Culturing Resiliency with Data: A Taxonomy of Outages

Explore Uber's Approach to Outage Categorization and Incident Management. This presentation offers a comprehensive look at how Uber systematically categorizes past outages based on their underlying root causes.

We'll begin with essential context, covering key definitions, the incident management framework employed, and established best practices for preventive techniques. Understand the structure behind our categorization system, including individual categories and sub-categories, and their distribution.

A deep dive will then focus on two significant outage categories: deployment issues and capacity problems. Learn how time series data mining techniques are used to improve outage detection and simulation of common root causes within these areas.

Finally, we'll examine how insights gained from these analyses are translated into tangible improvements through policy changes and process improvements, ensuring greater system reliability. This session provides valuable insights into system reliability, incident response, and operational excellence within a large-scale technology organization.