The field of smart fuzzing has opened up serious bug hunting to a whole new generation of hackers. Tools like AFL allow for speedy discovery of once subtle parsing bugs in software, often leading to serious exploitation opportunities. While AFL’s genetic algorithm based approach is fast for what it is, ultimately the fuzzing of such software is slower than it needs to be for quick software validation.In this talk John Dunlap will present a method for building a small scale compute cluster oriented toward large scale smart fuzzing on a home budget. John will present methods for converting performance concepts normally reserved for scientific computing applications into practical “fuzz farm” techniques. Topics such as high speed multithreading, vectorization, process management, and cluster node management will be discussed in a manner friendly to those new to scientific computing.A basic fuzzer management program will also be presented.