2:30 pm ET|
Oct 26, 2012
Netflix is all in for Big Data–using it to optimize the quality and stability of its video streams, and to assess customer entertainment preferences, so it can do a better job of targeting its users with offers for shows they might like to see. Those efforts are critical for the company, which made a successful transition from renting DVDs to providing digital video over the last few years, but now faces more competition and pressure on its business.
The company turned to Big Data following an outage in 2008 that left some customers without service for three days, said Kurt Brown, Netflix’s director of data science and engineering, during a presentation at the Strata Conference in New York. It’s an example of how the tools of the CIO and the IT department are growing in strategic importance.
The event was less catastrophic than it would be today, but it was severe enough to expose the company’s vulnerabilities, and spur it to action. Far fewer customers were using streaming at the time and Netflix was processing a smaller amount of information on its managed data center. “Most customers already had their DVDs for the next three days,” Brown said. “But we knew that as we started doing streaming and going international this was going to be a problem,” Brown said.
As the company prepared to offer more streaming, Netflix moved its storage from internal data centers to Amazon’s cloud. The main advantage of the cloud is that its architecture—which includes the highly scalable open source data processing platform known as Hadoop—allows the company to quickly provision computing resources as its needs grow.
That responsiveness is critical for a company that is under intense pressure. Netflix announced Tuesday that net income for the third quarter dropped 88% to $7.68 million. And subscriber growth isn’t up to expectations, either. Netflix added just 1.2 million U.S. digital subscribers last quarter, for a total of 25.1 million. The company has cut its domestic forecast for new streaming subscriptions to a range of 4.7 million to 5.4 million, down from an original forecast of 7 million.
Hadoop processing power allows the company to run massive data analyses, such as graphing traffic patterns for every type of device across multiple markets.
That effort helps Netflix improve the reliability of video feeds on different platforms and plan for future growth of streaming movies and shows. For example, the greater processing capabilities can allow engineers to see where traffic on the network is running slower, allowing them to plan for additional network capacity. The technology—which can manipulate larger data sets– also helps Netflix to better analyze customer preferences so that it can make improved recommendations.
“It scales beautifully and it’s superfast,” Brown said.