Data simulation tool - open to bidding

I would like you to create a tool that is capable of simulating requests made to a notional website - including attributes like the users IP, location, form fields, etc. The idea is to build up a corpus of data representing "normative" behaviour for a number of simulated users. Via some stastical process, events will be injected into the data stream that are non-normative - i.e. differ from the standard pattern of activity but some understood level of variance. The ultimate goal of this is to produce a corpus of data that can be used to train and test an unsupervised learning algorithm ( to be developed separately). The system should be capable of producing large volumes of data, representative of large time spans of simulated activity, very quickly More detail on the attributes required (and two more basic sample implementations of this system) can be provided if this project is of interest. Deliverable would be the tool that is able to generate the data, with configuration settings that allow the properties of the generated data to be adjusted to fit different models.