The high scalability feature of cloud storage systems benefits many companies and organizations. However, most available cloud storage systems lack for providing a full transaction processing support that is really needed by many daily-use applications such as on-line ticket booking. Although a few cloud-based transaction processing systems have been proposed, they achieve barely satisfactory throughput when the conflict-intensive workload is performed. In this context, this paper presents a cloud-based transaction processing system called "TSorter" that uses a conflict-aware scheduling scheme for achieving the high throughput when the conflict-intensive workload is performed. Moreover, Tsorter uses a data caching and an affinity-based scheduling schemes to improve the per-node performance. The experiment results indicate that Tsorter achieves the high throughput, irrespective of the workload types (i.e. the conflict-intensive workload or the conflict-free workload).