Data center power management is evolving from ad hoc methods based on maximum node power usage to systematic methods that employ power-scalable components. In addition, it is possible to exploit the power and throughput relationship to increase the total work performed and safely overprovision the rack space while staying below an aggregate power limit. This research describes a general framework for boosting throughput at a local level while load-balancing the available aggregate power under a set of operating constraints. Our solution is useful for those data centers that cannot expand the number of power circuits or seek effective usage of their available power budget due to unplanned power fluctuations. The framework is particularly well suited for environments with a heterogeneous workload and hence, a non-uniformpower allocation requirement. Based on a representative workload for a two minute period, this paper shows a non-uniform power allocation scheme increases throughput by over 16% versus a uniform power allocation mechanism.