Adequate testbeds for conducting security experiments and test under controlled, safe, repeatable and asrealistic- as-possible conditions, are a key element for the research and development of adequate security solutions and the training of security personnel and researchers. In this paper, we report on the construction and operations of isolated virtualised testbeds used in two separate security research labs in Canada and France, as part of a joint collaborative effort. The main idea was to use mid- to large-scale isolated computing clusters to obtain high levels of scale, manageability and safety by heavily leveraging virtualisation technology, open-source cluster management tools and a network architecture separating experiment and control traffic. Both facilities have been used for conducting different types of security research experiments, including in-lab reconstructions of botnets, denial-of-service attacks, and virus detection experimentation. They have also been used for teaching and training students in experimental security methods. We describe these facilities and the criteria that we used to design them, the research and training activities that were conducted, and close by discussing the lessons learned and the pros and cons of this approach.