In recent cloud systems, data-intensive applications have been widely deployed for solving non-trivial problems, which often involve large amount of dataset. As a standard service for offering data availability and reliability, data replication service plays an important role and is proven to be effective in many real-world cloud environments. However, many existing replication services are designed for improving the system-oriented metrics instead of user-oriented metrics. In this paper, we first formulate the quality-of-service (QoS) constrained replication placing problem in different manners and then a novel replication service is proposed which uses multi-objective technique to solve the replica placement problem via a time-efficient approach. A set of experiments are conducted to evaluate the effectiveness and efficiency of the proposed service and the results indicate it can significantly reduce the average response time metric comparing with other existing replication services and therefore improve the QoS for user applications. In addition, it also exhibits a good robustness when the cloud storage system is facing intensive replication workloads.