For nearly a week, a hospital’s core systems were exceptionally slow, disrupting critical operations and putting the internal IT team under intense pressure to find answers fast. At first glance, the issue looked like a typical “server overload” situation.
But instead of focusing only on the current state of the environment, our team went back in time to understand when the abnormal behavior first started.
By analyzing performance trends over time, we identified a very clear turning point where several metrics began spiking simultaneously:
- Sharp rise in Disk Read / Write activity
- Disk utilization jumped from a normal range of 30–40% to nearly 90%
- Clear spikes in Max CPU, Max Disk, and Max IO
At around the same time, while reviewing database performance metrics, a significant increase in Average CPU Waits accompanied a sharp rise in transaction volume, indicating growing pressure on the database workload and a direct correlation between system contention and increased processing activity.
Excessive CPU load caused by inefficient function usage
Going deeper into what was overloading the database, a heavy use of database functions was identified as a major contributor to CPU pressure.
On a server with 10 CPU cores, the workload generated by these functions created significant contention, leading to noticeable system slowness experienced by end users in milliseconds.
A key observation was that the function calls were executed directly within SELECT statements, rather than being embedded or optimized through indirect usage patterns. This resulted in unnecessary repeated execution and increased CPU strain.
