In its recently released 2021 State of DevOps Report, Puppet looks at the historical evolution of DevOps and highlights the importance of automation, cloud, and people and culture to its success. While it found that “many of the teams that are ‘doing DevOps’ well don’t even talk about DevOps anymore—it’s simply how they work,” there are other organizations mired at the halfway point.
In a follow-on webinar, “The State of DevOps: Stuck in the Middle,” BMC Product Manager Jon Stevens-Hall spoke with Puppet CTO Nigel Kersten about the report and how organizations need to get a better handle on technology and people in order to keep moving forward.
First, they talked about how size matters when it comes to DevOps. “DevOps in large enterprises is very different from the DevOps in smaller companies and startups, and that seems to be a big theme of the state of DevOps report,” says Stevens-Hall.
“The people I was meeting [at conferences] back in 2015 were smallish companies and the biggest and most prominent company…was Spotify…who…have been a very big pioneer in terms of team organization and things. Just four years later…it was banks, it was telcos They felt they’d done agile development, but the ops side hadn’t really emerged as agile.”
“The most interesting thing for me was that [they said] legacy isn’t going away…overnight. [The bank said] a lot of their transactions [might] involve the newest technology, but they also involve a lot of legacy products that can’t, or won’t be changed for years…they need new work practices, new technologies to [meet] that challenge of allowing DevOps to thrive amongst all this governance and complexity.”
Kersten adds that “legacy” is also a variable term. “When customers start to talk about the legacy environment, it’s not all the same. It is not actually the oldest stuff they have. Often there’s these old mainframe systems, which are really well run by small teams that operate in a very, very DevOps mindset just without sharing those practices across the rest of the organization,” he says.
“Legacy often tends to refer to those half-baked applications built in the nineties and two thousands [by people] who are all self-taught [so] there wasn’t a lot of discipline. Everyone was working this out as they went along. I think we have this whole generation of applications that are much harder to monetize than many of the things on the mainframes…And I think there’s many categories of what kinds of problems people have within legacy.”
In addressing the report’s categorization methodology, Kersten explained, “We introduced this idea of low, medium, and high evolutionary groups…deployment frequency at the low level [occurs] monthly [to every] three months, six months to get a deployment from done to actually shipping value to an end user, medium is between daily and weekly, and a high group is on demand.”
“Deployment frequency [can] be a vanity metric if people focus on it too much [but] there is something transformative for the business moving from monthly to every couple of days, it’s a really big jump. [When] we look at the data from the survey itself…there’s relatively low-hanging fruit in most organizations to move from low to medium. And then there’s a relatively concrete set of tasks to move from medium to high. The low group is getting smaller. The high group is getting larger, but there is this massive group of nearly 80 percent of folks who aren’t getting out of the middle area. And so that’s what we really wanted to focus on this year.”
“[None] of these [challenges] are technical problems…the things inhibiting large organizations [are] the people interactions and the team interactions and communication and all of [these] socio-technical systems. And when we make changes to the people side, it has an influence on the technology and vice versa…and it’s really inhibited…a lot of progress for people.”
Stevens-Hall points out that the current complexity faced by many organizations adds to that challenge. “[To] get people out of that stuck-in-the middle state…the technology department in the enterprise can [ensure] that we work with that tool chain and the owners of that tool chain. [We] can also [bring] this into the service management system [and] people from the service organization can be working with people from the DevOps team and…they can be sharing information, creating knowledge,” he says.
Kersten has tips for management, too. “From the…operations perspective…if you’re in management, the most important thing you can do is make sure your teams have an obvious purpose and identity [and] have well-defined interactions between them,” he explains.
“It’s all about collaborating across functional boundaries, but it turns out collaboration is actually a very time-intensive and inefficient way to operate at scale. And if you are a relatively small service management team [or] operations team with hundreds or thousands of developers there, you can’t possibly collaborate with them all.”
“The more evolved organizations are the ones that share practices and create spaces for it. As long as you don’t call it a center of excellence, which tends to be the death knell for actual progress for most environments. [Call] it something else [like a] community of practice…something a little less hierarchical and autonomous and authoritarian. [Create] spaces where people can just listen and learn and share. I think people are genuinely pretty smart and generally pretty engaged. And if you give them the space to find out about solutions, they’ll implement them.”
To learn more, download The 2021 State of DevOps Report and for more of this conversation, watch the entire webinar on demand.
