How to grow a DevOps practice
What is the best approach for creating a culture and installing processes that orchestrate technology development and operations?
It's a trick question: There isn't one.
Many in industry are after a silver bullet for making technology practices more efficient and weave in a DevOps mindset, but none exists. And those experts who could build a five-step, how-to guide for DevOps know it will not take effect.
"Everyone wants a maturity model," Nicole Forsgren, DevOps Research and Assessment (DORA) lead researcher and research and strategy at Google Cloud, told CIO Dive.
"They want this easy, prescriptive path telling them exactly where it should go and what we've seen is that that doesn't work because no two organizations look alike, no two teams in an organization look alike," she said.
That means organizations have to develop and design a strategy particular to their business. To start, it requires a constraints-based approach which tackles a businesses' biggest problem.
When creating a strategy, business leaders have to identify their biggest problem and focus there first, Forsgren said. "When it's no longer your problem, do your next problem."
Drawing on six years of research and data from more than 31,000 professions, DORA analyzed what works — and what doesn't — when companies across sectors develop and deploy technology.
While no two strategies are the same, the research group found common threads of success.
- High performers scale DevOps by creating "community structures" at the management and tech worker level.
- Large-scale change comes from management, rather than a grassroots movement.
- Software "speed, stability and availability" dictate how an organization performs and boosts technology transformation.
- Software delivery performance and availability is more effective when companies tap into the cloud.
Enterprises, companies with 5,000 employees or more, have lower software delivery and operational (SDO) performance than smaller organizations, according to the 2019 DORA report, released Thursday. Larger organizations can struggle from legacy process, controls and infrastructure, which can hinder DevOps adoption.
DORA breaks down company SDO performance into four distinct categories: elite, high, medium and low performers.
This year, 20% are elite performers, up from 7% last year. The number of medium performers is growing, up from 37% to 44% this year, and the proportion of low performers is shrinking.
The gaps between the elite and low performers highlights the chasm companies have to cross when competing with each other. Elite performers are faster, deploying code 208 times more than low performers.
To put that in context, elite performers deploy on-demand and multiple times per day. Amazon, Google and Netflix deploy "thousands of times per day," according to DORA. Low performers deploy once a month or once every six months.
Elite performers are also more resilient, 2,604 times faster to recover from incidents than low performers.
Approaching its 10-year anniversary, the DevOps movement has evolved. Some even want to adapt the name to emphasize the importance of security, changing it to DevSecOps.
Regardless of the name, DevOps implementation sets the tone for adoption and determines long-term success.
"Only leaders can change your process," Forsgren said. "That's a bureaucracy thing."
Previously DevOps adoption relied on a grassroots movement, which built momentum until gaining the attention of the C-suite.
Now, middle management has emerged as a critically important layer of DevOps adoption.
"TheMiddle management layer is critically important in the DevOps evolution as they can translate both the business needs and the on the ground needs to make it happen across those teams," Alanna Brown, senior director of developer relations at Puppet and lead researcher of the State of DevOps report, told CIO Dive.
Executive support acts as touch points across teams and curbs the emergence of siloed knowledge and communication, she said.
After identifying who will allow DevOps to emerge, it's a question of scale. Highest performers build a community, which is more resilient to reorgs and product changes, DORA research found.
System administrators or developers cannot set into motion an enterprise's approach to process and development. They can, however, at the team level work to deploy automation, continuous integration and automate testing.
They can automate that and take advantage of efficiencies "where leaders and VPs should probably not be putting their hands on code," Forsgren said.
Together teams and the managerial level can use cloud services and execute disaster recovery testing, according to the report.
Article top image credit: Getty Images