Inside Tumblr’s Efforts to Make the Site More Reliable

By Cameron Scott 

blogging, social networks, social media, tumblrLast week, Tumblr hit a major milestone: The platform now hosts more than 100 million blogs and sees more than 150 million unique visitors a month.

Yet, Tumblr is still something of an enigma. The company hasn’t clearly defined how it will make money or fully solved its technical challenges.

It has grown impressively over the last year as it has drawn younger users away from Facebook. It’s currently knocking at the door of Quantcast’s top 20 Web properties.

But, in his presentation at South by Southwest, founder and CEO David Karp sounded more like the editor of a literary magazine than the founder of a major tech startup. He lauded the site’s users but didn’t lay out any concrete plans for the growing platform.

The company has barely gotten its feet wet with promoted content offerings launched in May 2012.

Before it can monetize its millions of users, the company will have to provide a reliable platform, free of the period outages that have plagued it.

“We’ve had a few hiccups lately, but overall we’re pretty happy with where were at,” said the company’s vice-president of product, Derek Gottfrid, in an interview this week.

In 2011, the company averaged 3.8 hours of downtime a month, according to Pingdom. In 2012, averaged 15.6, but the individual test blog Pingdom watches averaged just 2.4.

Gottfrid said the company went out of its way to make sure the individual blogs it hosts would experience minimal downtime.

“This is really people’s blog presence. We’re putting a lot of effort to make sure that part stays up,” he said.

Big domain hosts like tech companies can host different parts of their content on different servers and also usually have redundancy in their saved data.

Many companies, notably Twitter, struggle to support all of their traffic and data as they grow. But pure capacity has not been to blame for Tumblr’s reliability woes.

“It’s more about an ability to manage complexity — the number of machines, the network configuration, the amount of data we’re moving around,” he said.

“In the abstract it’s pretty simple, but in the detailed level, it’s nontrivial,” he said.

As more users in more countries connect using more devices and customize their blogs, Tumblr will have more, not less, complexity to face.

But, Gottfrid said, the company has set up appropriate processes to handle the complexity.

Meanwhile, Tumblr is conducting initial studies to measure the benefits to brands of promoting content on the site. It hopes those results will lure more advertisers. But to be able to show concrete benefits to advertisers, the platform will have to prove it can serve up content even in the face of network glitches.