90% of performance is data access patterns

A recent analysis highlighted that 90% of application performance issues stem from data access patterns. A platform team faced a significant performance bug linked to excessive API requests, particularly from a session middleware that fetched patient appointment histories unnecessarily. Initially, this caching method was effective when user numbers were low, but as the application evolved into a more complex single-page application, the inefficiencies became apparent. The middleware was causing a bottleneck by repeatedly fetching appointment data, leading to increased CPU usage and latency.

Upon investigation, it was discovered that the client application no longer required this data to be fetched on every request, as it had its own method for retrieving appointments. By removing a single line of code that triggered the redundant fetching, the team reduced the number of daily calls from 150,000 to 1,000, resulting in a 66% drop in CPU usage and a significant improvement in latency across all API requests.

This experience underscores the importance of observability in software development, as it allows teams to identify and address performance issues before they escalate into outages. The lesson learned is that while code may have been appropriate at the time of writing, changes in the environment necessitate ongoing evaluation and adaptation to maintain performance.