Stochastic processing networks (SPNs) model the operations of many complex systems, such as data centers and communication networks. These networks have been an active subject of research for more than 40 years. In this talk, Dai will introduce M-COF (Multi-scale Closed-form Optimization Framework), a new framework for the optimal control of a class of SPNs known as multi-class queueing networks. M-COF is scalable, can incorporate practical performance metrics such as tail-latency or fairness, and is rooted in recently developed multi-scale heavy-traffic theory. He will also discuss the relationship between M-COF (simulation-free) and recent methods (with simulation) based on deep PDEs and deep reinforcement learning. This talk is based on joint work with Jin Guang (Chicago Booth) and Lucy Huo (HKUST).