How Pivotal's PaaS Transformed Comcast
Using microservices and the Pivotal platform-as-a-service (PaaS), Comcast managed to build a team of 1,500 developers and modernize its application development in roughly a year.
That, in turn, sets up the operator to apply those same software principles and people to customer-facing applications, said James Watters, senior vice president of products at Pivotal , speaking during a VMware media dinner during last week's Open Networking Summit. (Pivotal and VMware were both owned by EMC, which is now owned by Dell Technologies (Nasdaq: DELL).)
The story for now, though, is about Comcast Corp. (Nasdaq: CMCSA, CMCSK) being able to build up a cloud-native software development team rapidly. It's a common trend among large companies. Enterprises such as Target are transforming their IT departments to take advantage of open-source software and deliver services through the cloud, because they see themselves becoming quickly obsolete if they don't change.(See Target Looks to Open Source to Hit Bullseye.)
Banks provide some of the most aggressive examples. "Every bank is terrified they're going to become the Blockbuster of banking," said Bruce Davie, VMware's CTO for Asia/Pacific and Japan, during the dinner event.
According to Watters, Comcast started fiddling with Pivotal's PaaS in 2014 but didn't really start working with microservices until 2015. By 2016, the company was running 10,000 containers and had built 1,000 applications based on microservices.
The result was a marked improvement in network resiliency and, more directly relevant to Pivotal, a drop in the time it took to deliver new services. (Watters backed up his story with statistics that he asked us not to share. Much of the data came not from Comcast directly but from an accounting firm's audit of the operator.)
Comcast's back-end systems had been based on Oracle. Comcast was running 700 applications "all painted red," Watters said -- and they were contributing to Comcast's low net promoter score. "When you call Comcast and you can't get what you want, sometimes it was because the Oracle stack was down," Watters said.
Part of the idea behind microservices is to get infrastructure to "act in concert" to deliver what the application needs, Watters said. Until now, it's been the other way around: Applications have to be built to accommodate the limitations of the infrastructure.
A similar story applies to carrier-provided services, which is why service providers are so intrigued by network functions virtualization (NFV).
"IT has run on 'Deploy my VM and hope the network holds' for 30 years," Watters said. "Microservices are going to win, hands down."
— Craig Matsumoto, Editor-in-Chief, Light Reading