GAME FREAK inc. has produced many hit titles, mainly for home video game consoles. In recent years, the company has also been involved in the development of smartphone and PC games.

The company is unique in that it produces world-class game titles with a small group of elite professionals. The latest game consoles and game systems have evolved remarkably, and game development is becoming increasingly large-scale. As a result, on-premise systems were increasingly suffering from a lack of resources. Therefore, the development environment had a significant impact on maintenance and other operational burdens.
"The data center's resources were also limited, and lack of rack space and power were also major issues," recalls Haruki Tachihara, Manager of Research and Development Dept. “Especially after the COVID-19 pandemic, it is difficult to even procure machines even if one wants to augment the system. We needed a highly scalable infrastructure that could be operated by a small number of people, had minimal physical limitations, and could respond flexibly to changes in the environment."

This is why they focused on the cloud. The build environment, in particular, was often noticeably under-resourced as the project progressed, and expansion of the CI/CD environment, including program compilation and performance checks, was seen as a challenge. In addition, these tasks were often performed on the developer's work terminal, and individual settings often affected CI/CD. By moving the build environment to the cloud, you can respond to increasing or decreasing needs without worrying about machine procurement, and you can integrate CI/CD.

GAME FREAK inc. chose AWS from among a number of cloud services. The high availability and flexibility of the managed service, which requires no detailed setup, reduces the operational load, and the high cost effectiveness was attractive. The GitLab community that the company uses is very active, and there is a wealth of know-how and documentation on GitLab integration with AWS.

GAME FREAK inc. builds its environment on Amazon EC2 and uses it to build some of its programs. Because game development involves updating programs on a daily basis, it uses EC2 Auto Scaling Groups to ensure that the capacity is appropriate for the scale of the project at any given time, and it also applies Amazon EC2 spot instances to optimize costs.

Traditionally, most game builds are run in a Windows environment, but the company is leveraging the Amazon Elastic Kubernetes Service (Amazon EKS) and also using a cost-effective Linux environment.
　"Initially, we started utilizing AWS on our own, but we wanted to get some expert pointers on the core aspects, so we consulted with the AWS account team. They gave us a lot of practical advice, such as how to combine Auto Scaling Groups and spot instances suitable for game development, how to use Savings Plans, and the best way to store data, which helped us optimize costs and strengthen compliance," says Reona Takayama of the Research and Development Department's Operations Management Team.

GAME FREAK inc. constructs not only the build environment but also the data analysis infrastructure essential for game development on AWS. As mentioned above, the evolution of games and the accompanying increase in the scale of game development have created points that cannot be improved by human test play alone. With Amazon OpenSearch Service at its core, the company is committed to continuous game improvement through in-depth analysis of game data and play logs.
"We were able to migrate smoothly from our on-premise analysis infrastructure, and by integrating with Amazon CloudWatch Logs, we were able to build an environment that is easy to operate. Every day, we receive hundreds of thousands of requests from within the company, which are processed in real time using Amazon Kinesis Data Stream, AWS Lambda, and other integrations. It is extremely difficult to build such a large scale analysis environment in-house. With AWS managed services, it is relatively easy to enhance the many services we operate in-house," said Mr. Takayama.

The company also provides other development support functions on AWS, including AWS CodeArtifact for sharing and distributing in-house libraries, AWS Lambda for batch processing, and a proprietary storage service that had to be developed to meet compliance requirements, using a combination of Amazon Elastic Container Service (Amazon ECS), AWS Fargate, and Amazon Elastic File System (Amazon EFS).

By converting the game development environment to AWS, flexibility and availability have been greatly improved. In particular, the procurement of development machines, which used to take two months in the COVID-19 pandemic, can now be prepared in about one hour with AWS. No matter how many times your project data grows, you don't have to worry about running out of capacity in the cloud.
“A few years ago, before AWS, we had about 30 machines running a single title development project. The latest title required 50 machines, 70% of which eventually ran on AWS. We are no longer plagued by machine failures, and the operation of our build environment has become very stable. AWS' advice also extends to security, pointing out specific areas for improvement, such as how to write S3 bucket policies. We are in the process of implementing safe operation initiatives, including data protection in the cloud," said Tachihara.

By reducing the operational burden with AWS, GAME FREAK inc. can now focus its human resources on game development. Tachihara says, "The biggest advantage is that we were able to support a large project even with a small, elite environmental team," which would not have been possible with an on-premise system. Another important effect was that they were able to support game development while the pandemic limited their attendance at work.

GAME FREAK inc. plans to continue to promote multi-platform support for home video game consoles and smartphones. The company is in the process of considering converting our on-premise test play environment to AWS as one of the issues. There are also plans to utilize AI/machine learning as a new system, and efforts are underway to leverage AWS AI services, including Amazon SageMaker.
"As games continue to evolve, development projects are expected to grow in scale. Therefore, we are looking at services such as Amazon WorkSpaces in order to continue our telecommuting system and to develop our employees' freedom without being tied to where they work or where they sit. We want to make the most of AWS to create even more fun games," says Mr. Tachihara.

Return to the list of customer case studies in Japan »