Detecting abnormalities in time-series data from an online professional network

摘要

The disclosed embodiments relate to a system for detecting abnormalities in time-series performance data obtained from machines that implement an online professional network. During operation, the system receives the time-series data, including throughput measurements and/or latency measurements for requests made to back-end systems associated with the online professional network. Next, the system attempts to detect abnormalities in the time-series data. If such an abnormality is detected, the system looks up associated system metrics, which are temporally proximate to the abnormality. The system then generates a notification about the abnormality along with the associated system metrics to facilitate determining a root cause of the abnormality.