With the rapid development of information technology, the Internet of Things (IoT) is getting more attentions, which has promoted a new wave of information and industrial tide. In order to reduce the CAPEX, OPEX and TIEX of IoT systems, before the deployment, people often evaluate IoT architectures, protocols, as well as configurations on testbeds. As physical resources of any testbed are limited, it is challenging to conduct large-scale IoT experiments. In this paper, we apply the virtualization technology to optimize an existing network testbed, adopt VMNet to emulate wireless sensors in the IoT experiment, and improve the scalability of the experiment to support large-scale IoT assessments. Our scheme leverages a multi-host collaborative with static multi-sink architecture to solve the bottleneck problem of the largescale IoT emulation and multiple interpolation algorithms to supplement the time continuity and spatial integrity of the sensed data for enhanced fidelity of the IoT experiment. According to our experiments, the virtualized IoT testbed not only reduced the TIEX of IoT emulation sharply, but also enhanced the scalability of IoT experiments.