Abstract. Closed-circuit television (CCTV) and Internet protocol (IP) cameras have been applied to a surveillance or monitoring system, from which users can remotely monitor video streams. The system has been employed for many applications such as home surveillance, traffic monitoring, and crime prevention. Currently, cloud computing has been integrated with the video monitoring system for achieving value-added services such as video adjustment, encoding, image/video recognition, and backup services. One of the challenges in this integration is due to the size and geographical scalability problems when video streams are transferred to and retrieved from the cloud services by numerous cameras and users, respectively. Unreliable network connectivity is a major factor that causes the problems. To deal with the scalability problems, this paper proposes a framework designed for a cloud-based video monitoring (CVM) system. In particular, this framework applies two major approaches, namely stream aggregation (SA) and software-defined networking (SDN). The SA approach can reduce the network latency between cameras and cloud services. The SDN approach can achieve the adaptive routing control which improves the network performance. With the SA and SDN approaches applied by the framework, the total latency for transferring video streams can be minimized and the scalability of the CVM system can be significantly enhanced.