Velocity 09: John Allspaw, "10+ Deploys Per Day: Dev and Ops Cooperation at Flickr"
“In the last week there were 67 deploys of 496 changes by 18 people” – Flickr Dev Blog, December 17th 2008.
For deploying new code, planning projects, and crisis management, communications and cooperation between development and operations is paramount. As web infrastructures grow, the line between systems and software become quite blurred. Operations and development are disciplines that historically have been limited to a predictable list of responsibilities and have suffered from a culture of finger-pointyness.
We’ll talk about why and how these roles have mixed in Flickr’s environment for the better.
We’ll discuss tools, techniques, culture, and the processes of transparency that enables a healthy respect for the Big Red Deploy Button™ as well as incident management, escalation, and troubleshooting.Tools, techniques, and culture:
- Overview of our one-button code deployment, application-level logging, and dev/ops communications tools.
- Examples of adaptive monitoring and metrics collection, capacity awareness and planning must-haves.
- Rethinking the ops/developer relationship and mutual respect for responsibilities and expertise.
John has worked in systems operations for over ten years in biotech, government and online media. He started out tuning parallel clusters running vehicle crash simulations for the U.S. government, and then moved on to the Internet in 1997. He built the backing infrastructures at Salon.com, InfoWorld.com, Friendster.com and Flickr.com, where he currently manages the Operations Engineering group.
He is the author of “The Art of Capacity Planning: Scaling Web Resources”
Paul Hammond is a web developer, product manager and father. He has been building websites for as long as he can remember, and now leads a team of hard-working supernerds at Flickr. Before then he was part of the Yahoo Brickhouse team, and previously led technical project management at BBC Radio and Music interactive.