Analytics-Based Solutions for Improving Alert Management Service for Enterprise Systems

Today´s enterprise systems are continuously monitored for timely detection of behavioral anomalies. The tools for monitoring these systems generate alerts on observing abnormal conditions. These alerts are then acted upon by the service desk personnel for timely resolution of the problems. However, there are several drawbacks in today´s alert management service for alert generation and resolution. Present approach of generating and analyzing alerts is highly manual, ad-hoc, and intuition-driven. The fixes are often temporary and ineffective thereby making the system unstable. We propose to replace this manual and intuition-based approach with an automated and analytics led approach. We present algorithms to detect duplicate alerts, infer inter-alert relationships, and derive temporal signature of alerts. We validate the proposed ideas by presenting a real-world case-study.