View a markdown version of this page

GAMEPERF06-BP02 Categorize and store game data based on access patterns - Games Industry Lens
Implementation guidance

GAMEPERF06-BP02 Categorize and store game data based on access patterns

Categorize your game data into different types based on their access patterns and storage requirements.

Level of risk exposed if this best practice is not established: High

Implementation guidance

Common categories include player data, game saves, persistent world storage, and analytics data.

Implementation steps

Use appropriate storage solutions for each data type to optimize performance and cost-efficiency:

  • Player data: Use Amazon DynamoDB, a fast and scalable NoSQL database, to store player profiles, preferences, and progression data. The low-latency access and automatic scaling capabilities of DynamoDB provide efficient retrieval and update of player data.

  • Game saves: Use Amazon S3 to store game saves and checkpoints. S3 provides high durability and scalability for storing large amounts of game save data. Consider using S3 Transfer Acceleration or Amazon CloudFront for faster uploads and downloads of game saves.

  • Persistent world storage: For games with persistent world states or shared game data, consider using Amazon DynamoDB, Amazon ElastiCache or Amazon MemoryDB. ElastiCache and MemoryDB provide in-memory key-value store while DynamoDB is an SSD backed NoSQL database. These services provide fast access to stored data, reducing the time it takes for the game server process to save game state which improves overall process performance.

  • Analytics data: Use Amazon Managed Streaming for Apache Kafka or Kinesis Data Streams to ingest data streams from your game data producers. Amazon Managed Service for Apache Flink can be used for real-time transformation and analysis and sent to Amazon Data Firehose for processing and delivery into backend data lakes, warehouses and analytics services. Guidance for Game Analytics Pipeline on AWS illustrates how the services work together to provide near real-time and batch analytics.