do you mean sizing in the meaning of the hardware necessary to manage the server efficiently? i.e. RAM, processor speed, disk architecture etc? obviously the storage size itself is a trivial exercise.
for the rest, it's pretty much the same issue as any hardware sizing exercise. you need to make traffic estimates; provide app information on the likely mix of inserts, other non-idempotent transactions and selects. information on your network architecture is also relevant.
then you need to consider redundancy requirements and thus whether you should be running a master-slave configuration or using master-master replication or other forms of clustering. sometimes clustering is useful not just for redundancy but simply for performance reasons.