Instead of focussing on the volume/size you have to size you MARS appliance using the approximate number of EPS (Events per second) and the no. of Netflows per second.
You can get a sizing document from the Cisco Account team for this.
Once you have the box running you can adopt archiving, tuning etc. to ensure all data is 'avaialble' when required as per the compliance requirements.
Regards
Farrukh
