DevOps platforms seen as key to enterprise evolution

DevOps platforms get teams unstuck

Many enterprises have reached what the report described as a middle phase of executing a DevOps roadmap: one in which organizational issues inhibit the ultimate goal of fast application deployments and efficient feedback loops that create continuous improvement. For teams at the high end of this middle phase, these blockers are entirely nontechnical, such as a failure to create sufficient feedback mechanisms or unclear DevOps roles and responsibilities, according to the report.

That's where DevOps platforms come in, the report authors suggest. Teams of platform engineers build such digital platforms for enterprises -- here, the report cites a 2018 definition of the term digital platform by Evan Bottcher, head of engineering at global software consultancy ThoughtWorks: "A foundation of self-service APIs, tools, services, knowledge and support which are arranged as a compelling internal product."

Some enterprises refer to these internal digital platforms as "the paved road" for developer teams -- not mandatory to use, but the easiest way to push code to production. Platform engineers, sometimes called site reliability engineers, are responsible for maintaining DevOps platforms as an internal product.

Achieving a product mentality and a mindset focused on providing services to customers is also an important part of building a successful DevOps platform, the report's authors said in a virtual roundtable discussion last week.

"You can give developers a set of things to assemble so that they can get on to the differentiation of what matters," said Michael Stahnke, vice president of platform at CI/CD vendor CircleCI. "We're basically making products for developers to move faster."

DevOps platform engineers as product owners

However, enterprises must avoid falling into counterproductive patterns of requiring that developers must use centralized DevOps platforms at all times or letting an enterprise architecture team removed from daily IT work dictate product decisions without flexibility. DevOps platform engineers must actually market their platform products to developers, and measure their products' quality using metrics such as net promoter score rather than traditional infrastructure performance data.

Few DevOps platform engineers have experience being product managers, the panelists acknowledged.

My generation of infrastructure engineers is dog-paddling catch up, because in order to be a modern software team you have to treat it like a product.

"My generation of infrastructure engineers is dog-paddling to catch up, because in order to be a modern software team you have to treat it like a product," said Charity Majors, co-founder at observability vendor Honeycomb.io.

This aspect of the report resonated with Andy Domeier, senior director of technology operations at SPS Commerce, a Minneapolis-based communications network for supply chain and logistics businesses.

About 18 months ago, the company created a technical product owner (TPO) role specifically to instill this product mentality, and to act as a kind of technical account manager for DevOps teams after the company struggled with some aspects of communication about the digital platform.

Initially, developers had to contact specific DevOps platform engineers to report feedback, depending on the infrastructure component with which they were dealing. That proved a blocker to communication between developer and platform teams, since components such as API gateways can span multiple IT disciplines.

"We have a TPO that can be a single point of contact, so we've insulated the development team from needing to know who to send their feedback to," Domeier said.

SPS filled this role from within its developer ranks, where a former lead developer who wanted to get into a management role now oversees a small team of TPOs. This team directs product lifecycle management for both customer-facing and internal-only services.

SPS's experience highlights another key component of escaping the DevOps plateau cited by the State of DevOps report -- upper management buy-in.

"That's validated in the idea that they were willing to fund this role at all," Domeier said. "Before that, the managers of the different teams responsible for the platform were doing that job ... a lot of it boils down to [prioritizing] developer experience."

From platforms to value streams

The State of DevOps report found that most of the teams it describes as highly evolved -- those that ship code fastest and most frequently, with the tightest feedback loops for developers -- make the heaviest use of internal platforms for their teams for functions such as user and service-to-service authentication and container orchestration.

The report does not cite Capital One as an example, but the financial services company is widely considered a highly evolved technical organization. This year, the company closed its last on-premises data center, and now bases all of its IT operations in the cloud.

Mindy Ferguson

The company maintains a platform of trusted APIs for partners called Capital One DevExchange, but overall, has moved past centralizing specific IT functions into thinking in terms of product domains, said Mindy Ferguson, managing vice president of technology at Capital One.

"We're about as full-stack as it gets," Ferguson said. "But we define platform a little bit differently."

Instead of separating DevOps platform and software development teams, Capital One's engineers work on all aspects of applications and infrastructure, aligned around business challenges such as fighting fraud, rather than technical areas of expertise, Ferguson said.

This can be a hallmark of a highly evolved organization, according to the State of DevOps report.

"Highly evolved teams tend to do a good job of limiting extraneous cognitive load on delivery teams (through good practices, automation and support from other teams), leaving more capacity to focus on the business needs," the report said.

At the highly evolved stage, according to the report, major challenges include dealing with legacy applications and maintaining team members' skills as technology evolves. While Capital One has dealt with the first challenge through its cloud migration, adding more skilled technology employees is among the company's top goals for 2021.

Capital One plans to hire 3,000 more technical employees before the year is out to bolster artificial intelligence and machine learning efforts as well as expand its cloud engineering teams, Ferguson said.

Despite industry talk of an IT skills shortage in recent years, Ferguson is confident that Capital One can attract qualified candidates. The company has multiple technical training programs for new employees, as well as an ongoing education program for all staff called Tech College. It also has partnerships with colleges, universities and industry associations such as Women Who Code to help it meet diversity and inclusion goals.

However, the nature of work itself is in flux following a year of pandemic-related upheavals, which will be an additional variable for Capital One to factor in as it expands its hybrid workforce, Ferguson said.

"We do have an audacious goal," Ferguson said. "The challenge is less about the number of people and more about how we continue to do the work that we're doing today at scale ... opening up our culture, helping people see our mission, during a little bit of a different time in this hybrid work situation."

Beth Pariseau, senior news writer at TechTarget, is an award-winning veteran of IT journalism. She can be reached at [email protected] or on Twitter @PariseauTT.