In this podcast Manuel Pais, InfoQ Lead Editor for DevOps, talks to Gene Kim, co-author of the "DevOps Handbook" and "The Phoenix Project" books, on how to scale DevOps in large organizations ("horses") and the need for continuous learning and adaptation.
Key Takeaways
- DevOps Enterprise Summit focuses on continuously learning from leaders elevating technical practices and cultural norms at large, complex organizations.
- Remaining challenges in large organizations include largely outsourced IT workforces, rigid change management processes, and powerful regulatory offices.
- Effective DevOps transformations require a certain political savvyness to protect teams and prevent initiatives getting killed too early.
- It's not the org chart that dictates outcomes, it's how people act and react. Reorganization is far less important than setting cultural norms and expectations.
- Metrics that matter in high performing organizations are code deployment lead time, deployment frequency, change success rate, or mean time to repair.
Subscribe on:
Show Notes
DevOps Enterprise Summit London Goals, Expectations and Take Aways
- 01:49 - Large complex organizations, or "horses", are achieving similar outcomes to Google, Amazon or Facebook (the "unicorns").
- 02:58 - DevOps Enterprise Summit is about learning from courageous leaders elevating technical practices and cultural norms at large organizations.
- 03:44 - After the summit learnings get codified in the books or in some way that can help others improve as well.
- 04:50 - Presenters at the summit are pioneers of practices that we will become the norm in 10 years from now.
- 08:09 - Summit has been following leaders and their DevOps journeys for the last four years.
- 09:10 - Main goal is to see what worked and what didn't as they progress in their DevOps journey, making that knowledge available to the rest of us.
- 09:25 - Another goal is to continually build case studies that can serve as selling points for enterprise practitioners in their organizations.
- 09:48 - This year's London edition will showcase global banks like ING, insurances like Allianz Deutchland, and IT services companies like IBM, covering topics like mainframe moving from annual to monthly releases.
- 10:40 - These are organizations we don't typically think of as leaders in building world class IT organizations.
- 11:40 - Creating platforms helps keep developers productive.
- 14:40 - One of the recurrent take aways from the conference is people realizing they are not alone in their struggles, and that it's worthwhile pursuing.
DevOps Handbook Purpose
- 03:21 - Most case studies in the book were from the DevOps Enterprise Summit to show evidence of real achievements.
- 06:18 - Book targeted enterprise practitioners in order to help them navigate complex organizations towards DevOps practices.
- 07:41 - Leaders in large organizations are being promoted as they achieve major gains with DevOps.
- 20:41 - The book took five and a half years of co-authoring before being published.
Scaling DevOps Challenges, Tips and Examples
- 12:45 - There are still unique challenges in large organizations, such as largely outsourced IT workforces, rigid change management processes, or powerful regulatory offices.
- 15:30 - Capital One was able to change their practices in the entire technology department, with 6k to 8k in-house software engineers.
- 16:23 - Difficulty of scaling resides in the large surface area of people you must influence for it to work: developers, QA, ops, security, risk management, product owners, etc.
- 16:40 - DevOps approach requires adopting many practices which are new to large organizations, such as funding products, single piece flow or extensive automated testing.
- 17:10 - Capital One's approach mandated faster time-to-market for every business unit. They identified the two top impediments to that being the number of manual approval steps required to deploy changes to production, and long lived feature branches, and attacked those.
- 18:15 - We will see more and more organizations making large scale changes that require strong technical leaders.
- 18:25 - Simply asking people to explain their practices to others triggered substantial changes in the branching/merging/deploying approaches at Capital One.
- 21:05 - Operations work can become inhumane, we need DevOps to prevent that by promoting shared production responsibilities with developers and frequent deployments during the day.
- 21:50 - DevOps helps developers become autonomous in testing, deployment, and fixing issues in production.
- 22:00 - ITIL and security compliance made it difficult to achieve autonomous teams. I had a part to play in that as well.
- 23:13 - Find DevOps business cases from the summit that are closer to your own situation (industry, challenges, etc) and use those to create a coalition of like-minded early adopters in your organization.
- 25:00 - One of the obstacles is convincing leadership to try a different way. Sometimes need to take the risk of making changes even before full commitment.
- 25:35 - Fostering transformations without activating the corporate immune system requires a certain political sofistication and savvyness to protect teams and prevent initiatives getting killed too early.
- 30:30 - Two schools of thought on the need for cross-functional teams in DevOps. One says yes because DevOps is fundamentally a reorganization. Other says it's a matter of leadership roles and team behaviors.
- 31:55 - It's not the org chart that dictates outcomes, it's how people act and react. For example, Toyota had centralized functional oriented teams but members worked closely with and behaved as if they were cross-functional members of the product teams.
- 33:10 - Reorganization is far less important than setting cultural norms and expectations.
State of DevOps Report Findings and Relevant Metrics
- 22:25 - Report results from 25000 respondents shows organizations aligned with DevOps goals are twice as likely to exceed expected profitability and productivity.
- 22:47 - DevOps can also help improve public services such as the UK's HMRC (revenue collection) service.
- 27:25 - Metrics that matter are code deployment lead time, deployment frequency, change success rate, mean time to repair.
- 28:10 - Employees in high performing organizations are twice as likely to recommend their workplace (employeed net promoter score) to others.
- 28:33 - More informal metric of value creation is how many people are saying thank you.
People mentioned
- Jez Humble
- Patrick Debois
- John Willis
- Jonathan Smart
- Robert Scherrer
- Daniele Romano
- Jason Cox
- Andrea Hirzle-Yager
- Tom Clark
- Etelvina Melo Pires
- Bruno Meniere
- Tapabrata Pal
- Scott Potter
- Adrian Cockroft
- Ron van Kemenade
- Mike Rother
- Steve Spear
Companies mentioned
- Amazon
- Barclays
- ING
- Ingenico
- Disney
- Allianz Deutchland
- IBM
- Compuware
- Ericsson
- Orange
- Worldpay
- Capital One
- ITV
- News UK
- HMRC
- Toyota
Conferences mentioned
Books mentioned
