Title

AWS re:Invent 2023 - Exiting the Data Center: A League of Legends and VALORANT Story (GAM304)

Summary

• Presenters: Hadrian Baran (AWS Technical Account Manager), Aditya Kumar (Senior Network Engineer at Riot Games), and Rory Power (Infrastructure Engineer at Riot Games).
• Topic: The session covered Riot Games' extensive cloud migration journey, moving their popular games League of Legends and VALORANT from data centers to AWS cloud infrastructure.
• Key Points:
  • Riot Games' growth from a single game company to multiple titles and the need for global infrastructure.
  • The architecture of League of Legends and the challenges of maintaining low latency for a good player experience.
  • The evolution of Riot's infrastructure from data centers to cloud, including the development of their container engine (Rcluster) and the migration to AWS.
  • The launch of new games like Teamfight Tactics (TFT) and VALORANT, which were cloud-first and required a combination of AWS regions, outposts, and Riot's own ISP (Riot Direct) to meet latency requirements.
  • The migration of League of Legends to the cloud, including the development of a release train for automated deployments and the creation of new shards.
  • The partnership with AWS, including the use of AWS professional services and the importance of infrastructure as code and elasticity of AWS compute.
  • The decommissioning of data centers and the benefits realized from the migration, such as improved player experience, operational efficiency, and strategic partnership with AWS.

Insights

• Technical Investments: Riot's early investments in Terraform and containerization (Rcluster) were crucial for a smooth transition to the cloud, highlighting the importance of infrastructure as code and abstraction of infrastructure from developers.
• Player Experience: Throughout the migration, Riot maintained a player-first approach, ensuring that game servers were live, new content was delivered, and player experience was not compromised.
• Global Reach: The migration to AWS allowed Riot to scale globally, reaching more players and standardizing gameplay modes across different regions.
• Strategic Partnership: The collaboration between Riot and AWS went beyond a vendor-client relationship, with AWS advocating for Riot's needs internally and collaborating on new features and infrastructure decisions.
• Operational Efficiency: The cloud migration led to operational efficiencies such as reduced time to build new shards, better handling of DDoS attacks, and no longer needing to manage physical data center infrastructure.
• Continuous Improvement: Riot's approach to continuous improvement, including live rollbacks, refinement of migration steps, and post-migration optimizations, was key to the success of the project.
• Cost Optimization: Post-migration efforts included optimizing cloud resources, such as instance types and auto-scaling, to improve the total cost of ownership.
• Decommissioning Data Centers: The final stages of the project involved the physical decommissioning of data centers, including the shredding of hard drives, signifying the end of Riot's reliance on physical infrastructure.