As the founder and CEO of software maker AppDynamics, Jyoti Bansal had heard his share of horror stories from customers, but a tale relayed by a large bank CIO a couple years ago caught his interest. The CIO told him he had 600 software engineers spending most of their time writing automation scripts to enable DevOps, an increasingly popular methodology for building software in the digital age.
Moreover, the CIO estimated that he would have to hire another 2,000 engineers to achieve the continuous integration and continuous delivery (CICD) processes for which DevOps is prized. “Everyone talks about DevOps and automation, but every DevOps team is writing similar scripts to verify, deploy or roll back something,” Bansal tells CIO.com. “It’s become its own cottage industry.”
The CIO’s dilemma nagged at Bansal, who shared the tale with his friend Rishi Singh. As it happened, Singh encountered the same challenge deploying and scaling DevOps in his platform engineering role at Apple. They agreed that engineers are spending an inordinate amount of time crafting scripts to automate the various steps required to build, test, deploy and refine software. “Almost everyone struggles to make it work,” Bansal says.
DevOps challenge presents an opportunity
For an entrepreneur like Bansal, the problem presented an opportunity. In February 2017, Bansal left AppDynamics and launched Harness.io, with Singh as his co-founder and CTO. The startup quickly built a continuous delivery-as-a-service (CDaaS) platformthat leverages artificial intelligence (AI), machine learning (ML) and other capabilities to help enterprises achieve CICD without paying a premium for more DevOps engineers who will spend far too much time writing scripts.
Using Harness.io’s templates and wizards native to Amazon Web Services’ cloud and Lambda services, as well as container technologies such as Docker and Kubernetes, DevOps engineers create a model of software they want to build and upload it into the platform. Developers can then watch their CD pipelines execute in real time across development, deployment and quality assurance, a far cry from the two months or longer it takes to write scripts to automate such processes. “You configure it and tell us what you want to achieve and we’ll create the automation,” Bansal says.
Harness.io audits every step in the CD process so that if something breaks, an engineer can learn about it immediately through continuous verification. In the event the platform detects an error, it automatically rolls a build back to its previous working version. Bansal says this is possible because Harness.io dynamically versions software artifacts and runtime configurations every time a change is made. “You can move fast without the fear of breaking things,” Bansal says.
This is particularly crucial for enterprises such as banks and insurance carriers, which can’t afford to deploy flawed software. In short, Harness.io aims to shrink both errors and application delivery time that progressive IT organizations prize as the Holy Grail of modern software development. “It doesn’t make sense for everyone to do this themselves,” Bansal says.
Automating script-writing for release processes and other aspects of the development chain remains a relatively green field for AI and ML tools, and one that Forrester Research analyst Diego Lo Giudice says has the potential to reduce manual labor associated with DevOps. However, Lo Guidice says the challenge is collecting enough data to make the platform smarter.
Pilot proves faster deployment
One way to gather more data is to test it using real customer workloads. Jobvite is among the approximately 20 companies beta testing Harness.io. John Stuart, Jobvite’s vice president of DevOps, security and IT, says that with only four full-time DevOps engineers the company struggled to write unique PERL and Python scripts to continuously integrate and deploy software.
When Stuart’s vice president of engineering, Daniel Lipkin, suggested he look at Harness.io, Stuart became intrigued and set up a meeting with the company. The Harness.io team told him it would enable CICD, including the ability to roll back software for bug fixes. Recognizing the time and effort his team could save using such a platform, Stuart began piloting Harness.io.
So far, Harness.io has helped the company halve software deployment times, which is no mean feat for a company executing hundreds of regular deployments. Moreover, Stuart reallocated a release engineer who was tracking code deployments to other duties. “We need engineers to do other work, and this accelerates our ability to deliver software more efficiently,” Stuart says.
If all goes to Bansal’s plan, Harness.io will help level the playing field for IT leaders who look enviously (and incredulously) at the likes of Facebook, Amazon.com and Netflix, whose DevOps engineers boast on conference panels about their thriving and thoroughly automated CICD pipelines. But those Internet companies employ thousands of engineers dedicated to software delivery. And, at roughly $200,000 a head, paying let alone finding that much programming talent is virtually impossible for even the largest enterprises, Bansal says.
“You can’t recruit them because they’re not available,” Bansal says. “There are only a handful of companies who can pull off continuous delivery where developers check the code and where it gets delivered to production in less than a day.”
Harness.io has banked $20 million in funding to date, including $15 million from Menlo Ventures and $5 million of Bansal’s BIG Labs, a startup studio from which Harness.io graduated. Bansal says Harness.io plans to officially launch its production platform in early 2018. In the meantime, Bansal says he’s working on adding new solutions and services.
These include self-service capabilities that will give DevOps engineers templates with which to build new microservices and deploy them in minutes. Another feature will assign customers spending quotas to help the challenge of managing cloud costs. Harness.io is also working on continuous security, in which developers will build software at their own pace while hewing to corporate compliance and risk requirements.
“We don’t want to slow anyone down, but build guard rails so people can operate within them and if they go out of bounds, Harness.io will stop them,” Bansal says.
Harness.io is providing a technical demo of its platform on Jan. 30.
Related DevOps articles: