Interview: Developers spend too much time ‘not coding’ says Harness CEO – DEVCLASS

Harness CEO and co-founder Jyoti Bansal told DevClass that “developers spend 60 to 70 percent of their time not coding” because of all the things “that happen after coding, like testing, deployment, security, governance, compliance” – to the detriment of the developer experience.

According to Bansal, companies “want to modernize how they do software engineering … moving to the cloud, modernize infrastructure, rearchitect application to be agile and nimble with microservices, Kubernetes, containers” but cannot achieve it because of poor developer experience resulting in low productivity.

A Harness-sponsored developer experience survey concluded that 60 percent of organizations release code on a monthly or quarterly basis, rather than the frequent updates that are associated with an effective DevOps system. One of the reasons is that code reviews are time-consuming: 67 percent of devs, the survey reported, take more than a week to complete them. Other issues include manual rollbacks in the event of a failed deployment (again 67 percent); insufficient unit test coverage, according to 32 percent of developers surveyed; and security requirements, cited by 59 percent as a reason for delay in deploying.

Jyoti Bansal co-founded Harness in 2017, to market a platform for software developers, the implication being that the right tooling will improve developer productivity. It is common to hear though that the biggest barrier to productivity is organization culture rather than tools. Gene Kim, for example, told DevClass that it is leaders within an organization that set “conditions for everyone to do the work easily and well.”

Bansal disagrees. “My point of view is that tools are the driver to change the culture … you can ask your teams to ship once a day … it’s really impossible to make the change culturally until you show them the path, which actually happens with the helper tools. Organizations who drive the culture change, what they do is they bring a new toolchain, a new set of processes, into some part of the organization. If you have 5000 developers you can’t change it for all of them right away. So you start with some teams and some applications, and these are the 100 or 300 developers, they use the new set of tooling, new set of processes, to make that successful and showcase it as the driver of culture change for the other 4500 engineers.”

Why are so many organizations still locked into slow software delivery processes, when the value of continuous delivery (CD) has been known for so many years? “The tools haven’t kept up,” Bansal told us. “80 percent of industries still use Jenkins and that was built about 20 years ago now. What people did was, took Jenkins and bolted a lot of scripts on top of it to home-grow CD … the industry has out-grown Jenkins and home-grown tooling and scripting, and almost everyone is struggling because of that.”

Bansal’s comments are in the context of Harness, which provides what is intended to be a complete software delivery platform, so this is a sales pitch, but one that has some resonance. Now though we have hyperscale platforms like AWS, or GitHub and Azure, or Google Cloud Platform, all of which offer their own CI/CD tooling. Does that make it difficult for Harness to compete?

“The hyperscalers always have some tool chain for everything,” said Bansal. “Harness products are significantly superior to what the hyperscalers can provide,” he claimed. “We are heavily focused on it,” he said. “If you have a small app or a small environment you can probably use those but we normally don’t see them as competition.

“The other element is that most of the industry are also looking to create their engineering practices and tool chain to be multi-cloud or cloud-agnostic,” he said.

AI also has a role, Bansal said. “A deployment can fail for so many reasons … a developer has to look in the logs and figure out what happened. We can easily help synthesize it all and in simple English say this is what failed and how you fix it.”

AI can make mistakes though – in code generation as well as elsewhere. The consequence is that time saved by AI has to be invested in “the next generation of checks and balances,” said Bansal. “The role of QA may become even more important than the role of developer, in the world of AI. AI can write the code but the human has to validate and verify that it works.”

How does open source fit into what Harness does? “Our core CI product is an open source project called Drone. Last year we launched a code repo called Gitness, which is the first major code repo launch after GitHub and GitLab and Bitbucket. It’s Apache licensed. And we are one of the biggest sponsors of the OpenTofu project which is a fork from Terraform. And our team runs the project called Litmus Chaos which is a CNCF project … we integrate very well with the open source tool chain, it’s a plug and play kind of model and our main philosophy is that we have to co-exist, we cannot go with the walled garden approach.”