Netflix: We don’t need no stinkin’ data centers!

Updated: Netflix(s nflx), the streaming media giant, turned to Amazon (s amzn) for its cloud infrastructure because Amazon Web Services was — and remains — the biggest, baddest cloud around. That’s according to Adrian Cockroft, director of architecture of cloud systems for Netflix, who spoke on the topic at SVForum on Tuesday.
The AWS decision means that Netflix has not had to add data center capacity since 2008, according to Cockroft’s presentation on why Netflix built its platform as a service (PaaS) atop Amazon Web Services.
Netflix is a good example of a company that uses Amazon for what it does best — provide scale and power — but layers its own key technologies — in the form of a “thin” PaaS atop those Amazon services. However, Cockcroft’s presentation notes that Netflix would like its own PaaS to get thinner over time as Amazon moves up the stack.
The overriding reason for the Netflix move to a public cloud generally, and to AWS in particular, was that the company could not build data centers fast enough to meet the often-spiky demand of its users and that AWS was the only game in town in terms of the scale that Netflix needed.
One of the tweets from the SVForum event also piqued our curiosity because it indicates the pricing concessions — 33 percent off of  list — that Netflix, as a ginormous AWS customer, is able to negotiate.

(Update:  Adrian Cockroft wrote in to clarify: This  AWS pricing “is what I was talking about. We use the Heavy Usage 3 year reservations and on top of that AWS gives volume discounts. Compared to full hourly price and depending on what instance types are used, a large customer can save a lot. I was NOT talking about any specific AWS/Netflix discount scheme, just the publicly listed options.”)
Here are some key takeaways from Cockroft’s talk:
Why Netflix built its own PaaS atop AWS: Alternatives like VMware’s CloudFoundry are developer friendly but lack scale. RightScale, which many customers use as a management window into AWS, creates its own “lock-in” scenario, he said. Netflix didn’t want another vendor between it an AWS.
What that Netflix PaaS does:  Supports all AWS zones and regions; supports hundreds of developers on one project, monitors millions of metrics.
Core PaaS services used:  AWS S3 storage and Simple Queue Service; Netflix-based EVCache memcached ephemeral cache; Cassandra distributed data store. Some services like GeoIP lookup and Keystore HSM still run in the Netflix data center.
2012 roadmap: Work toward IPv6 support; add more automation and orchestration; prepare groundwork for global Netflix launches.
Why public versus private cloud?  Netflix needed the size and scale of a massive infrastructure and private clouds are “harder to build and support than you might think.”
Why the Amazon public cloud? AWS has more mature and less varied APIs, offers far more features faster, and much bigger scale than the “seven dwarf” options, Cockroft said. There may be a viable AWS alternative but not for “two or three years.”
Cockcroft also directly addressed the concern of many Amazon partners — that the platform provider is coming up the stack to compete with its PaaS customers. His response: The whole notion of “coopetition” between tech partners is as old as IT itself. In other words: Get over it. Netflix will be tied to AWS infrastructure for quite some time.
But these highlights are just a taste. Cloud tire kickers should check out the presentation itself as well as a longer version given at the qCon Conference in London earlier this month here here and here.