P90 is a statistical term commonly used in observability to understand the distribution of data. For example, if we measure the response times of a service and calculate the P90, it tells us that 90% of the response times are below this value. This is useful for understanding the worst-case performance scenario in a system.
If the P90 of the request latencies in your application is 300ms, it means that 90% of requests completed with a latency below 300ms, and 10% were above. With large datasets, it is recommended to measure worse case scenarios using P90 or other percentiles, as relying on the maximum latency can be misleading due to outliers.
Explore related concepts
Percentile
A percentile is a measure used in statistics to indicate the value below which a given percentage of observations in a group of observations fall. In the context of observability, percentiles are often used to measure the distribution of response times and latencies.
Latency
Latency is the delay between a request being made and a response being received. It is a key metric as it directly impacts user experience and system performance.
Observability
Observability is the ability to understand how your application is working and behaving in production through telemetry data (logs, metrics, traces, wide events, etc.). It enables you detect, diagnose and resolve issues in your app before they impact your users and become problems.