The answer to this question is dependent upon many things, and is not something that we can just give a number for. The size of your virtual services, the complexity of processing required and the transaction rate that you wish to achieve are all factors that would need to be considered - the answer is therefore very specific to your usage.
The best approach would be for you to put a typical subset workload on the VSE and observe the performance - this could then be used as an indication of your workload and used to scale accordingly