Author: Jan Michael Go Tan

When clients start running queries on new Amazon Aurora replicas, they will notice a longer runtime for the first few times that queries are run; this is due to the cold cache of the replica. As the database runs more queries, the cache gets populated and the clients notice faster runtimes. In this post, we focus on how to address the cold cache so clients that are connecting through a load-balanced endpoint get a consistent experience regardless of whether the replicas are automatically or manually scaled. In addition, we also look at other caching solutions such as Amazon ElastiCache, a fully managed Memcached, Redis, and Valkey compatible service, that can further improve the overall experience for latency-sensitive applications and, in some situations (such as higher cache hits), lead to less frequent auto-scaling events of the Aurora read replicas.

April 2024: This post was reviewed and updated for accuracy. Gaming leaderboards enable players to gauge their performance against each other. This important social feature increases players’ engagement and encourages competition. Leaderboard data can also inform in-game algorithms that match players against competitors of similar skill levels. In this post, I explore the challenges around […]