The current cloud approach is show to be very appealing for users, mainly within their pay per use and scalability aspect. But they are not so clearly attractive for the providers, especially in the case of Infrastructure as a Service (IaaS), that need a very efficient way to use their resources, to keep in business under extremely changing conditions. The target of the experiment is to experimentally determine the elasticity requirements for cloud based web applications, that will help providers to comfortably keep within SLA levels, without excessive over-provisioning of resources. To achieve this goal, in a first phase we will stress web applications with different load patterns and different provisioned infrastructures. The results of this first phase will be consolidated, thus creating some scalability policies that will be verified in a second phase, by studying their behaviour under changing loads.