Candy Crush maker King in Hadoop cluster move to Google Cloud

Candy Crush online game-maker King is moving a sizeable portion of its on-premise, big data processing capability to the Google public cloud. The migration is earmarked for completion in early 2019.

The company, whose games are reportedly played by more than 270 million people around the world, claims to operate one of Europe’s largest Hadoop clusters, which its data scientists and engineers use to process “double-digit” petabytes of data generated by players of its games.

Åsa Bredin, first vice-president of technology at King, told Computer Weekly that analysing this data plays a critical role in ensuring the company can stay one step ahead of changes in player behaviour.

“One of the challenges in our market is speed, both in terms of being able to act quickly on player feedback, but also for us to scale as the number of players and games grows,” she said.

“It is very hard to know exactly what lies ahead, so by being able to act quickly on insights, we can also handle challenges better.”

For this reason, the company has spent the last couple of years evaluating whether continuing to run an on-premise Hadoop cluster is the best way to achieve the rapid insights it requires, particularly in view of the advancing big data capabilities that the public cloud provider community has to offer, she said.

“As the cloud ecosystem grows and more options are ‘put on the shelf’, we expected that innovation in machine learning, query processing and speed would appear on the cloud providers first,” said Bredin.

So the company began weighing up its public cloud options, and decided the Google Cloud Platform would be the best fit for its current requirements. The firm’s burgeoning portfolio of machine learning tools was also an important supplier consideration.

“We will continue to be a very data-driven company where game data is used continuously to make the games even more fun to play,” she said.

“One area where we expect to see a lot of innovation in mobile gaming is around our relevance. With machine learning, we predict offers and suggestions for what to tackle next and what will become increasingly relevant.”

The company’s Hadoop migration began in early 2018 and is intended to be complete by the end of the year, said Bredin. “A substantial part of the business is using Google Cloud Platform for its data now. By the beginning of next year, our intention is that all data will be in it.”

At that point, King’s “core data infrastructure” will be running in the public cloud, but the firm is still deciding whether more of its estate is likely to follow. “We have run smaller systems in cloud solutions before, but this move is by far the largest,” said Bredin.

“We always look at ways of running things faster and smarter, but don’t have any set plans for any other migrations right now.”

Sunil Rayan, managing director of gaming at Google Cloud, said the project should put King on a path to provide “even richer gaming experiences” for its global user base  as it reaps the benefits of having access to Google Cloud’s data processing capabilities.

“King has long been a leader in delivering delightful mobile gaming experiences to millions of players around the world,” said Rayan.

“It will continue to innovate and demonstrate its leadership position as a global innovator by utilising our big data, AI and machine learning capabilities to give its engineers the next generation of tools to build great experiences.”