In this talk we introduce Apache Aurora, a framework built on top of Apache Mesos with the goal of providing a general-purpose process scheduler.  Aurora has been in development at Twitter for about four years (growing alongside Mesos), and powers a large portion of Twitter's services and ad-hoc processes.  In November 2013, Aurora was accepted into the Apache Incubator program.

Apache Aurora is self-described as a service scheduler, which means that it aims to reliably run many replicas of a serving system.  Additionally, it encourages decoupling service management from machine management, allowing a relatively small number of administrators to manage very large fleets of machines.

We will present a walkthrough of Aurora from a user's perspective and describe some forthcoming features on the project roadmap.