The document presents an adaptive replication scheme for elastic data stream processing designed to optimize scaling decisions and minimize overprovisioning and SLA violations. It highlights the challenges of configuring such systems, enabling fault tolerance through active replication and upstream backup, and using historical data for recovery time estimation. Future work will focus on further formalizing replication strategies and addressing network-bound workloads.