This paper describes a data center energy management system which reduces power consumption of cooling and a novel method for semi-optimal workload placement and cooling. The energy management is formulated as an optimization problem and the optimal workload and cooling are calculated using mathematical programming such as mixed integer programming (MIP), which needs long computation time for large data centers. The computation of cooling control has to be completed within its control interval. Therefore, the semi-optimal solution is used for actual data centers. In this paper, an algorithm for semi-optimal solution is introduced. First, semi-optimal workload placement is calculated based on an equation of cooling power consumption and workload placement. This includes only matrix calculations and sorting. Then, optimal cooling is calculated using the resultant workload placement and liner programming, which is faster than MIP. Simulation results indicate that its computation is fast and accurate enough to apply it to large data centers.