Redis as a scalable feature store

By Zohaib Sibte Hassan, Arbaz Khan

In this session we discuss various techniques we use at DoorDash to build and improve our tail latencies in our feature store. With billions of requests coming in, it is imperative for us to squeeze every last bit of performance out of Redis. We share our best practices on serialization, hot keys, stampede (Thundering Herd), predictive cache expiry, compression, and other optimization tips.