As part of the Cloud journey, New Relic implemented the cell architecture pattern: a cell is a self-contained installation with a limited lifespan that can satisfy all the operations for a range of users in a specific domain. It's like microservices but for data in motion services.

This architecture provides isolation, cloud cost reduction, and the ability to deploy in different regions or cloud providers, test upgrades, implement rolling upgrades and test different software versions.

But it also introduces challenges. How do we replicate data between cells if they can be created or decommissioned at any moment? How can the replicator react to changes in the traffic flow? In this talk, we'll cover our initial approach to this problem, what we learn in the process and how we plan to improve it in the future.