BOSTON – T-Mobile is currently operating 18,000 containers across its virtualized cloud infrastructure, mostly in support of customer facing applications. Those containers handle up to 10,000 transactions per second for services like payments and device leases.
Chuck Knostman, vice president of digital technology and strategy at T-Mobile, rolled out those stats during a keynote address at this week’s Cloud Foundry Summit in Boston. The venue was appropriate as the carrier’s container operations include significant support from Cloud Foundry.
Cloud Foundry’s annual event is where T-Mobile started its cloud-native push. During a white-board session in 2015 Pivotal focused on the benefits of going cloud native. Knostman admitted that at the time that concept did not register with the carrier. However, he soon realized that this was what the carrier needed to do to update its operations.
T-Mobile’s initial container work with Cloud Foundry started with middleware “to see if the platform could scale to our needs,” Knostman explained.
That initial work quickly escalated as the carrier began deploying containers in support of service applications. The carrier had 12,000 containers in pre-production and production by the end of 2017. Of the 18,000 containers in operation today, around 8,000 containers support production applications.
That effort is also supported by approximately 1,700 developers, with the carrier pushing around 30 Cloud Foundry updates per day.
Knostman noted that like many large enterprises moving toward cloud native, the biggest challenge has been getting people on board with the new direction.
“You get used to doing things a certain way,” Knostman said. “To make that leap of faith is a tough thing for people to go through, but since then it has caught on.”
The carrier is also working through legacy and monolithic internal operations that need to be converted to cloud native. Knostman said one of the goals in that process is removing as many barriers as possible for developers.
“You really need to make it as frictionless as possible,” Knostman said. “DevOps hates friction, so you have to make it as easy as possible to get on the platform.”