Scaling an ""always-on"" 24/7 continuous video recording product to support thousands of cameras, uploading terabytes of video per day comes with a unique set of challenges. Cassandra is well know for its stability and amazing write performance, but less know for its flexibility. During this talk, I'll discuss the lessons learned during the first 6 months in production. I will compare the actual state of production with initial assumptions around schema design, multi-DC architecture, and capacity planning prior to launch. Together we will walk through issues the team faced, and learn that Cassandra may lack in query flexibility but it makes up for that in operational flexibility. The nature of 24/7 video recording requires high availability. That didn't stop us from modifying the data model or even migrating the entire database to new data centers multiple times, without an outage or planned down time. You'll leave with more insight into ""always on devices"" and how to support them. About the Speaker James Miller Principle Engineer, Comcast James Miller is a Principle Engineer working at Comcast on Xfinity Home security and IoT related projects. He enjoys building and scaling large distributed systems, with a focus on resilience and performance. He has over 15 years of software industry experience, and has been involved in Java and Scala application development using Cassandra for last 4 years.