It is tempting to simply do 100x30x5. Well, given the consequences of out of disk (often causing system outage and shared storage has become single point of failure for HA), I would suggest run 'realistic' load tests with your cluster with HA not enabled. Get the maximum disk space used among all nodes, then times that with number of nodes and then time a factor of 2 to 3. Storage these days is fairly inexpensive but a full disk space will cost you more in terms of business loss due to system outage. You should be aware that starting 7.1.9, database will be the default passivation storage if HA is enabled, so you do not have to worry about that anymore.