Typically AirWave is licensed based on the number of network infrastructure devices monitored. From there, if you're looking to focus on a particular portion like VisualRF, you can allocate memory specifically for that feature.
Determining the amount of memory to allocate then depends on the customer deployment. How many switches, controllers, APs are in the customer environment? Of the devices, what brand are the devices (this will help take into account devices that are slow to respond)? How many wireless clients do they typically see at peak hours? What's the behavior of these clients (do they move a lot, do they show up everyday, do they go between floors often)? What type of deployment is the customer environment (office, education, airport, hotspot, storefront)? Is the customers deployment in multiple locations or a single location? How many floor plans would they be trying to track? These factors will help determine the kind of server resources needed to calculate the location proficiently (if location is their main goal).
Another thing to consider is if the deployment is setup for getting better location calculation? If it's not, is customer planning to add/move APs in order to facilitate better location mapping?
You should collect the answers to the above and then work with a sales rep or partner reseller who can help provide additional guidance.