The invention relates to a method and device for scheduling request of Cloud Data Center. The method includes performing the following steps at each timestamp: S1, at the start time of the timestamp, obtaining the queue length for a server in each data center, S2, calculating a first objective function value in each data center, and delivering all service requests in the timestamp to data center with the largest first objective function value, S3, initializing the second objective function value to negative infinity, S4, for each operating frequency in an operating frequency set of each data center, determining optimal operating frequency and optimal operating quantity, S5, for each data center, starting a server of the optimal operating quantity determined in the S4, regulating the operating frequency of the server to the optimal operating frequency determined in the S4. Under the condition of increasing the system benefit, queue length can be reduced, and the tradeoff between system benefit and queue length can be achieved.