'A whole new world': How Chef DevOps tools help JustGiving make new deployments in minutes
'Ultimately we're getting far more out the door,' JustGiving CIO Richard Atkinson tells Computing's Danny Palmer
The exact definition of DevOps can sometimes be hard to pin down, because it's one of those IT terms that there doesn't quite seem to be a consensus on.
But for Richard Atkinson, CIO at JustGiving, the online charity fund-raising platform that has over 15 million users and registers 50,000 donations a day, the definition is simple. "DevOps is building things so that we can operate them," he told Computing.
Before the deployment of DevOps tools and processes, the rollout of any new JustGiving service was "manual intensive", he said.
"It would always start with a developer writing code then once they were happy with that code they'd hand it over to QA who'd check it worked on a system-wide basis.
"Then they'd pass it on to some Ops people who'd have the job of getting it to run at scale," he explained, referring to the end product as a "poisoned chalice" for the person in operations who was tasked with making it work.
To get away from this labour-intensive process, JustGiving adopted the DevOps toolset provided by Chef, after it was recommended by some recently hired developers.
"As is often the case for us with a new initiative, it started with a software re-engineer," Atkinson said.
"We had a legacy that was one large application. We started by saying this isn't the way to build software, so we brought in a whole new architectural capability. We hired architects and they brought in knowledge and skills, and one of those was use of Chef," Atkinson explained.
"It wasn't a large procurement, we picked hires that we trusted who took the tool out of the box to see what they could do. Then after six months we reviewed it and it was working just great. We're happy with Chef," he said.
Before switching to a DevOps-based approach, deployments at JustGiving could be time consuming and a big drain on resources.
"It could take an hour or more to get the product out into the wild to be publicly accessible," he explained.
"And whilst that release is happening, you're actually suffering a degraded capacity, because you're taking servers out of the pool to upgrade them," Atkinson added. However, since deploying Chef, that's all changed.
"With Chef, we can deploy a service in minutes. We can run through the whole deployment pipeline. So any automated testing, any integration and build work, all of that is being run by Chef," he explained, describing how it means "whenever a developer has a feature to push, they just push it".
"We can do many deployments per day and each one takes minutes, if that, without suffering capacity degradation," he added.
What does this mean for JustGiving users? Ultimately, it allows Atkinson and his IT team to get updates and products deployed much more quickly and smoothly than previously possible.
"We can get product to market quicker. We suffer less disruption, whether that's degraded capacity or anything else you care to mention as things go into service. And I think it benefits our external stakeholders by proxy because our efficiency is at a consistent level," he said.
That, he told Computing, is preferable to the old stop-start way of doing things.
"When you have to stop every week to do a deployment, it creates a very spike-and-trough effort level across developers," Atkinson said.
"Whereas with developers pushing their own features through to deployment, there's a constant level of business that makes us develop code more effectively," he continued, adding "ultimately we're getting far more out the door".
Atkinson said it's a "whole new world" for developers now because they "have a lot more responsibility".
"It's not just about judging their own work on the features, but on much wider issues of scalability and operability," he said.
Computing's DevOps Review 2015 is available to download for free here.