LEVERAGING USER BEHAVIOR-BASED SERVICE LEVEL INDICATORS AND OBJECTIVES TO PREVENT AND MITIGATE SOFTWARE INCIDENTS

Abstract

Incidents, defined as unplanned disruptions to software services, can lead to degraded quality and significant losses for organizations. By categorizing incidents based on triggers, type, actors, and impact, valuable insights can be gained to drive improvements. Service Level Indicators (SLIs) and Objectives (SLOs) are key metrics that reflect service health and reliability. However, the current approach of defining SLIs/SLOs in isolation from actual user behavior patterns limits their effectiveness in predicting and preventing user-impacting incidents. This paper proposes a novel approach to defining User Behavior-Based SLIs and SLOs that are directly mapped to key user journeys and features. By cascading these user-centric SLOs down to the underlying microservices and components, more relevant and proactive reliability targets can be set. Potential benefits include fewer incidents, improved Mean Time to Resolution (MTTR), better user experiences, and reduced operational risk. Integrating this approach with machine learning techniques also shows promise for AI-assisted incident forecasting and prevention.