PGDay Lowlands2026

User-generated content (UGC) systems often become critical bottlenecks as they scale. What happens when a high-traffic, read-heavy service starts to struggle under growing load, technical debt, and operational complexity?

This talk explores how a large-scale reviews system evolved to handle hundreds of millions of daily requests more efficiently. Over time, inefficient database access patterns, excessive JOINs, and overloaded responsibilities led to rising latency, SLO breaches, and on-call strain.

Instead of a full rewrite, we’ll walk through pragmatic, incremental improvements:

Reducing database complexity through targeted denormalization of frequently accessed fields Simplifying read paths and removing unnecessary caching layers Converting expensive O(n) computations into O(1) through precomputation Achieving consistent low-latency responses at scale

We’ll cover how these changes were rolled out safely with zero downtime, how consistency was maintained, and the trade-offs involved.

If you’re dealing with a read-heavy system under pressure, this talk offers practical strategies to improve performance and scalability without starting from scratch.