Concurrent job execution is a hot topic in large-scale resource scheduling research. Existing efforts employ queueing model with local optimal solution to schedule co-located tasks, thus can only fit specific requirement. Hence, how to design a single scheduler to meet diverse requirements is challenging. This paper introduces Sirius, a new framework for resource scheduling based on minimum cost maximum flow network. This new approach makes it easy to express scheduling requirements, including fairness, priority and placement constraint, on a unified way as a typical graph construction and solution problem. Meanwhile, an incremental algorithm is implemented to speed up the flow network solver, significantly reducing its runtime by 90 percent.