Harness’s founder Jyoti Bansal learned some lessons from selling his previous company, App Dynamics, to Cisco for $3.7 billion in 2017 – Cisco had swooped in with a strong offer, just days before a long-planned IPO.
One of them is, perhaps, despite the large amount of money: Don’t do that.
Bansal has previously said he regrets the decision to exit and to feeling “a bit lost” after the sale; App Dynamics could have grown a lot more, he reflects; and the itch to take a startup to IPO remains strong.
He learned a tonne of lessons about scaling fast along the way however – and after swiftly deciding that retirement wasn’t for him, returned afresh.
His current company, Harness, founded in 2017, provides a multifaceted platform for DevOps teams spanning a code repository, Continuous Integration (CI), Continuous Development (CD) and feature flag capabilities, through to cloud cost management, integrations with OpenTofu and Terraform for Infrastructure as Code management, more.
“Everything is designed to be modular”, he explains. “You don’t need to boil the ocean. So use [us] like Lego blocks. Pick which block you want…”
Harness has already grown to what he tells The Stack loosely is “a few hundred million dollars of revenue”, with Citibank and United Airlines some of the impressive logos among its 1,000+ enterprise customers.
Like many fast-growing SaaS firms it’s not profitable yet, “because we are investing so much back into growth…but we'll be profitable as we build towards being a public company at some point” he tells us at re:Invent.
Replacing Jenkins and developer "toil"
Bansal built Harness because in his view, there simply wasn’t a unified platform for developers that could “reduce their toil”; instead most organisations are relying on a loosely knit plethora of third-party platforms and homegrown Jenkins or equivalent pipelines, he says; toggling between disjointed toolchains and getting lumbered with rampant “context switching” or random task-changing. As a result, DIY environments are a large part of what Harness is displacing, he says.
“Historically, there were not really good platforms to do a software delivery pipeline properly. So people build homegrown systems. People build homegrown systems on top of Jenkins; people build homegrown systems on top of, you know, something like a GitHub Actions, which is a CI, but then you have to do the deployment automation, security governance, [and you] build your homegrown scripting on top of that.
“Typically people have glued things together; [there’s] hundreds of engineers toiling at maintaining and updating these kinds of things [homegrown CI/CD and associated environments]… in most companies, developers are spending [only] about 30% of time coding.”
“Let's say you wrote some code, and now you have to deploy the code to production. Are you spending four hours watching a deployment? Are you running manual steps that could take another two hours? Or something fails. Does it take two hours of trouble-shooting to fix it?
“You have your compliance teams [who] want an audit trail of all the changes you made. Does it take you one hour of manual work to produce that trail? Does it take one hour to build, or does it take 10 minutes?”
14 modules; 14 "startups"
Harness has built 14 specific modules (fill your boots with acronyms here); all aim to easily plugin other tools where appropriate. Bansal says these are run as “startups within a startup; like 14 independent startups focused on building the best of breed solution for their problem area…”
A Security Testing Orchestration (STO) module for example “consolidates, prioritizes, and deduplicates findings” from various scanners (open-source or otherwise) whilst a chaos engineering module lets you stress-test application performance under a range of scenarios. Want to do canary deployments,or feature-flag management where you define the characteristics of the users who will experience a feature first? Bansal says Harness makes this easy; ditto the ability to do progressive delivery with automated rollbacks – hardly cutting edge, but still not widely done.
See also: Fidelity reveals "chaos buffet" approach, says 7,700 apps now on public cloud
In many organisations developer teams take a pull-based GitOps approach, where Git is the source of truth for what should be deployed, developers update files to a desired version, merge the change through a pull-request and an open-source “reconciler” like ArgoCD or Flux brings the environment into compliance with the new desired state.
Harness holds that to “bridge the gap between good open-source reconcilers and full continuous delivery, teams need to trigger further automation pipelines”; it adds a management layer over such tools, “making it easy to install and maintain reconcilers and centralising reporting so it’s easy to see what is where without digging through multiple Git repositories… [and] layering in all the testing and security scans needed.” That, in a nutshell, captures the approach – bringing together a wealth of open-source and other tools but making them easier to use together in a joined up fashion and with more features baked in.
Harness has its own open-source proposition too. That builds on the 2020 acquisition of “Drone”, a popular container-native CI platform built by Brad Rydzewski that Harness effectively built its CI module on top of.
First dubbed “Gitness” it has since rebranded this simply “Harness Open Source” – Bansal says Harness still contributes upstream to Drone and has subsequently made elements of multiple Harness modules open-source.
“Last year we decided to take the Drone project, which is CI, but also create a bigger open source project, ‘Harness open source’... we bring CI, CD, code repo and a few other DevOps capabilities as like one open source download you can get started [with]” Bansal tells The Stack.
"We are very deliberate about OSS"
Pressed on the dubious track record of VC-backed companies pulling a freemium strategy open-source bait-and-switch, he insists that Harness’s approach to OSS is well thought out: “We are very deliberate about what is open-source and what is not. Like if we make something open source, we are not going to take it back… we are [also] making choices where we're not making everything open source; some things are in a source-available kind of license, and some things are in an open-source license. We want to be very clear with our intention. We don't want to surprise the community who are using or using or relying on us” he says.
Customers can get Harness as a SaaS service “where it's everything in the cloud… but we also have a secure architecture where you have these worker processes called ‘delegates’ that you run in your own environment. So our SaaS will do the control plane, and then the worker process will do the groundwork. So it's very secure from that sense.”
Around 10% of Harness’s customers just want to deploy fully on-premises and it supports that, he says. (“Mostly government agencies, etc, where, you know, they're not comfortable with the cloud.”) Everything is containerised so “you just bring in a Kubernetes cluster” he explains.
Wrapping up, Bansal says “we have an excellent list of customers. We have gotten efficiencies of 20, 30, 40% on developer time; customers where their deployment times have come down significantly; companies that were shipping every two weeks, they can ship every day. Companies that are shipping every three months, they can ship every week. [Their] quality has increased, costs have come down. So it's very fulfilling.”
