Cloud computing is a major trend in distributed computing environments. Software virtualization technologies allow cloud Infrastructure-as-a-Service (IaaS) providers to instantiate and run a large number of virtual appliances. However, one of the major challenges is to reduce the disk space footprint of such virtual appliances to improve their storage and transfer across cloud servers. In this paper, we propose to use a Software Product Line (SPL) approach and describe the virtual appliance as a set of common and variable elements modeled by means of Feature Model (FM). We describe a solution to reverse engineer a FM from a virtual appliance and we show how we take advantage of the SPL configuration mechanisms to signifi cantly reduce the size of a virtual appliance.