Have you ever been in a situation where you need to guarantee 100% reliability in a client-server application pipeline? Did the last 0.1% feel impossible? How did you design the system to be resilient against unpredictable failures?

In this talk, I'll do a technical walk-through of how we guarantee at-least-once semantics throughout the LinkedIn instant messaging delivery pipeline. You will learn how to reliably transmit sender actions from the client to the server even under bad network conditions, how to reliably process operations in a durable retry queue in the backend, and how to reliably deliver events to recipient client devices when losing even a single message in the pipeline is unacceptable.