Technology managers at video distributor Netflix Inc., which last year took a big plunge into the public cloud, have started a remarkably candid blog about their experiences with mission-critical cloud computing.
Netflix redeployed most of its customer-facing applications -- including its Web site, search, recommendations, video streaming and huge data stores -- to Amazon.com Inc.'s cloud service. It's been a "Herculean effort," wrote John Ciancutti, one of three Netflix vice presidents of engineering, in The Netflix Tech Blog (http://techblog.netflix.com).
One reason for moving to the cloud was the need to completely rethink IT strategy as the company's business model moved from mailing DVDs to streaming video over the Internet, he said. It made sense to have Amazon handle data center infrastructure so that Netflix engineers could focus on innovation and customer experience.
But Ciancutti acknowledged some big challenges. "There were some dark days as we struggled with the sheer size of the task we'd taken on, and some of the differences between how [Amazon's cloud] operates vs. our own data centers," he wrote.
For example, Netflix must cope with hardware failures and slow response times in the cloud, where IT resources are shared with other tenants. "Co-tenancy is hard," Ciancutti said. To address that problem, Netflix developed a software architecture that it calls "Rambo," in which each application can succeed even if related systems fail, he said.