Sony Interactive Entertainment Unifies Platform on EKS, achieving 5x Faster Deployments & 60% Lower Operational Costs

The launch of PlayStation 5 in November 2020—in the middle of the COVID-19 pandemic—created an increased demand for gaming. In addition to the continued support of the existing PlayStation 4, SIE was challenged with supporting multiple consoles while consistently rolling out new services and updates. Moreover, SIE’s global user base wanted extremely low latency in their game experiences.

Various departments across the global organization at that time managed a mix of AWS services, self-managed Kubernetes, and on-premises infrastructure. Different teams used different tools, programming languages, and schedules to handle diverse workloads, from game servers to enterprise applications. “Our global teams were constantly racing against time to deliver a great experience to end users, but maintaining system reliability was an issue because of inconsistent infrastructure,” says Chris “Mac” McEniry, a principal software engineer at SIE. “Tracking down problems, identifying responsible teams, and understanding the context of issues took time away from our focus in offering great experiences to our users.” That fragmented approach also caused difficulties in global governance and security.

“At SIE, we put our gaming community first, and the mandate from our leadership was clear: to give gamers the best play experience possible,” says McEniry. “We knew that we had to be innovative and resourceful to achieve our goals.”

SIE aimed to increase deployment speed and improve the platform experience. “We needed a solution where everyone could align without making too many changes—where if the effort is high, the impact is high too,” says Swathi Somanchi, senior director of engineering at SIE.

That determination led to the creation of an internal development platform called Platform Unification, centered around a cohesive runtime called Unified Kubernetes Service (UKS). UKS was built on Amazon Elastic Kubernetes Service (Amazon EKS), a managed Kubernetes service (see figure 1). UKS improved access management, observability, cluster management, troubleshooting, advanced deployment strategies, and automated upgrades across all PlayStation Network backend services. By creating UKS, SIE aimed to facilitate the standardization of environments and systems, help various teams operate more efficiently, and empower senior developers to focus on their core coding and product development responsibilities.
 

Figure 1: Layered architecture

For cost-effectiveness, SIE transitioned workloads to the Platform Unification infrastructure of UKS on AWS, reducing operational costs by 60 percent. “Sony Interactive Entertainment, with support from Amazon EKS, has efficiently managed compute resources and provided developers their preferred tools,” says Somanchi. “This has removed common obstacles, accelerated kernel and node updates, and enhanced the overall developer experience.”

SIE developed its initial unified platform over 6 months, with support from AWS solutions architects and subject matter experts, and then migrated services in phases. The team held regular scrum stand-ups and workshops to verify global alignment and adherence to best practices. “In the workshops, we gathered together the engineers responsible for the services,” says Somanchi. “We successfully onboarded 40 microservices to a nonproduction environment in 2 days, which is a major milestone.”

Platform Unification is a centralized hosting environment where developers deploy, update, and run their microservices (see figure 2). It has streamlined developer operations with its unified interface, cutting training time from 16 hours to 3 hours and boosting deployment speed. “Previously, team members juggled different setups, but now they handle a single, consistent environment, simplifying their workflow,” says Somanchi.
 

Figure 2: User interaction architecture

Application management variability has been significantly reduced, resulting in virtually 100 percent standardization and simplified troubleshooting. “SIE’s migration to Amazon EKS and standardization of application environments has removed the cognitive load of managing multiple configurations,” says McEniry.

A unified registry management was also deployed with support from Amazon Elastic Container Registry (Amazon ECR), a fully managed container registry.

Additionally, with Platform Unification, the company has transitioned from diverse security solutions to a unified and automated solution. SIE also uses AWS Secrets Manager—a service that helps companies manage access to their applications, services, and IT resources—to retrieve and rotate credentials and other assets. The platform engineering team developed a continuous integration and continuous delivery pipeline for standardized testing and security scans across regions. There is uniform integration and consistent enforcement of the security tools for all application deployments and uniform runtime integration with the security tools. Previously, there were four runtime integrations at different levels of completeness; only one integration is needed now. “The incorporation of security and operations into the pipeline has made everything more efficient,” says McEniry. The automation also provides developers with all the information they need for deployment.

Platform unification has thus helped SIE reduce redundant compute costs by 10 percent and redundant operational time by 50 percent. The reduction in the number of implementations from four to one has also helped reduce operational costs by 60 percent.