Abstract: Smart Grid data processing system needs to store the large-scale data collected from the mass device endpoint periodically,handle the high-concurrent read and write requests,and the high-scalability is more important.Cloud-based data serving systems provide a good choice for us.However the large variety of cloud-based data serving systems and the various performance make it difficult for developers to choose an appropriate system.Here we propose a method to benchmark the performance of cloud-based data serving systems based on the special scene of smart grid.Finally we benchmark the performance of HBase,Voldemort and Cassandra through our method.