone,” said the executive, who spoke on condition of anonymity because he didn’t want
to criticize the competition even this late in the game. “That was useful for LinkedIn,
but not for social networking. We ignored that, and only looked at mutual friends you had
with someone, which is a much easier problem. They wasted unknowable amounts of energy on
that problem, which didn’t add much to their traffic.”
Friendster, in keeping with industry practice at the time, tended to put lots of information into
a single database, one of the reasons for system slowdowns and crashes as the business grew.
Facebook avoided the problem by accident; Facebook started at Harvard, then went through limited
rollouts at other Ivy Leage schools, then other colleges, then to the public.
The idea was to build anticipation, and learn by doing.
Mr. Zuckerberg created a separate database for each college, and wrote software to connect
people between them if they wanted to communicate. That turned out to be a much more stable way
to manage a social network. That model still exists, with lots of separate databases for Facebook’s
900 million users. They are now called “shards,” and new users are assigned to them based on what
the overall system can handle. If one database goes down, no one in the others notices.
Friendster, in a storied tale, turned down an offer from Google of $30 million in pre-I.P.O.
stock in 2003. Its membership topped out at about 115 million users in 2008, then quickly fell
away as Facebook grew. It was sold in late 2009, and has reemerged as a social gaming site,
largely serving customers in Asia. You can log into it with your Facebook account.