The effectiveness of anti-spam techniques is an important question: after all, spam has a real cost to legitimate users in terms of time and resources. The problem is how we determine effectiveness, especially for anti-spam techniques that are distributed, and require global scale to function. A scientific approach would suggest that we conduct controlled experiments to evaluate global-scale anti-spam techniques, but that requires controlling the Internet. Or does it? In this paper, we describe a system we have constructed to test global anti-spam techniques that is built on top of the Spamulator, a system that simulates relevant parts of the Internet on a single computer. We demonstrate the ability of our testing system to conduct large-scale experiments with some proof-of-concept experiments on SMTP tarpits and a variation on the Distributed Checksum Clearinghouse.