Common ground
The origins of ITV’s Common Platform began in 2014 within its Online division, when a colleague of Clark’s, Rob Taylor, was tasked with creating a version of the broadcaster’s catch-up service to run on Samsung internet-connected TVs.
At the time, the company – and the MSP to which it outsources its infrastructure requirements – was tightly wedded to the waterfall method of project delivery. Taylor, however, decided to adopt a DevOps-style approach to creating a minimum viable version of the app.
“He seconded one of the developers and a platform engineer, put them together and told them they had control to make their own decisions, and he set about requesting some virtual machines [VMs] from the MSP for the project,” says Clark, which is where the firm’s agile ambitions ran into trouble.
After requesting the VMs, Taylor was told to expect a wait of about four weeks for them, which prompted him to provision his own from cloud giant Amazon Web Services (AWS) instead.
“When you are trying to develop in an agile way, a four-week lead time for a VM is not compatible really,” says Clark.
The success of this relatively small project gave Online the confidence that DevOps, as a software delivery approach, could thrive elsewhere at ITV, along with the use of cloud.
“From there, I was asked to take the minimum viable product that Rob built as inspiration to build the hosting platform for the whole modernisation programme for ITV,” says Clark.
“There are 16 or so key lines of business systems that take billions of revenue through them, that run on this DevOps, open source Common Platform, and it basically underpins the whole of ITV.
“It started with the simple, low-risk stuff and building that up until we get to where we are today.”
That includes having services considered to be “broadcast-critical” running on the Common Platform, including its scheduling system, which dictates what programmes are shown when and on which channel.
“If the broadcast scheduling system breaks, ITV goes off air,” he says. “Not immediately, but within a few hours – but that runs on the Common Platform in AWS.”
Although the organisation has reached a point where mission-critical workloads can run on the Common Platform, that is not to say the modernisation programme is complete, says Clark.
“We have systems that still run Cobol, for example, like the one we use to pay Ant and Dec. We are replacing those with a brand-new Java-based microservices stack running in the cloud. We need to get Cobol out of those systems because the people who write it are dying out – literally.”