My Interview with Adrian Cockcroft – Netflix Cloud Pioneer

Adrian-Cockroft-bigOn Wednesday July 2nd, Adrian Cockcroft from Battery Venture is going to lecture on an IGTCloud and IamOnDemand meetup. Prior to joining Battery, Adrian  lead Netflix’s migration to a large scale, highly available public-cloud architecture and the open sourcing of the cloud-native NetflixOSS platform – the Netflix cloud management platform.

The Netflix Cloud

Netflix, the popular American provider of internet video streaming, dominates the market with over 50 million subscribers and a third of all internet traffic during peak hours. Netflix currently operates in over 40 countries and has announced its expansion to 6 more European countries. Netflix is not only the leader in internet streaming services but an example of the largest pure cloud service based on an open-source stack running on Amazon Web Services. Netflix is a widely referenced case study for how to effectively operate a cloud application at scale.
Following his arrival to Israel, I had the pleasure of interviewing the Netflix cloud pioneer.

1. How much has the cloud progressed since you became familiar with Amazon AWS cloud until today? In which aspects/areas?

I started on AWS in 2009. At that time it was missing features, support mechanisms and was overall a rather immature platform suitable for small startups. Netflix worked with AWS over the years to develop the features, get an enterprise support model in place, and to shake out a lot of bugs and single points of failure. Now, in 2014 AWS is getting a lot of enterprise adoption and is the clear leader in cloud.

2. Currently, at Netflix do they look for “Devops” or for developers? Does the NoOps perception really exist?

DevOps is a re-org, not a team. You get to DevOps by merging the development and operations processes into a single team that runs both. There are two flavors of DevOps. In many cases, particularly when there are data centers to manage, DevOps migration is driven by the operations team. They speed up and automate the operations processes and build common tools for developers and operations. In the case of Netflix, the cloud migration moved part of the operations responsibility to AWS and a developer driven migration to DevOps occurred. Many operations roles and processes were eliminated and cloud native tools were built. At one point this looked very different and was called NoOps, but a better name is developer driven DevOps. Netflix is a very developer oriented company. They teach their developers to operate their own code using continuous delivery tooling and automation.

3. How do you see enterprise IT in 10 years from now?

In ten years, only the very biggest organizations will run their own infrastructure. Everyone else will use a public cloud from a handful of huge generic low cost vendors like AWS, Azure, and GCE. Other cloud vendors and many of the current enterprise vendors will have died out. Most business functions will be run as SaaS hosted on public cloud, and most applications will be portable across the three big public clouds.

4. Israel, “The Start-up Nation” has lots of new technology companies. Especially ones related to the cloud. – Could you give 3 tips for the modern scalable R&D manager?

  1. Optimize for speed of development by removing handoffs from your processes and automating what’s left.
  2. Make sure that developers are on call for their code in production. They write much better code if they are directly responsible for how it behaves.
  3. There is a huge amount of open source code coming from web scale companies that is better than commercial products. There is a strong tendency to reinvent the wheel, but it’s far better to leverage existing code. Take a good look at the forty or more Netflix open source projects on Github, and keep track of new releases.

5. In your presentation, you are going to discuss the notions of “Cloud Native” and “MicroService” Could you give a few brief words about each?

Cloud Native architectures are dynamic, scalable and highly available and are built of ephemeral components that can be updated or fail and be replaced without impacting customers. To continuously deploy new code, bounded contexts are created using groups of single function loosely coupled “micro” services that can be updated independently and safely, many times a day.
To register for the meetup that will also include practical Devops sessions dealing with Speed and Scale brought to you by Redislabs, Ravello and Emind:
Cross posted on

Related posts