Monday, February 11, 2013

The man behind Google Docs is now trying to reinvent the web app at Box


Sam Schillace (Credit: Google Ventures)
When Sam Schillace first crossed paths with Box Founder and CEO Aaron Levie in 2006, it didn’t exactly go how Levie expected. Box wanted to buy Schillace’s document-collaboration startup called Writely, but shortly after discussing it, the team at Writely went dark. A few days later, Levie got to hear from the press that Google had purchased Writely instead.
The rest, as they say, is history. Writely went on to become the foundation the foundation of Google Docs. Schillace went on to lead just about every product that falls under the Google Apps umbrella, leave for a stint to launch a new company, then come back as a member of the Google Ventures team. Box, well, it’s flush with cash and set for a 2014 IPO.

But although their paths diverged, Schillace and Levie would find each other again. In 2011, while still with Google Ventures, Schillace joined Box’s technical advisory board along with then-Facebook VP Jonathan Heiliger, LinkedIn SVP Kevin Scott, then-Flipboard CTO Arthur van Hoff and some other notable Silicon Valley names. (Schillace calls himself “the dumb guy to make all the other guys seem even smarter.”) It was good camaraderie and a lot of fun, Schillace said, and “once a quarter they’d bring all their really hard technical problems to you.”
Then, in mid-2012, Box COO Dan Levin asked Schillace to come on as the company’s vice president of engineering. He actually didn’t want to take the job initially, but when he realized the decision was really between continuing to have conversations with the smart people around him — both on the advisory board and within Box itself — he knew what he had to do. ”I really liked the people and I really liked the team,” Schillace told me.
He officially joined the company in August 2012, and is now bringing his considerable expertise to bear on Box’s very pressing technical challenges. With crushing data volumes and millions of users, Schillace explained, “There’s a little bit of a success disaster we’re having to deal with here.”

Laying the groundwork at Google

It’s a good thing Schillace has some experience dealing with some difficult issues, in both technology and business. When he started Writely, even his co-founders thought it was a stupid idea; web browsers then were so primitive. There was no documentation, no standardization (even around how they handled HTML), and browsers were slow and underpowered. The only solution was to “just poke at the browser until it behaves,” Schillace joked.
“The crown jewel of hard problems … is around collaboration,” he explained. Although the Google Docs team eventually rebuilt Writely as a native word processor, Schillace and Writely co-founder Steve Newman didn’t have that luxury. Because the browsers all behaved differently, they had to write some “hairy” logic to merge changes on the server and then send it back down to the user in a format that particular browser could understand. Sometimes, they’d run into what Schillace calls “edit fights,” where a browser would actually fight itself.
“It was totally Wild West,” he said. “We had to invent a lot of stuff even to get it to even work a little.”
Credit: Google
Credit: Google
Schillace also knows how to overcome, or at least manage, the non-technical issues that come as a company grows from a small team into something much, much larger. When he started at Google, his team was about 40 people. By the time he left, it had grown to 600.
But despite its size, Schillace said, Docs was never a huge priority at Google because of its minimal effect on the company’s bottom line.
“The only reason we managed to make Docs happen in the first place, really, was Eric Schmidt was such a big supporter and we were kind of tyrants because we were so feral and so much in the startup mentality,” he acknowledged.
Hellbent on earning every dollar possible from their contracts, Schillace and his team worried much more about continuous execution than about following the rules. “Every once in a while we’d get in trouble and Eric would protect us and get us out of it,” he said. “We just sort of shoved through it and made it happen.”
Fast-forward a few years and there’s “not a lot of room for the feral programmer at Google anymore,” Schillace explained. Because the company is so big, everything you do is under extreme scrutiny and it becomes impossible to recover from even small mistakes (e.g., Wave and Buzz) because there’s so much exposure. And as part of such a large company working on such large projects, it can become hard for individuals to stand out or to see the direct effects of their work.

Building a better Box …

Box, Schillace said, feels a lot like Google did when he first came on board there. It’s a big enough company (with nearly 700 employees) that everyone is doing real work, but not so big that all the problems have been solved. He thinks there’s a good two to four years of really fun stuff to work on, and the company can bring in smart people from places like Facebook and Google to help find the solutions.
For Schillace personally, Box’s size also fits in with his current state of mind: He’s done with the emotional roller coaster of startups after spending more than 20 years building them. “When you’re initially a programmer, you learn to program with 10 fingers,” he said. “And then maybe if you go to management you learn to program with 10 people. And then if you go to the level of management that I’m at, you get to program with 10 teams of people.”
But works that’s “fun” isn’t necessarily work that’s easy. Technologically, he said, the hardest problems at Box are dealing with the company’s scale. From a data volume perspective, it’s growing at 10 percent a month, which means it’s doubling roughly every seven months and growing by about a factor of four every year. Box’s storage capacity is into the multiple petabytes now, Schillace said.
On the infrastructure side, that means Box is always working on the next generation storage of scale-out storage architecture. It takes between 6 and 12 months to build one, he explained, but they only last about 18 to 24 months.
In the case of Box, though, its focus on real-time collaboration means mere storage capacity isn’t enough when designing a database. That’s why Schillace’s team ends up building a lot of its own technology to deal with the company’s unique needs. It’s able to buy a few things off the shelf and use some available open source tools, but, as Schillace said, “every company has its own set of tradeoffs you have to make, so you basically end up having to build this stuff yourself.”
And as Box signs up more deals with companies that have tens of thousands of users — sometimes more — it has to make changes around caching and metadata to account for that many people potentially sharing a folder or a document. ”That does strange things to your product,” Schillace explained. “… All complicated by the fact that everything’s moving all the time.”

No comments:

Post a Comment