This document discusses dynamic weighted round robin (DWRR) algorithms designed for CPU provisioning in cloud-based environments to ensure service differentiation across server clusters. It demonstrates that these algorithms can meet service level agreements (SLAs) without needing detailed statistical knowledge about system demands, even under overloaded conditions. Additionally, the evaluation of these policies through simulations provides insights into their effectiveness and operational trade-offs.